fix(ci): dump server/container logs when tests fail (#3873)

Output last 100 lines of server.log or docker container logs when
integration tests fail to aid debugging.

.github/actions/run-and-record-tests/action.yml

@@ -82,11 +82,13 @@ runs:
           echo "No recording changes"
         fi
 
-    - name: Write inference logs to file
+    - name: Write docker logs to file
       if: ${{ always() }}
       shell: bash
       run: |
-        sudo docker logs ollama > ollama-${{ inputs.inference-mode }}.log || true
+        # Ollama logs (if ollama container exists)
+        sudo docker logs ollama > ollama-${{ inputs.inference-mode }}.log 2>&1 || true
+        # Note: distro container logs are now dumped in integration-tests.sh before container is removed
 
     - name: Upload logs
       if: ${{ always() }}

.github/workflows/integration-auth-tests.yml

@@ -73,6 +73,24 @@ jobs:
           image_name: kube
           apis: []
           providers: {}
+          storage:
+            backends:
+              kv_default:
+                type: kv_sqlite
+                db_path: $run_dir/kvstore.db
+              sql_default:
+                type: sql_sqlite
+                db_path: $run_dir/sql_store.db
+            stores:
+              metadata:
+                namespace: registry
+                backend: kv_default
+              inference:
+                table_name: inference_store
+                backend: sql_default
+              conversations:
+                table_name: openai_conversations
+                backend: sql_default
           server:
             port: 8321
           EOF

.github/workflows/integration-vector-io-tests.yml

@@ -169,9 +169,7 @@ jobs:
         run: |
           uv run --no-sync \
             pytest -sv --stack-config="files=inline::localfs,inference=inline::sentence-transformers,vector_io=${{ matrix.vector-io-provider }}" \
-            tests/integration/vector_io \
-            --embedding-model inline::sentence-transformers/nomic-ai/nomic-embed-text-v1.5 \
-            --embedding-dimension 768
+            tests/integration/vector_io
 
       - name: Check Storage and Memory Available After Tests
         if: ${{ always() }}

.github/workflows/pre-commit.yml

@@ -37,7 +37,7 @@ jobs:
             .pre-commit-config.yaml
 
       - name: Set up Node.js
-        uses: actions/setup-node@a0853c24544627f65ddf259abe73b1d18a591444 # v5.0.0
+        uses: actions/setup-node@2028fbc5c25fe9cf00d9f06a71cc4710d4507903 # v6.0.0
         with:
           node-version: '20'
           cache: 'npm'

.github/workflows/precommit-trigger.yml

@@ -99,7 +99,7 @@ jobs:
               owner: context.repo.owner,
               repo: context.repo.repo,
               issue_number: ${{ steps.check_author.outputs.pr_number }},
-              body: `⏳ Running pre-commit hooks on PR #${{ steps.check_author.outputs.pr_number }}...`
+              body: `⏳ Running [pre-commit hooks](https://github.com/${context.repo.owner}/${context.repo.repo}/actions/runs/${context.runId}) on PR #${{ steps.check_author.outputs.pr_number }}...`
             });
 
       - name: Checkout PR branch (same-repo)
@@ -141,7 +141,7 @@ jobs:
 
       - name: Set up Node.js
         if: steps.check_author.outputs.authorized == 'true'
-        uses: actions/setup-node@a0853c24544627f65ddf259abe73b1d18a591444 # v5.0.0
+        uses: actions/setup-node@2028fbc5c25fe9cf00d9f06a71cc4710d4507903 # v6.0.0
         with:
           node-version: '20'
           cache: 'npm'

.github/workflows/providers-list-deps.yml

@@ -36,7 +36,7 @@ jobs:
       distros: ${{ steps.set-matrix.outputs.distros }}
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
 
       - name: Generate Distribution List
         id: set-matrix
@@ -55,7 +55,7 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
 
       - name: Install dependencies
         uses: ./.github/actions/setup-runner
@@ -79,7 +79,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
 
       - name: Install dependencies
         uses: ./.github/actions/setup-runner
@@ -92,7 +92,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
 
       - name: Install dependencies
         uses: ./.github/actions/setup-runner

.github/workflows/python-build-test.yml

@@ -24,7 +24,7 @@ jobs:
       uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
 
     - name: Install uv
-      uses: astral-sh/setup-uv@eb1897b8dc4b5d5bfe39a428a8f2304605e0983c # v7.0.0
+      uses: astral-sh/setup-uv@3259c6206f993105e3a61b142c2d97bf4b9ef83d # v7.1.0
       with:
         python-version: ${{ matrix.python-version }}
         activate-environment: true

.github/workflows/ui-unit-tests.yml

@@ -29,7 +29,7 @@ jobs:
         uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
 
       - name: Setup Node.js
-        uses: actions/setup-node@a0853c24544627f65ddf259abe73b1d18a591444 # v5.0.0
+        uses: actions/setup-node@2028fbc5c25fe9cf00d9f06a71cc4710d4507903 # v6.0.0
         with:
           node-version: ${{ matrix.node-version }}
           cache: 'npm'

benchmarking/k8s-benchmark/stack-configmap.yaml

@@ -98,21 +98,30 @@ data:
       - provider_id: model-context-protocol
         provider_type: remote::model-context-protocol
         config: {}
-    metadata_store:
-      type: postgres
-      host: ${env.POSTGRES_HOST:=localhost}
-      port: ${env.POSTGRES_PORT:=5432}
-      db: ${env.POSTGRES_DB:=llamastack}
-      user: ${env.POSTGRES_USER:=llamastack}
-      password: ${env.POSTGRES_PASSWORD:=llamastack}
-      table_name: llamastack_kvstore
-    inference_store:
-      type: postgres
-      host: ${env.POSTGRES_HOST:=localhost}
-      port: ${env.POSTGRES_PORT:=5432}
-      db: ${env.POSTGRES_DB:=llamastack}
-      user: ${env.POSTGRES_USER:=llamastack}
-      password: ${env.POSTGRES_PASSWORD:=llamastack}
+    storage:
+      backends:
+        kv_default:
+          type: kv_postgres
+          host: ${env.POSTGRES_HOST:=localhost}
+          port: ${env.POSTGRES_PORT:=5432}
+          db: ${env.POSTGRES_DB:=llamastack}
+          user: ${env.POSTGRES_USER:=llamastack}
+          password: ${env.POSTGRES_PASSWORD:=llamastack}
+          table_name: ${env.POSTGRES_TABLE_NAME:=llamastack_kvstore}
+        sql_default:
+          type: sql_postgres
+          host: ${env.POSTGRES_HOST:=localhost}
+          port: ${env.POSTGRES_PORT:=5432}
+          db: ${env.POSTGRES_DB:=llamastack}
+          user: ${env.POSTGRES_USER:=llamastack}
+          password: ${env.POSTGRES_PASSWORD:=llamastack}
+      references:
+        metadata:
+          backend: kv_default
+          namespace: registry
+        inference:
+          backend: sql_default
+          table_name: inference_store
     models:
     - metadata:
         embedding_dimension: 768
@@ -137,5 +146,4 @@ data:
       port: 8323
 kind: ConfigMap
 metadata:
-  creationTimestamp: null
   name: llama-stack-config

benchmarking/k8s-benchmark/stack_run_config.yaml

@@ -95,21 +95,30 @@ providers:
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
     config: {}
-metadata_store:
-  type: postgres
-  host: ${env.POSTGRES_HOST:=localhost}
-  port: ${env.POSTGRES_PORT:=5432}
-  db: ${env.POSTGRES_DB:=llamastack}
-  user: ${env.POSTGRES_USER:=llamastack}
-  password: ${env.POSTGRES_PASSWORD:=llamastack}
-  table_name: llamastack_kvstore
-inference_store:
-  type: postgres
-  host: ${env.POSTGRES_HOST:=localhost}
-  port: ${env.POSTGRES_PORT:=5432}
-  db: ${env.POSTGRES_DB:=llamastack}
-  user: ${env.POSTGRES_USER:=llamastack}
-  password: ${env.POSTGRES_PASSWORD:=llamastack}
+storage:
+  backends:
+    kv_default:
+      type: kv_postgres
+      host: ${env.POSTGRES_HOST:=localhost}
+      port: ${env.POSTGRES_PORT:=5432}
+      db: ${env.POSTGRES_DB:=llamastack}
+      user: ${env.POSTGRES_USER:=llamastack}
+      password: ${env.POSTGRES_PASSWORD:=llamastack}
+      table_name: ${env.POSTGRES_TABLE_NAME:=llamastack_kvstore}
+    sql_default:
+      type: sql_postgres
+      host: ${env.POSTGRES_HOST:=localhost}
+      port: ${env.POSTGRES_PORT:=5432}
+      db: ${env.POSTGRES_DB:=llamastack}
+      user: ${env.POSTGRES_USER:=llamastack}
+      password: ${env.POSTGRES_PASSWORD:=llamastack}
+  references:
+    metadata:
+      backend: kv_default
+      namespace: registry
+    inference:
+      backend: sql_default
+      table_name: inference_store
 models:
 - metadata:
     embedding_dimension: 768

client-sdks/stainless/README.md

@@ -0,0 +1,8 @@
+These are the source-of-truth configuration files used to generate the Stainless client SDKs via Stainless.
+
+- `openapi.yml`: this is the OpenAPI specification for the Llama Stack API.
+- `openapi.stainless.yml`: this is the Stainless _configuration_ which instructs Stainless how to generate the client SDKs.
+
+A small side note: notice the `.yml` suffixes since Stainless uses that suffix typically for its configuration files.
+
+These files go hand-in-hand. As of now, only the `openapi.yml` file is automatically generated using the `run_openapi_generator.sh` script.

client-sdks/stainless/openapi.stainless.yml

@@ -0,0 +1,610 @@
+# yaml-language-server: $schema=https://app.stainlessapi.com/config-internal.schema.json
+
+organization:
+  # Name of your organization or company, used to determine the name of the client
+  # and headings.
+  name: llama-stack-client
+  docs: https://llama-stack.readthedocs.io/en/latest/
+  contact: llamastack@meta.com
+security:
+  - {}
+  - BearerAuth: []
+security_schemes:
+  BearerAuth:
+    type: http
+    scheme: bearer
+# `targets` define the output targets and their customization options, such as
+# whether to emit the Node SDK and what it's package name should be.
+targets:
+  node:
+    package_name: llama-stack-client
+    production_repo: llamastack/llama-stack-client-typescript
+    publish:
+      npm: false
+  python:
+    package_name: llama_stack_client
+    production_repo: llamastack/llama-stack-client-python
+    options:
+      use_uv: true
+    publish:
+      pypi: true
+    project_name: llama_stack_client
+  kotlin:
+    reverse_domain: com.llama_stack_client.api
+    production_repo: null
+    publish:
+      maven: false
+  go:
+    package_name: llama-stack-client
+    production_repo: llamastack/llama-stack-client-go
+    options:
+      enable_v2: true
+      back_compat_use_shared_package: false
+
+# `client_settings` define settings for the API client, such as extra constructor
+# arguments (used for authentication), retry behavior, idempotency, etc.
+client_settings:
+  default_env_prefix: LLAMA_STACK_CLIENT
+  opts:
+    api_key:
+      type: string
+      read_env: LLAMA_STACK_CLIENT_API_KEY
+      auth: { security_scheme: BearerAuth }
+      nullable: true
+
+# `environments` are a map of the name of the environment (e.g. "sandbox",
+# "production") to the corresponding url to use.
+environments:
+  production: http://any-hosted-llama-stack.com
+
+# `pagination` defines [pagination schemes] which provides a template to match
+# endpoints and generate next-page and auto-pagination helpers in the SDKs.
+pagination:
+  - name: datasets_iterrows
+    type: offset
+    request:
+      dataset_id:
+        type: string
+      start_index:
+        type: integer
+        x-stainless-pagination-property:
+          purpose: offset_count_param
+      limit:
+        type: integer
+    response:
+      data:
+        type: array
+        items:
+          type: object
+      next_index:
+        type: integer
+        x-stainless-pagination-property:
+          purpose: offset_count_start_field
+  - name: openai_cursor_page
+    type: cursor
+    request:
+      limit:
+        type: integer
+      after:
+        type: string
+        x-stainless-pagination-property:
+          purpose: next_cursor_param
+    response:
+      data:
+        type: array
+        items: {}
+      has_more:
+        type: boolean
+      last_id:
+        type: string
+        x-stainless-pagination-property:
+          purpose: next_cursor_field
+# `resources` define the structure and organziation for your API, such as how
+# methods and models are grouped together and accessed. See the [configuration
+# guide] for more information.
+#
+# [configuration guide]:
+#   https://app.stainlessapi.com/docs/guides/configure#resources
+resources:
+  $shared:
+    models:
+      agent_config: AgentConfig
+      interleaved_content_item: InterleavedContentItem
+      interleaved_content: InterleavedContent
+      param_type: ParamType
+      safety_violation: SafetyViolation
+      sampling_params: SamplingParams
+      scoring_result: ScoringResult
+      message: Message
+      user_message: UserMessage
+      completion_message: CompletionMessage
+      tool_response_message: ToolResponseMessage
+      system_message: SystemMessage
+      tool_call: ToolCall
+      query_result: RAGQueryResult
+      document: RAGDocument
+      query_config: RAGQueryConfig
+      response_format: ResponseFormat
+  toolgroups:
+    models:
+      tool_group: ToolGroup
+      list_tool_groups_response: ListToolGroupsResponse
+    methods:
+      register: post /v1/toolgroups
+      get: get /v1/toolgroups/{toolgroup_id}
+      list: get /v1/toolgroups
+      unregister: delete /v1/toolgroups/{toolgroup_id}
+  tools:
+    methods:
+      get: get /v1/tools/{tool_name}
+      list:
+        endpoint: get /v1/tools
+        paginated: false
+
+  tool_runtime:
+    models:
+      tool_def: ToolDef
+      tool_invocation_result: ToolInvocationResult
+    methods:
+      list_tools:
+        endpoint: get /v1/tool-runtime/list-tools
+        paginated: false
+      invoke_tool: post /v1/tool-runtime/invoke
+    subresources:
+      rag_tool:
+        methods:
+          insert: post /v1/tool-runtime/rag-tool/insert
+          query: post /v1/tool-runtime/rag-tool/query
+
+  responses:
+    models:
+      response_object_stream: OpenAIResponseObjectStream
+      response_object: OpenAIResponseObject
+    methods:
+      create:
+        type: http
+        endpoint: post /v1/responses
+        streaming:
+          stream_event_model: responses.response_object_stream
+          param_discriminator: stream
+      retrieve: get /v1/responses/{response_id}
+      list:
+        type: http
+        endpoint: get /v1/responses
+      delete:
+        type: http
+        endpoint: delete /v1/responses/{response_id}
+    subresources:
+      input_items:
+        methods:
+          list:
+            type: http
+            endpoint: get /v1/responses/{response_id}/input_items
+
+  conversations:
+    models:
+      conversation_object: Conversation
+    methods:
+      create:
+        type: http
+        endpoint: post /v1/conversations
+      retrieve: get /v1/conversations/{conversation_id}
+      update:
+        type: http
+        endpoint: post /v1/conversations/{conversation_id}
+      delete:
+        type: http
+        endpoint: delete /v1/conversations/{conversation_id}
+    subresources:
+      items:
+        methods:
+          get:
+            type: http
+            endpoint: get /v1/conversations/{conversation_id}/items/{item_id}
+          list:
+            type: http
+            endpoint: get /v1/conversations/{conversation_id}/items
+          create:
+            type: http
+            endpoint: post /v1/conversations/{conversation_id}/items
+
+  inspect:
+    models:
+      healthInfo: HealthInfo
+      providerInfo: ProviderInfo
+      routeInfo: RouteInfo
+      versionInfo: VersionInfo
+    methods:
+      health: get /v1/health
+      version: get /v1/version
+
+  embeddings:
+    models:
+      create_embeddings_response: OpenAIEmbeddingsResponse
+    methods:
+      create: post /v1/embeddings
+
+  chat:
+    models:
+      chat_completion_chunk: OpenAIChatCompletionChunk
+    subresources:
+      completions:
+        methods:
+          create:
+            type: http
+            endpoint: post /v1/chat/completions
+            streaming:
+              stream_event_model: chat.chat_completion_chunk
+              param_discriminator: stream
+          list:
+            type: http
+            endpoint: get /v1/chat/completions
+          retrieve:
+            type: http
+            endpoint: get /v1/chat/completions/{completion_id}
+  completions:
+    methods:
+      create:
+        type: http
+        endpoint: post /v1/completions
+        streaming:
+          param_discriminator: stream
+
+  vector_io:
+    models:
+      queryChunksResponse: QueryChunksResponse
+    methods:
+      insert: post /v1/vector-io/insert
+      query: post /v1/vector-io/query
+
+  vector_stores:
+    models:
+      vector_store: VectorStoreObject
+      list_vector_stores_response: VectorStoreListResponse
+      vector_store_delete_response: VectorStoreDeleteResponse
+      vector_store_search_response: VectorStoreSearchResponsePage
+    methods:
+      create: post /v1/vector_stores
+      list:
+        endpoint: get /v1/vector_stores
+      retrieve: get /v1/vector_stores/{vector_store_id}
+      update: post /v1/vector_stores/{vector_store_id}
+      delete: delete /v1/vector_stores/{vector_store_id}
+      search: post /v1/vector_stores/{vector_store_id}/search
+    subresources:
+      files:
+        models:
+          vector_store_file: VectorStoreFileObject
+        methods:
+          list: get /v1/vector_stores/{vector_store_id}/files
+          retrieve: get /v1/vector_stores/{vector_store_id}/files/{file_id}
+          update: post /v1/vector_stores/{vector_store_id}/files/{file_id}
+          delete: delete /v1/vector_stores/{vector_store_id}/files/{file_id}
+          create: post /v1/vector_stores/{vector_store_id}/files
+          content: get /v1/vector_stores/{vector_store_id}/files/{file_id}/content
+      file_batches:
+        models:
+          vector_store_file_batches: VectorStoreFileBatchObject
+          list_vector_store_files_in_batch_response: VectorStoreFilesListInBatchResponse
+        methods:
+          create: post /v1/vector_stores/{vector_store_id}/file_batches
+          retrieve: get /v1/vector_stores/{vector_store_id}/file_batches/{batch_id}
+          list_files: get /v1/vector_stores/{vector_store_id}/file_batches/{batch_id}/files
+          cancel: post /v1/vector_stores/{vector_store_id}/file_batches/{batch_id}/cancel
+
+  models:
+    models:
+      model: Model
+      list_models_response: ListModelsResponse
+    methods:
+      retrieve: get /v1/models/{model_id}
+      list:
+        endpoint: get /v1/models
+        paginated: false
+      register: post /v1/models
+      unregister: delete /v1/models/{model_id}
+    subresources:
+      openai:
+        methods:
+          list:
+            endpoint: get /v1/models
+            paginated: false
+
+  providers:
+    models:
+      list_providers_response: ListProvidersResponse
+    methods:
+      list:
+        endpoint: get /v1/providers
+        paginated: false
+      retrieve: get /v1/providers/{provider_id}
+
+  routes:
+    models:
+      list_routes_response: ListRoutesResponse
+    methods:
+      list:
+        endpoint: get /v1/inspect/routes
+        paginated: false
+
+
+  moderations:
+    models:
+      create_response: ModerationObject
+    methods:
+      create: post /v1/moderations
+
+
+  safety:
+    models:
+      run_shield_response: RunShieldResponse
+    methods:
+      run_shield: post /v1/safety/run-shield
+
+
+  shields:
+    models:
+      shield: Shield
+      list_shields_response: ListShieldsResponse
+    methods:
+      retrieve: get /v1/shields/{identifier}
+      list:
+        endpoint: get /v1/shields
+        paginated: false
+      register: post /v1/shields
+      delete: delete /v1/shields/{identifier}
+
+  synthetic_data_generation:
+    models:
+      syntheticDataGenerationResponse: SyntheticDataGenerationResponse
+    methods:
+      generate: post /v1/synthetic-data-generation/generate
+
+  telemetry:
+    models:
+      span_with_status: SpanWithStatus
+      trace: Trace
+      query_spans_response: QuerySpansResponse
+      event: Event
+      query_condition: QueryCondition
+    methods:
+      query_traces:
+        endpoint: post /v1alpha/telemetry/traces
+        skip_test_reason: 'unsupported query params in java / kotlin'
+      get_span_tree: post /v1alpha/telemetry/spans/{span_id}/tree
+      query_spans:
+        endpoint: post /v1alpha/telemetry/spans
+        skip_test_reason: 'unsupported query params in java / kotlin'
+      query_metrics:
+        endpoint: post /v1alpha/telemetry/metrics/{metric_name}
+        skip_test_reason: 'unsupported query params in java / kotlin'
+      # log_event: post /v1alpha/telemetry/events
+      save_spans_to_dataset: post /v1alpha/telemetry/spans/export
+      get_span: get /v1alpha/telemetry/traces/{trace_id}/spans/{span_id}
+      get_trace: get /v1alpha/telemetry/traces/{trace_id}
+
+  scoring:
+    methods:
+      score: post /v1/scoring/score
+      score_batch: post /v1/scoring/score-batch
+  scoring_functions:
+    methods:
+      retrieve: get /v1/scoring-functions/{scoring_fn_id}
+      list:
+        endpoint: get /v1/scoring-functions
+        paginated: false
+      register: post /v1/scoring-functions
+    models:
+      scoring_fn: ScoringFn
+      scoring_fn_params: ScoringFnParams
+      list_scoring_functions_response: ListScoringFunctionsResponse
+
+  benchmarks:
+    methods:
+      retrieve: get /v1alpha/eval/benchmarks/{benchmark_id}
+      list:
+        endpoint: get /v1alpha/eval/benchmarks
+        paginated: false
+      register: post /v1alpha/eval/benchmarks
+    models:
+      benchmark: Benchmark
+      list_benchmarks_response: ListBenchmarksResponse
+
+  files:
+    methods:
+      create: post /v1/files
+      list: get /v1/files
+      retrieve: get /v1/files/{file_id}
+      delete: delete /v1/files/{file_id}
+      content: get /v1/files/{file_id}/content
+    models:
+      file: OpenAIFileObject
+      list_files_response: ListOpenAIFileResponse
+      delete_file_response: OpenAIFileDeleteResponse
+
+  alpha:
+    subresources:
+      inference:
+        methods:
+          rerank: post /v1alpha/inference/rerank
+
+      post_training:
+        models:
+          algorithm_config: AlgorithmConfig
+          post_training_job: PostTrainingJob
+          list_post_training_jobs_response: ListPostTrainingJobsResponse
+        methods:
+          preference_optimize: post /v1alpha/post-training/preference-optimize
+          supervised_fine_tune: post /v1alpha/post-training/supervised-fine-tune
+        subresources:
+          job:
+            methods:
+              artifacts: get /v1alpha/post-training/job/artifacts
+              cancel: post /v1alpha/post-training/job/cancel
+              status: get /v1alpha/post-training/job/status
+              list:
+                endpoint: get /v1alpha/post-training/jobs
+                paginated: false
+
+      eval:
+        methods:
+          evaluate_rows: post /v1alpha/eval/benchmarks/{benchmark_id}/evaluations
+          run_eval: post /v1alpha/eval/benchmarks/{benchmark_id}/jobs
+          evaluate_rows_alpha: post /v1alpha/eval/benchmarks/{benchmark_id}/evaluations
+          run_eval_alpha: post /v1alpha/eval/benchmarks/{benchmark_id}/jobs
+
+        subresources:
+          jobs:
+            methods:
+              cancel: delete /v1alpha/eval/benchmarks/{benchmark_id}/jobs/{job_id}
+              status: get /v1alpha/eval/benchmarks/{benchmark_id}/jobs/{job_id}
+              retrieve: get /v1alpha/eval/benchmarks/{benchmark_id}/jobs/{job_id}/result
+        models:
+          evaluate_response: EvaluateResponse
+          benchmark_config: BenchmarkConfig
+          job: Job
+
+      agents:
+        methods:
+          create: post /v1alpha/agents
+          list: get /v1alpha/agents
+          retrieve: get /v1alpha/agents/{agent_id}
+          delete: delete /v1alpha/agents/{agent_id}
+        models:
+          inference_step: InferenceStep
+          tool_execution_step: ToolExecutionStep
+          tool_response: ToolResponse
+          shield_call_step: ShieldCallStep
+          memory_retrieval_step: MemoryRetrievalStep
+        subresources:
+          session:
+            models:
+              session: Session
+            methods:
+              list: get /v1alpha/agents/{agent_id}/sessions
+              create: post /v1alpha/agents/{agent_id}/session
+              delete: delete /v1alpha/agents/{agent_id}/session/{session_id}
+              retrieve: get /v1alpha/agents/{agent_id}/session/{session_id}
+          steps:
+            methods:
+              retrieve: get /v1alpha/agents/{agent_id}/session/{session_id}/turn/{turn_id}/step/{step_id}
+          turn:
+            models:
+              turn: Turn
+              turn_response_event: AgentTurnResponseEvent
+              agent_turn_response_stream_chunk: AgentTurnResponseStreamChunk
+            methods:
+              create:
+                type: http
+                endpoint: post /v1alpha/agents/{agent_id}/session/{session_id}/turn
+                streaming:
+                  stream_event_model: alpha.agents.turn.agent_turn_response_stream_chunk
+                  param_discriminator: stream
+              retrieve: get /v1alpha/agents/{agent_id}/session/{session_id}/turn/{turn_id}
+              resume:
+                type: http
+                endpoint: post /v1alpha/agents/{agent_id}/session/{session_id}/turn/{turn_id}/resume
+                streaming:
+                  stream_event_model: alpha.agents.turn.agent_turn_response_stream_chunk
+                  param_discriminator: stream
+
+  beta:
+    subresources:
+      datasets:
+        models:
+          list_datasets_response: ListDatasetsResponse
+        methods:
+          register: post /v1beta/datasets
+          retrieve: get /v1beta/datasets/{dataset_id}
+          list:
+            endpoint: get /v1beta/datasets
+            paginated: false
+          unregister: delete /v1beta/datasets/{dataset_id}
+          iterrows: get /v1beta/datasetio/iterrows/{dataset_id}
+          appendrows: post /v1beta/datasetio/append-rows/{dataset_id}
+
+
+settings:
+  license: MIT
+  unwrap_response_fields: [ data ]
+
+openapi:
+  transformations:
+    - command: renameValue
+      reason: pydantic reserved name
+      args:
+        filter:
+          only:
+            - '$.components.schemas.InferenceStep.properties.model_response'
+        rename:
+          python:
+            property_name: 'inference_model_response'
+
+    # - command: renameValue
+    #   reason: pydantic reserved name
+    #   args:
+    #     filter:
+    #       only:
+    #         - '$.components.schemas.Model.properties.model_type'
+    #     rename:
+    #       python:
+    #         property_name: 'type'
+    - command: mergeObject
+      reason: Better return_type using enum
+      args:
+        target:
+          - '$.components.schemas'
+        object:
+          ReturnType:
+            additionalProperties: false
+            properties:
+              type:
+                enum:
+                  - string
+                  - number
+                  - boolean
+                  - array
+                  - object
+                  - json
+                  - union
+                  - chat_completion_input
+                  - completion_input
+                  - agent_turn_input
+            required:
+              - type
+            type: object
+    - command: replaceProperties
+      reason: Replace return type properties with better model (see above)
+      args:
+        filter:
+          only:
+            - '$.components.schemas.ScoringFn.properties.return_type'
+            - '$.components.schemas.RegisterScoringFunctionRequest.properties.return_type'
+        value:
+          $ref: '#/components/schemas/ReturnType'
+    - command: oneOfToAnyOf
+      reason: Prism (mock server) doesn't like one of our requests as it technically matches multiple variants
+    - reason: For better names
+      command: extractToRefs
+      args:
+        ref:
+          target: '$.components.schemas.ToolCallDelta.properties.tool_call'
+          name: '#/components/schemas/ToolCallOrString'
+
+# `readme` is used to configure the code snippets that will be rendered in the
+# README.md of various SDKs. In particular, you can change the `headline`
+# snippet's endpoint and the arguments to call it with.
+readme:
+  example_requests:
+    default:
+      type: request
+      endpoint: post /v1/chat/completions
+      params: &ref_0 {}
+    headline:
+      type: request
+      endpoint: post /v1/models
+      params: *ref_0
+    pagination:
+      type: request
+      endpoint: post /v1/chat/completions
+      params: {}

containers/Containerfile

@@ -60,6 +60,17 @@ ENV RUN_CONFIG_PATH=${RUN_CONFIG_PATH}
 # Copy the repository so editable installs and run configurations are available.
 COPY . /workspace
 
+# Install the client package if it is provided
+# NOTE: this is installed before llama-stack since llama-stack depends on llama-stack-client-python
+RUN set -eux; \
+    if [ -n "$LLAMA_STACK_CLIENT_DIR" ]; then \
+        if [ ! -d "$LLAMA_STACK_CLIENT_DIR" ]; then \
+            echo "LLAMA_STACK_CLIENT_DIR is set but $LLAMA_STACK_CLIENT_DIR does not exist" >&2; \
+            exit 1; \
+        fi; \
+        uv pip install --no-cache-dir -e "$LLAMA_STACK_CLIENT_DIR"; \
+    fi;
+
 # Install llama-stack
 RUN set -eux; \
     if [ "$INSTALL_MODE" = "editable" ]; then \
@@ -83,16 +94,6 @@ RUN set -eux; \
         fi; \
     fi;
 
-# Install the client package if it is provided
-RUN set -eux; \
-    if [ -n "$LLAMA_STACK_CLIENT_DIR" ]; then \
-        if [ ! -d "$LLAMA_STACK_CLIENT_DIR" ]; then \
-            echo "LLAMA_STACK_CLIENT_DIR is set but $LLAMA_STACK_CLIENT_DIR does not exist" >&2; \
-            exit 1; \
-        fi; \
-        uv pip install --no-cache-dir -e "$LLAMA_STACK_CLIENT_DIR"; \
-    fi;
-
 # Install the dependencies for the distribution
 RUN set -eux; \
     if [ -z "$DISTRO_NAME" ]; then \

docs/docs/building_applications/rag.mdx

@@ -88,18 +88,19 @@ Llama Stack provides OpenAI-compatible RAG capabilities through:
 To enable automatic vector store creation without specifying embedding models, configure a default embedding model in your run.yaml like so:
 
 ```yaml
-models:
-  - model_id: nomic-ai/nomic-embed-text-v1.5
-    provider_id: inline::sentence-transformers
-    metadata:
-      embedding_dimension: 768
-      default_configured: true
+vector_stores:
+  default_provider_id: faiss
+  default_embedding_model:
+    provider_id: sentence-transformers
+    model_id: nomic-ai/nomic-embed-text-v1.5
 ```
 
 With this configuration:
-- `client.vector_stores.create()` works without requiring embedding model parameters
-- The system automatically uses the default model and its embedding dimension for any newly created vector store
-- Only one model can be marked as `default_configured: true`
+- `client.vector_stores.create()` works without requiring embedding model or provider parameters
+- The system automatically uses the default vector store provider (`faiss`) when multiple providers are available
+- The system automatically uses the default embedding model (`sentence-transformers/nomic-ai/nomic-embed-text-v1.5`) for any newly created vector store
+- The `default_provider_id` specifies which vector storage backend to use
+- The `default_embedding_model` specifies both the inference provider and model for embeddings
 
 ## Vector Store Operations
 
@@ -108,14 +109,15 @@ With this configuration:
 You can create vector stores with automatic or explicit embedding model selection:
 
 ```python
-# Automatic - uses default configured embedding model
+# Automatic - uses default configured embedding model and vector store provider
 vs = client.vector_stores.create()
 
-# Explicit - specify embedding model when you need a specific one
+# Explicit - specify embedding model and/or provider when you need specific ones
 vs = client.vector_stores.create(
     extra_body={
-        "embedding_model": "nomic-ai/nomic-embed-text-v1.5",
-        "embedding_dimension": 768
+        "provider_id": "faiss",  # Optional: specify vector store provider
+        "embedding_model": "sentence-transformers/nomic-ai/nomic-embed-text-v1.5",
+        "embedding_dimension": 768  # Optional: will be auto-detected if not provided
     }
 )
 ```

docs/docs/distributions/configuration.mdx

@@ -44,18 +44,32 @@ providers:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
     config:
-      persistence_store:
-        type: sqlite
-        namespace: null
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/agents_store.db
+      persistence:
+        agent_state:
+          backend: kv_default
+          namespace: agents
+        responses:
+          backend: sql_default
+          table_name: responses
   telemetry:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
     config: {}
-metadata_store:
-  namespace: null
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/registry.db
+storage:
+  backends:
+    kv_default:
+      type: kv_sqlite
+      db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/kvstore.db
+    sql_default:
+      type: sql_sqlite
+      db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/sqlstore.db
+  references:
+    metadata:
+      backend: kv_default
+      namespace: registry
+    inference:
+      backend: sql_default
+      table_name: inference_store
 models:
 - metadata: {}
   model_id: ${env.INFERENCE_MODEL}

docs/docs/distributions/k8s/stack-configmap.yaml

@@ -1,56 +1,155 @@
 apiVersion: v1
 data:
-  stack_run_config.yaml: "version: '2'\nimage_name: kubernetes-demo\napis:\n- agents\n-
-    inference\n- files\n- safety\n- telemetry\n- tool_runtime\n- vector_io\nproviders:\n
-    \ inference:\n  - provider_id: vllm-inference\n    provider_type: remote::vllm\n
-    \   config:\n      url: ${env.VLLM_URL:=http://localhost:8000/v1}\n      max_tokens:
-    ${env.VLLM_MAX_TOKENS:=4096}\n      api_token: ${env.VLLM_API_TOKEN:=fake}\n      tls_verify:
-    ${env.VLLM_TLS_VERIFY:=true}\n  - provider_id: vllm-safety\n    provider_type:
-    remote::vllm\n    config:\n      url: ${env.VLLM_SAFETY_URL:=http://localhost:8000/v1}\n
-    \     max_tokens: ${env.VLLM_MAX_TOKENS:=4096}\n      api_token: ${env.VLLM_API_TOKEN:=fake}\n
-    \     tls_verify: ${env.VLLM_TLS_VERIFY:=true}\n  - provider_id: sentence-transformers\n
-    \   provider_type: inline::sentence-transformers\n    config: {}\n  vector_io:\n
-    \ - provider_id: ${env.ENABLE_CHROMADB:+chromadb}\n    provider_type: remote::chromadb\n
-    \   config:\n      url: ${env.CHROMADB_URL:=}\n      kvstore:\n        type: postgres\n
-    \       host: ${env.POSTGRES_HOST:=localhost}\n        port: ${env.POSTGRES_PORT:=5432}\n
-    \       db: ${env.POSTGRES_DB:=llamastack}\n        user: ${env.POSTGRES_USER:=llamastack}\n
-    \       password: ${env.POSTGRES_PASSWORD:=llamastack}\n  files:\n  - provider_id:
-    meta-reference-files\n    provider_type: inline::localfs\n    config:\n      storage_dir:
-    ${env.FILES_STORAGE_DIR:=~/.llama/distributions/starter/files}\n      metadata_store:\n
-    \       type: sqlite\n        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/files_metadata.db
-    \ \n  safety:\n  - provider_id: llama-guard\n    provider_type: inline::llama-guard\n
-    \   config:\n      excluded_categories: []\n  agents:\n  - provider_id: meta-reference\n
-    \   provider_type: inline::meta-reference\n    config:\n      persistence_store:\n
-    \       type: postgres\n        host: ${env.POSTGRES_HOST:=localhost}\n        port:
-    ${env.POSTGRES_PORT:=5432}\n        db: ${env.POSTGRES_DB:=llamastack}\n        user:
-    ${env.POSTGRES_USER:=llamastack}\n        password: ${env.POSTGRES_PASSWORD:=llamastack}\n
-    \     responses_store:\n        type: postgres\n        host: ${env.POSTGRES_HOST:=localhost}\n
-    \       port: ${env.POSTGRES_PORT:=5432}\n        db: ${env.POSTGRES_DB:=llamastack}\n
-    \       user: ${env.POSTGRES_USER:=llamastack}\n        password: ${env.POSTGRES_PASSWORD:=llamastack}\n
-    \ telemetry:\n  - provider_id: meta-reference\n    provider_type: inline::meta-reference\n
-    \   config:\n      service_name: \"${env.OTEL_SERVICE_NAME:=\\u200B}\"\n      sinks:
-    ${env.TELEMETRY_SINKS:=console}\n  tool_runtime:\n  - provider_id: brave-search\n
-    \   provider_type: remote::brave-search\n    config:\n      api_key: ${env.BRAVE_SEARCH_API_KEY:+}\n
-    \     max_results: 3\n  - provider_id: tavily-search\n    provider_type: remote::tavily-search\n
-    \   config:\n      api_key: ${env.TAVILY_SEARCH_API_KEY:+}\n      max_results:
-    3\n  - provider_id: rag-runtime\n    provider_type: inline::rag-runtime\n    config:
-    {}\n  - provider_id: model-context-protocol\n    provider_type: remote::model-context-protocol\n
-    \   config: {}\nmetadata_store:\n  type: postgres\n  host: ${env.POSTGRES_HOST:=localhost}\n
-    \ port: ${env.POSTGRES_PORT:=5432}\n  db: ${env.POSTGRES_DB:=llamastack}\n  user:
-    ${env.POSTGRES_USER:=llamastack}\n  password: ${env.POSTGRES_PASSWORD:=llamastack}\n
-    \ table_name: llamastack_kvstore\ninference_store:\n  type: postgres\n  host:
-    ${env.POSTGRES_HOST:=localhost}\n  port: ${env.POSTGRES_PORT:=5432}\n  db: ${env.POSTGRES_DB:=llamastack}\n
-    \ user: ${env.POSTGRES_USER:=llamastack}\n  password: ${env.POSTGRES_PASSWORD:=llamastack}\nmodels:\n-
-    metadata:\n    embedding_dimension: 384\n  model_id: all-MiniLM-L6-v2\n  provider_id:
-    sentence-transformers\n  model_type: embedding\n- metadata: {}\n  model_id: ${env.INFERENCE_MODEL}\n
-    \ provider_id: vllm-inference\n  model_type: llm\n- metadata: {}\n  model_id:
-    ${env.SAFETY_MODEL:=meta-llama/Llama-Guard-3-1B}\n  provider_id: vllm-safety\n
-    \ model_type: llm\nshields:\n- shield_id: ${env.SAFETY_MODEL:=meta-llama/Llama-Guard-3-1B}\nvector_dbs:
-    []\ndatasets: []\nscoring_fns: []\nbenchmarks: []\ntool_groups:\n- toolgroup_id:
-    builtin::websearch\n  provider_id: tavily-search\n- toolgroup_id: builtin::rag\n
-    \ provider_id: rag-runtime\nserver:\n  port: 8321\n  auth:\n    provider_config:\n
-    \     type: github_token\n"
+  stack_run_config.yaml: |
+    version: '2'
+    image_name: kubernetes-demo
+    apis:
+    - agents
+    - inference
+    - files
+    - safety
+    - telemetry
+    - tool_runtime
+    - vector_io
+    providers:
+      inference:
+      - provider_id: vllm-inference
+        provider_type: remote::vllm
+        config:
+          url: ${env.VLLM_URL:=http://localhost:8000/v1}
+          max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
+          api_token: ${env.VLLM_API_TOKEN:=fake}
+          tls_verify: ${env.VLLM_TLS_VERIFY:=true}
+      - provider_id: vllm-safety
+        provider_type: remote::vllm
+        config:
+          url: ${env.VLLM_SAFETY_URL:=http://localhost:8000/v1}
+          max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
+          api_token: ${env.VLLM_API_TOKEN:=fake}
+          tls_verify: ${env.VLLM_TLS_VERIFY:=true}
+      - provider_id: sentence-transformers
+        provider_type: inline::sentence-transformers
+        config: {}
+      vector_io:
+      - provider_id: ${env.ENABLE_CHROMADB:+chromadb}
+        provider_type: remote::chromadb
+        config:
+          url: ${env.CHROMADB_URL:=}
+          kvstore:
+            type: postgres
+            host: ${env.POSTGRES_HOST:=localhost}
+            port: ${env.POSTGRES_PORT:=5432}
+            db: ${env.POSTGRES_DB:=llamastack}
+            user: ${env.POSTGRES_USER:=llamastack}
+            password: ${env.POSTGRES_PASSWORD:=llamastack}
+      files:
+      - provider_id: meta-reference-files
+        provider_type: inline::localfs
+        config:
+          storage_dir: ${env.FILES_STORAGE_DIR:=~/.llama/distributions/starter/files}
+          metadata_store:
+            type: sqlite
+            db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/files_metadata.db
+      safety:
+      - provider_id: llama-guard
+        provider_type: inline::llama-guard
+        config:
+          excluded_categories: []
+      agents:
+      - provider_id: meta-reference
+        provider_type: inline::meta-reference
+        config:
+          persistence_store:
+            type: postgres
+            host: ${env.POSTGRES_HOST:=localhost}
+            port: ${env.POSTGRES_PORT:=5432}
+            db: ${env.POSTGRES_DB:=llamastack}
+            user: ${env.POSTGRES_USER:=llamastack}
+            password: ${env.POSTGRES_PASSWORD:=llamastack}
+          responses_store:
+            type: postgres
+            host: ${env.POSTGRES_HOST:=localhost}
+            port: ${env.POSTGRES_PORT:=5432}
+            db: ${env.POSTGRES_DB:=llamastack}
+            user: ${env.POSTGRES_USER:=llamastack}
+            password: ${env.POSTGRES_PASSWORD:=llamastack}
+      telemetry:
+      - provider_id: meta-reference
+        provider_type: inline::meta-reference
+        config:
+          service_name: "${env.OTEL_SERVICE_NAME:=\u200B}"
+          sinks: ${env.TELEMETRY_SINKS:=console}
+      tool_runtime:
+      - provider_id: brave-search
+        provider_type: remote::brave-search
+        config:
+          api_key: ${env.BRAVE_SEARCH_API_KEY:+}
+          max_results: 3
+      - provider_id: tavily-search
+        provider_type: remote::tavily-search
+        config:
+          api_key: ${env.TAVILY_SEARCH_API_KEY:+}
+          max_results: 3
+      - provider_id: rag-runtime
+        provider_type: inline::rag-runtime
+        config: {}
+      - provider_id: model-context-protocol
+        provider_type: remote::model-context-protocol
+        config: {}
+    storage:
+      backends:
+        kv_default:
+          type: kv_postgres
+          host: ${env.POSTGRES_HOST:=localhost}
+          port: ${env.POSTGRES_PORT:=5432}
+          db: ${env.POSTGRES_DB:=llamastack}
+          user: ${env.POSTGRES_USER:=llamastack}
+          password: ${env.POSTGRES_PASSWORD:=llamastack}
+          table_name: ${env.POSTGRES_TABLE_NAME:=llamastack_kvstore}
+        sql_default:
+          type: sql_postgres
+          host: ${env.POSTGRES_HOST:=localhost}
+          port: ${env.POSTGRES_PORT:=5432}
+          db: ${env.POSTGRES_DB:=llamastack}
+          user: ${env.POSTGRES_USER:=llamastack}
+          password: ${env.POSTGRES_PASSWORD:=llamastack}
+      references:
+        metadata:
+          backend: kv_default
+          namespace: registry
+        inference:
+          backend: sql_default
+          table_name: inference_store
+    models:
+    - metadata:
+        embedding_dimension: 768
+      model_id: nomic-embed-text-v1.5
+      provider_id: sentence-transformers
+      model_type: embedding
+    - metadata: {}
+      model_id: ${env.INFERENCE_MODEL}
+      provider_id: vllm-inference
+      model_type: llm
+    - metadata: {}
+      model_id: ${env.SAFETY_MODEL:=meta-llama/Llama-Guard-3-1B}
+      provider_id: vllm-safety
+      model_type: llm
+    shields:
+    - shield_id: ${env.SAFETY_MODEL:=meta-llama/Llama-Guard-3-1B}
+    vector_dbs: []
+    datasets: []
+    scoring_fns: []
+    benchmarks: []
+    tool_groups:
+    - toolgroup_id: builtin::websearch
+      provider_id: tavily-search
+    - toolgroup_id: builtin::rag
+      provider_id: rag-runtime
+    server:
+      port: 8321
+      auth:
+        provider_config:
+          type: github_token
 kind: ConfigMap
 metadata:
-  creationTimestamp: null
   name: llama-stack-config

docs/docs/distributions/k8s/stack_run_config.yaml

@@ -93,21 +93,30 @@ providers:
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
     config: {}
-metadata_store:
-  type: postgres
-  host: ${env.POSTGRES_HOST:=localhost}
-  port: ${env.POSTGRES_PORT:=5432}
-  db: ${env.POSTGRES_DB:=llamastack}
-  user: ${env.POSTGRES_USER:=llamastack}
-  password: ${env.POSTGRES_PASSWORD:=llamastack}
-  table_name: llamastack_kvstore
-inference_store:
-  type: postgres
-  host: ${env.POSTGRES_HOST:=localhost}
-  port: ${env.POSTGRES_PORT:=5432}
-  db: ${env.POSTGRES_DB:=llamastack}
-  user: ${env.POSTGRES_USER:=llamastack}
-  password: ${env.POSTGRES_PASSWORD:=llamastack}
+storage:
+  backends:
+    kv_default:
+      type: kv_postgres
+      host: ${env.POSTGRES_HOST:=localhost}
+      port: ${env.POSTGRES_PORT:=5432}
+      db: ${env.POSTGRES_DB:=llamastack}
+      user: ${env.POSTGRES_USER:=llamastack}
+      password: ${env.POSTGRES_PASSWORD:=llamastack}
+      table_name: ${env.POSTGRES_TABLE_NAME:=llamastack_kvstore}
+    sql_default:
+      type: sql_postgres
+      host: ${env.POSTGRES_HOST:=localhost}
+      port: ${env.POSTGRES_PORT:=5432}
+      db: ${env.POSTGRES_DB:=llamastack}
+      user: ${env.POSTGRES_USER:=llamastack}
+      password: ${env.POSTGRES_PASSWORD:=llamastack}
+  references:
+    metadata:
+      backend: kv_default
+      namespace: registry
+    inference:
+      backend: sql_default
+      table_name: inference_store
 models:
 - metadata:
     embedding_dimension: 768

docs/docs/distributions/self_hosted_distro/meta-reference-gpu.md

@@ -21,7 +21,7 @@ The `llamastack/distribution-meta-reference-gpu` distribution consists of the fo
 | inference | `inline::meta-reference` |
 | safety | `inline::llama-guard` |
 | scoring | `inline::basic`, `inline::llm-as-judge`, `inline::braintrust` |
-| tool_runtime | `remote::brave-search`, `remote::tavily-search`, `inline::rag-runtime`, `remote::model-context-protocol` |
+| tool_runtime | `remote::brave-search`, `remote::tavily-search`, `remote::model-context-protocol` |
 | vector_io | `inline::faiss`, `remote::chromadb`, `remote::pgvector` |
 
 

docs/docs/distributions/self_hosted_distro/nvidia.md

@@ -16,7 +16,7 @@ The `llamastack/distribution-nvidia` distribution consists of the following prov
 | post_training | `remote::nvidia` |
 | safety | `remote::nvidia` |
 | scoring | `inline::basic` |
-| tool_runtime | `inline::rag-runtime` |
+| tool_runtime |  |
 | vector_io | `inline::faiss` |
 
 

docs/docs/providers/agents/inline_meta-reference.mdx

@@ -14,16 +14,18 @@ Meta's reference implementation of an agent system that can use tools, access ve
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `persistence_store` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite |  |
-| `responses_store` | `utils.sqlstore.sqlstore.SqliteSqlStoreConfig \| utils.sqlstore.sqlstore.PostgresSqlStoreConfig` | No | sqlite |  |
+| `persistence` | `<class 'inline.agents.meta_reference.config.AgentPersistenceConfig'>` | No |  |  |
 
 ## Sample Configuration
 
 ```yaml
-persistence_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/agents_store.db
-responses_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/responses_store.db
+persistence:
+  agent_state:
+    namespace: agents
+    backend: kv_default
+  responses:
+    table_name: responses
+    backend: sql_default
+    max_write_queue_size: 10000
+    num_writers: 4
 ```

docs/docs/providers/batches/inline_reference.mdx

@@ -14,7 +14,7 @@ Reference implementation of batches API with KVStore persistence.
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | Configuration for the key-value store backend. |
+| `kvstore` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  | Configuration for the key-value store backend. |
 | `max_concurrent_batches` | `<class 'int'>` | No | 1 | Maximum number of concurrent batches to process simultaneously. |
 | `max_concurrent_requests_per_batch` | `<class 'int'>` | No | 10 | Maximum number of concurrent requests to process per batch. |
 
@@ -22,6 +22,6 @@ Reference implementation of batches API with KVStore persistence.
 
 ```yaml
 kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/batches.db
+  namespace: batches
+  backend: kv_default
 ```

docs/docs/providers/datasetio/inline_localfs.mdx

@@ -14,12 +14,12 @@ Local filesystem-based dataset I/O provider for reading and writing datasets to
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite |  |
+| `kvstore` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  |  |
 
 ## Sample Configuration
 
 ```yaml
 kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/localfs_datasetio.db
+  namespace: datasetio::localfs
+  backend: kv_default
 ```

docs/docs/providers/datasetio/remote_huggingface.mdx

@@ -14,12 +14,12 @@ HuggingFace datasets provider for accessing and managing datasets from the Huggi
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite |  |
+| `kvstore` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  |  |
 
 ## Sample Configuration
 
 ```yaml
 kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/huggingface_datasetio.db
+  namespace: datasetio::huggingface
+  backend: kv_default
 ```

docs/docs/providers/eval/inline_meta-reference.mdx

@@ -14,12 +14,12 @@ Meta's reference implementation of evaluation tasks with support for multiple la
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite |  |
+| `kvstore` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  |  |
 
 ## Sample Configuration
 
 ```yaml
 kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/meta_reference_eval.db
+  namespace: eval
+  backend: kv_default
 ```

docs/docs/providers/files/inline_localfs.mdx

@@ -15,7 +15,7 @@ Local filesystem-based file storage provider for managing files and documents lo
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `storage_dir` | `<class 'str'>` | No |  | Directory to store uploaded files |
-| `metadata_store` | `utils.sqlstore.sqlstore.SqliteSqlStoreConfig \| utils.sqlstore.sqlstore.PostgresSqlStoreConfig` | No | sqlite | SQL store configuration for file metadata |
+| `metadata_store` | `<class 'llama_stack.core.storage.datatypes.SqlStoreReference'>` | No |  | SQL store configuration for file metadata |
 | `ttl_secs` | `<class 'int'>` | No | 31536000 |  |
 
 ## Sample Configuration
@@ -23,6 +23,6 @@ Local filesystem-based file storage provider for managing files and documents lo
 ```yaml
 storage_dir: ${env.FILES_STORAGE_DIR:=~/.llama/dummy/files}
 metadata_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/files_metadata.db
+  table_name: files_metadata
+  backend: sql_default
 ```

docs/docs/providers/files/remote_s3.mdx

@@ -20,7 +20,7 @@ AWS S3-based file storage provider for scalable cloud file management with metad
 | `aws_secret_access_key` | `str \| None` | No |  | AWS secret access key (optional if using IAM roles) |
 | `endpoint_url` | `str \| None` | No |  | Custom S3 endpoint URL (for MinIO, LocalStack, etc.) |
 | `auto_create_bucket` | `<class 'bool'>` | No | False | Automatically create the S3 bucket if it doesn't exist |
-| `metadata_store` | `utils.sqlstore.sqlstore.SqliteSqlStoreConfig \| utils.sqlstore.sqlstore.PostgresSqlStoreConfig` | No | sqlite | SQL store configuration for file metadata |
+| `metadata_store` | `<class 'llama_stack.core.storage.datatypes.SqlStoreReference'>` | No |  | SQL store configuration for file metadata |
 
 ## Sample Configuration
 
@@ -32,6 +32,6 @@ aws_secret_access_key: ${env.AWS_SECRET_ACCESS_KEY:=}
 endpoint_url: ${env.S3_ENDPOINT_URL:=}
 auto_create_bucket: ${env.S3_AUTO_CREATE_BUCKET:=false}
 metadata_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/s3_files_metadata.db
+  table_name: s3_files_metadata
+  backend: sql_default
 ```

docs/docs/providers/vector_io/inline_chromadb.mdx

@@ -79,13 +79,13 @@ See [Chroma's documentation](https://docs.trychroma.com/docs/overview/introducti
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `db_path` | `<class 'str'>` | No |  |  |
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | Config for KV store backend |
+| `persistence` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  | Config for KV store backend |
 
 ## Sample Configuration
 
 ```yaml
 db_path: ${env.CHROMADB_PATH}
-kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/chroma_inline_registry.db
+persistence:
+  namespace: vector_io::chroma
+  backend: kv_default
 ```

docs/docs/providers/vector_io/inline_faiss.mdx

@@ -95,12 +95,12 @@ more details about Faiss in general.
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite |  |
+| `persistence` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  |  |
 
 ## Sample Configuration
 
 ```yaml
-kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/faiss_store.db
+persistence:
+  namespace: vector_io::faiss
+  backend: kv_default
 ```

docs/docs/providers/vector_io/inline_meta-reference.mdx

@@ -14,14 +14,14 @@ Meta's reference implementation of a vector database.
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite |  |
+| `persistence` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  |  |
 
 ## Sample Configuration
 
 ```yaml
-kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/faiss_store.db
+persistence:
+  namespace: vector_io::faiss
+  backend: kv_default
 ```
 ## Deprecation Notice
 

docs/docs/providers/vector_io/inline_milvus.mdx

@@ -17,14 +17,14 @@ Please refer to the remote provider documentation.
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `db_path` | `<class 'str'>` | No |  |  |
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | Config for KV store backend (SQLite only for now) |
+| `persistence` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  | Config for KV store backend (SQLite only for now) |
 | `consistency_level` | `<class 'str'>` | No | Strong | The consistency level of the Milvus server |
 
 ## Sample Configuration
 
 ```yaml
 db_path: ${env.MILVUS_DB_PATH:=~/.llama/dummy}/milvus.db
-kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/milvus_registry.db
+persistence:
+  namespace: vector_io::milvus
+  backend: kv_default
 ```

docs/docs/providers/vector_io/inline_qdrant.mdx

@@ -98,13 +98,13 @@ See the [Qdrant documentation](https://qdrant.tech/documentation/) for more deta
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `path` | `<class 'str'>` | No |  |  |
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite |  |
+| `persistence` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  |  |
 
 ## Sample Configuration
 
 ```yaml
 path: ${env.QDRANT_PATH:=~/.llama/~/.llama/dummy}/qdrant.db
-kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/qdrant_registry.db
+persistence:
+  namespace: vector_io::qdrant
+  backend: kv_default
 ```

docs/docs/providers/vector_io/inline_sqlite-vec.mdx

@@ -28,7 +28,7 @@ description: |
   #### Empirical Example
 
   Consider the histogram below in which 10,000 randomly generated strings were inserted
-  in batches of 100 into both Faiss and sqlite-vec using `client.tool_runtime.rag_tool.insert()`.
+  in batches of 100 into both Faiss and sqlite-vec.
 
   ```{image} ../../../../_static/providers/vector_io/write_time_comparison_sqlite-vec-faiss.png
   :alt: Comparison of SQLite-Vec and Faiss write times
@@ -233,7 +233,7 @@ Datasets that can fit in memory, frequent reads | Faiss | Optimized for speed, i
 #### Empirical Example
 
 Consider the histogram below in which 10,000 randomly generated strings were inserted
-in batches of 100 into both Faiss and sqlite-vec using `client.tool_runtime.rag_tool.insert()`.
+in batches of 100 into both Faiss and sqlite-vec.
 
 ```{image} ../../../../_static/providers/vector_io/write_time_comparison_sqlite-vec-faiss.png
 :alt: Comparison of SQLite-Vec and Faiss write times
@@ -408,13 +408,13 @@ See [sqlite-vec's GitHub repo](https://github.com/asg017/sqlite-vec/tree/main) f
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `db_path` | `<class 'str'>` | No |  | Path to the SQLite database file |
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | Config for KV store backend (SQLite only for now) |
+| `persistence` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  | Config for KV store backend (SQLite only for now) |
 
 ## Sample Configuration
 
 ```yaml
 db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/sqlite_vec.db
-kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/sqlite_vec_registry.db
+persistence:
+  namespace: vector_io::sqlite_vec
+  backend: kv_default
 ```

docs/docs/providers/vector_io/inline_sqlite_vec.mdx

@@ -17,15 +17,15 @@ Please refer to the sqlite-vec provider documentation.
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `db_path` | `<class 'str'>` | No |  | Path to the SQLite database file |
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | Config for KV store backend (SQLite only for now) |
+| `persistence` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  | Config for KV store backend (SQLite only for now) |
 
 ## Sample Configuration
 
 ```yaml
 db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/sqlite_vec.db
-kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/sqlite_vec_registry.db
+persistence:
+  namespace: vector_io::sqlite_vec
+  backend: kv_default
 ```
 ## Deprecation Notice
 

docs/docs/providers/vector_io/remote_chromadb.mdx

@@ -78,13 +78,13 @@ See [Chroma's documentation](https://docs.trychroma.com/docs/overview/introducti
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `url` | `str \| None` | No |  |  |
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | Config for KV store backend |
+| `persistence` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  | Config for KV store backend |
 
 ## Sample Configuration
 
 ```yaml
 url: ${env.CHROMADB_URL}
-kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/chroma_remote_registry.db
+persistence:
+  namespace: vector_io::chroma_remote
+  backend: kv_default
 ```

docs/docs/providers/vector_io/remote_milvus.mdx

@@ -408,7 +408,7 @@ For more details on TLS configuration, refer to the [TLS setup guide](https://mi
 | `uri` | `<class 'str'>` | No |  | The URI of the Milvus server |
 | `token` | `str \| None` | No |  | The token of the Milvus server |
 | `consistency_level` | `<class 'str'>` | No | Strong | The consistency level of the Milvus server |
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | Config for KV store backend |
+| `persistence` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  | Config for KV store backend |
 | `config` | `dict` | No | `{}` | This configuration allows additional fields to be passed through to the underlying Milvus client. See the [Milvus](https://milvus.io/docs/install-overview.md) documentation for more details about Milvus in general. |
 
 :::note
@@ -420,7 +420,7 @@ This configuration class accepts additional fields beyond those listed above. Yo
 ```yaml
 uri: ${env.MILVUS_ENDPOINT}
 token: ${env.MILVUS_TOKEN}
-kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/milvus_remote_registry.db
+persistence:
+  namespace: vector_io::milvus_remote
+  backend: kv_default
 ```

docs/docs/providers/vector_io/remote_pgvector.mdx

@@ -218,7 +218,7 @@ See [PGVector's documentation](https://github.com/pgvector/pgvector) for more de
 | `db` | `str \| None` | No | postgres |  |
 | `user` | `str \| None` | No | postgres |  |
 | `password` | `str \| None` | No | mysecretpassword |  |
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig, annotation=NoneType, required=False, default='sqlite', discriminator='type'` | No |  | Config for KV store backend (SQLite only for now) |
+| `persistence` | `llama_stack.core.storage.datatypes.KVStoreReference \| None` | No |  | Config for KV store backend (SQLite only for now) |
 
 ## Sample Configuration
 
@@ -228,7 +228,7 @@ port: ${env.PGVECTOR_PORT:=5432}
 db: ${env.PGVECTOR_DB}
 user: ${env.PGVECTOR_USER}
 password: ${env.PGVECTOR_PASSWORD}
-kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/pgvector_registry.db
+persistence:
+  namespace: vector_io::pgvector
+  backend: kv_default
 ```

docs/docs/providers/vector_io/remote_qdrant.mdx

@@ -26,13 +26,13 @@ Please refer to the inline provider documentation.
 | `prefix` | `str \| None` | No |  |  |
 | `timeout` | `int \| None` | No |  |  |
 | `host` | `str \| None` | No |  |  |
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite |  |
+| `persistence` | `<class 'llama_stack.core.storage.datatypes.KVStoreReference'>` | No |  |  |
 
 ## Sample Configuration
 
 ```yaml
 api_key: ${env.QDRANT_API_KEY:=}
-kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/qdrant_registry.db
+persistence:
+  namespace: vector_io::qdrant_remote
+  backend: kv_default
 ```

docs/docs/providers/vector_io/remote_weaviate.mdx

@@ -75,14 +75,14 @@ See [Weaviate's documentation](https://weaviate.io/developers/weaviate) for more
 |-------|------|----------|---------|-------------|
 | `weaviate_api_key` | `str \| None` | No |  | The API key for the Weaviate instance |
 | `weaviate_cluster_url` | `str \| None` | No | localhost:8080 | The URL of the Weaviate cluster |
-| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig, annotation=NoneType, required=False, default='sqlite', discriminator='type'` | No |  | Config for KV store backend (SQLite only for now) |
+| `persistence` | `llama_stack.core.storage.datatypes.KVStoreReference \| None` | No |  | Config for KV store backend (SQLite only for now) |
 
 ## Sample Configuration
 
 ```yaml
 weaviate_api_key: null
 weaviate_cluster_url: ${env.WEAVIATE_CLUSTER_URL:=localhost:8080}
-kvstore:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/weaviate_registry.db
+persistence:
+  namespace: vector_io::weaviate
+  backend: kv_default
 ```

docs/docs/references/llama_stack_client_cli_reference.md

@@ -32,7 +32,6 @@ Commands:
   scoring_functions  Manage scoring functions.
   shields            Manage safety shield services.
   toolgroups         Manage available tool groups.
-  vector_dbs         Manage vector databases.
 ```
 
 ### `llama-stack-client configure`
@@ -211,53 +210,6 @@ Unregister a model from distribution endpoint
 llama-stack-client models unregister <model_id>
 ```
 
-## Vector DB Management
-Manage vector databases.
-
-
-### `llama-stack-client vector_dbs list`
-Show available vector dbs on distribution endpoint
-```bash
-llama-stack-client vector_dbs list
-```
-```
-┏━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
-┃ identifier               ┃ provider_id ┃ provider_resource_id     ┃ vector_db_type ┃ params                            ┃
-┡━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩
-│ my_demo_vector_db        │ faiss       │ my_demo_vector_db        │                │ embedding_dimension: 768          │
-│                          │             │                          │                │ embedding_model: nomic-embed-text-v1.5 │
-│                          │             │                          │                │ type: vector_db                   │
-│                          │             │                          │                │                                   │
-└──────────────────────────┴─────────────┴──────────────────────────┴────────────────┴───────────────────────────────────┘
-```
-
-### `llama-stack-client vector_dbs register`
-Create a new vector db
-```bash
-llama-stack-client vector_dbs register <vector-db-id> [--provider-id <provider-id>] [--provider-vector-db-id <provider-vector-db-id>] [--embedding-model <embedding-model>] [--embedding-dimension <embedding-dimension>]
-```
-
-
-Required arguments:
-- `VECTOR_DB_ID`: Vector DB ID
-
-Optional arguments:
-- `--provider-id`: Provider ID for the vector db
-- `--provider-vector-db-id`: Provider's vector db ID
-- `--embedding-model`: Embedding model to use. Default: `nomic-embed-text-v1.5`
-- `--embedding-dimension`: Dimension of embeddings. Default: 768
-
-### `llama-stack-client vector_dbs unregister`
-Delete a vector db
-```bash
-llama-stack-client vector_dbs unregister <vector-db-id>
-```
-
-
-Required arguments:
-- `VECTOR_DB_ID`: Vector DB ID
-
-
 ## Shield Management
 Manage safety shield services.
 ### `llama-stack-client shields list`

docs/openapi_generator/pyopenapi/operations.py

@@ -196,16 +196,10 @@ def _get_endpoint_functions(
 def _get_defining_class(member_fn: str, derived_cls: type) -> type:
     "Find the class in which a member function is first defined in a class inheritance hierarchy."
 
-    # This import must be dynamic here
-    from llama_stack.apis.tools import RAGToolRuntime, ToolRuntime
-
     # iterate in reverse member resolution order to find most specific class first
     for cls in reversed(inspect.getmro(derived_cls)):
         for name, _ in inspect.getmembers(cls, inspect.isfunction):
             if name == member_fn:
-                # HACK ALERT
-                if cls == RAGToolRuntime:
-                    return ToolRuntime
                 return cls
 
     raise ValidationError(

docs/openapi_generator/run_openapi_generator.sh

@@ -30,3 +30,5 @@ fi
 stack_dir=$(dirname $(dirname $THIS_DIR))
 PYTHONPATH=$PYTHONPATH:$stack_dir \
   python -m docs.openapi_generator.generate $(dirname $THIS_DIR)/static
+
+cp $stack_dir/docs/static/stainless-llama-stack-spec.yaml $stack_dir/client-sdks/stainless/openapi.yml

docs/static/deprecated-llama-stack-spec.html

@@ -5547,7 +5547,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",
@@ -5798,7 +5798,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",
@@ -9024,6 +9024,10 @@
                         "$ref": "#/components/schemas/OpenAIResponseUsage",
                         "description": "(Optional) Token usage information for the response"
                     },
+                    "instructions": {
+                        "type": "string",
+                        "description": "(Optional) System message inserted into the model's context"
+                    },
                     "input": {
                         "type": "array",
                         "items": {
@@ -9901,6 +9905,10 @@
                     "usage": {
                         "$ref": "#/components/schemas/OpenAIResponseUsage",
                         "description": "(Optional) Token usage information for the response"
+                    },
+                    "instructions": {
+                        "type": "string",
+                        "description": "(Optional) System message inserted into the model's context"
                     }
                 },
                 "additionalProperties": false,

docs/static/deprecated-llama-stack-spec.yaml

@@ -4114,7 +4114,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark
@@ -4303,7 +4303,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark
@@ -6734,6 +6734,10 @@ components:
           $ref: '#/components/schemas/OpenAIResponseUsage'
           description: >-
             (Optional) Token usage information for the response
+        instructions:
+          type: string
+          description: >-
+            (Optional) System message inserted into the model's context
         input:
           type: array
           items:
@@ -7403,6 +7407,10 @@ components:
           $ref: '#/components/schemas/OpenAIResponseUsage'
           description: >-
             (Optional) Token usage information for the response
+        instructions:
+          type: string
+          description: >-
+            (Optional) System message inserted into the model's context
       additionalProperties: false
       required:
         - created_at

docs/static/experimental-llama-stack-spec.html

@@ -1850,7 +1850,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",
@@ -3983,7 +3983,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",

docs/static/experimental-llama-stack-spec.yaml

@@ -1320,7 +1320,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark
@@ -2927,7 +2927,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark

docs/static/llama-stack-spec.html

@@ -2624,89 +2624,6 @@
                 "deprecated": false
             }
         },
-        "/v1/tool-runtime/rag-tool/insert": {
-            "post": {
-                "responses": {
-                    "200": {
-                        "description": "OK"
-                    },
-                    "400": {
-                        "$ref": "#/components/responses/BadRequest400"
-                    },
-                    "429": {
-                        "$ref": "#/components/responses/TooManyRequests429"
-                    },
-                    "500": {
-                        "$ref": "#/components/responses/InternalServerError500"
-                    },
-                    "default": {
-                        "$ref": "#/components/responses/DefaultError"
-                    }
-                },
-                "tags": [
-                    "ToolRuntime"
-                ],
-                "summary": "Index documents so they can be used by the RAG system.",
-                "description": "Index documents so they can be used by the RAG system.",
-                "parameters": [],
-                "requestBody": {
-                    "content": {
-                        "application/json": {
-                            "schema": {
-                                "$ref": "#/components/schemas/InsertRequest"
-                            }
-                        }
-                    },
-                    "required": true
-                },
-                "deprecated": false
-            }
-        },
-        "/v1/tool-runtime/rag-tool/query": {
-            "post": {
-                "responses": {
-                    "200": {
-                        "description": "RAGQueryResult containing the retrieved content and metadata",
-                        "content": {
-                            "application/json": {
-                                "schema": {
-                                    "$ref": "#/components/schemas/RAGQueryResult"
-                                }
-                            }
-                        }
-                    },
-                    "400": {
-                        "$ref": "#/components/responses/BadRequest400"
-                    },
-                    "429": {
-                        "$ref": "#/components/responses/TooManyRequests429"
-                    },
-                    "500": {
-                        "$ref": "#/components/responses/InternalServerError500"
-                    },
-                    "default": {
-                        "$ref": "#/components/responses/DefaultError"
-                    }
-                },
-                "tags": [
-                    "ToolRuntime"
-                ],
-                "summary": "Query the RAG system for context; typically invoked by the agent.",
-                "description": "Query the RAG system for context; typically invoked by the agent.",
-                "parameters": [],
-                "requestBody": {
-                    "content": {
-                        "application/json": {
-                            "schema": {
-                                "$ref": "#/components/schemas/QueryRequest"
-                            }
-                        }
-                    },
-                    "required": true
-                },
-                "deprecated": false
-            }
-        },
         "/v1/toolgroups": {
             "get": {
                 "responses": {
@@ -6800,7 +6717,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",
@@ -7600,6 +7517,10 @@
                         "$ref": "#/components/schemas/OpenAIResponseUsage",
                         "description": "(Optional) Token usage information for the response"
                     },
+                    "instructions": {
+                        "type": "string",
+                        "description": "(Optional) System message inserted into the model's context"
+                    },
                     "input": {
                         "type": "array",
                         "items": {
@@ -8148,6 +8069,10 @@
                     "usage": {
                         "$ref": "#/components/schemas/OpenAIResponseUsage",
                         "description": "(Optional) Token usage information for the response"
+                    },
+                    "instructions": {
+                        "type": "string",
+                        "description": "(Optional) System message inserted into the model's context"
                     }
                 },
                 "additionalProperties": false,
@@ -10197,7 +10122,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",
@@ -10679,7 +10604,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",
@@ -11375,346 +11300,6 @@
                 "title": "ListToolDefsResponse",
                 "description": "Response containing a list of tool definitions."
             },
-            "RAGDocument": {
-                "type": "object",
-                "properties": {
-                    "document_id": {
-                        "type": "string",
-                        "description": "The unique identifier for the document."
-                    },
-                    "content": {
-                        "oneOf": [
-                            {
-                                "type": "string"
-                            },
-                            {
-                                "$ref": "#/components/schemas/InterleavedContentItem"
-                            },
-                            {
-                                "type": "array",
-                                "items": {
-                                    "$ref": "#/components/schemas/InterleavedContentItem"
-                                }
-                            },
-                            {
-                                "$ref": "#/components/schemas/URL"
-                            }
-                        ],
-                        "description": "The content of the document."
-                    },
-                    "mime_type": {
-                        "type": "string",
-                        "description": "The MIME type of the document."
-                    },
-                    "metadata": {
-                        "type": "object",
-                        "additionalProperties": {
-                            "oneOf": [
-                                {
-                                    "type": "null"
-                                },
-                                {
-                                    "type": "boolean"
-                                },
-                                {
-                                    "type": "number"
-                                },
-                                {
-                                    "type": "string"
-                                },
-                                {
-                                    "type": "array"
-                                },
-                                {
-                                    "type": "object"
-                                }
-                            ]
-                        },
-                        "description": "Additional metadata for the document."
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "document_id",
-                    "content",
-                    "metadata"
-                ],
-                "title": "RAGDocument",
-                "description": "A document to be used for document ingestion in the RAG Tool."
-            },
-            "InsertRequest": {
-                "type": "object",
-                "properties": {
-                    "documents": {
-                        "type": "array",
-                        "items": {
-                            "$ref": "#/components/schemas/RAGDocument"
-                        },
-                        "description": "List of documents to index in the RAG system"
-                    },
-                    "vector_db_id": {
-                        "type": "string",
-                        "description": "ID of the vector database to store the document embeddings"
-                    },
-                    "chunk_size_in_tokens": {
-                        "type": "integer",
-                        "description": "(Optional) Size in tokens for document chunking during indexing"
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "documents",
-                    "vector_db_id",
-                    "chunk_size_in_tokens"
-                ],
-                "title": "InsertRequest"
-            },
-            "DefaultRAGQueryGeneratorConfig": {
-                "type": "object",
-                "properties": {
-                    "type": {
-                        "type": "string",
-                        "const": "default",
-                        "default": "default",
-                        "description": "Type of query generator, always 'default'"
-                    },
-                    "separator": {
-                        "type": "string",
-                        "default": " ",
-                        "description": "String separator used to join query terms"
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "type",
-                    "separator"
-                ],
-                "title": "DefaultRAGQueryGeneratorConfig",
-                "description": "Configuration for the default RAG query generator."
-            },
-            "LLMRAGQueryGeneratorConfig": {
-                "type": "object",
-                "properties": {
-                    "type": {
-                        "type": "string",
-                        "const": "llm",
-                        "default": "llm",
-                        "description": "Type of query generator, always 'llm'"
-                    },
-                    "model": {
-                        "type": "string",
-                        "description": "Name of the language model to use for query generation"
-                    },
-                    "template": {
-                        "type": "string",
-                        "description": "Template string for formatting the query generation prompt"
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "type",
-                    "model",
-                    "template"
-                ],
-                "title": "LLMRAGQueryGeneratorConfig",
-                "description": "Configuration for the LLM-based RAG query generator."
-            },
-            "RAGQueryConfig": {
-                "type": "object",
-                "properties": {
-                    "query_generator_config": {
-                        "oneOf": [
-                            {
-                                "$ref": "#/components/schemas/DefaultRAGQueryGeneratorConfig"
-                            },
-                            {
-                                "$ref": "#/components/schemas/LLMRAGQueryGeneratorConfig"
-                            }
-                        ],
-                        "discriminator": {
-                            "propertyName": "type",
-                            "mapping": {
-                                "default": "#/components/schemas/DefaultRAGQueryGeneratorConfig",
-                                "llm": "#/components/schemas/LLMRAGQueryGeneratorConfig"
-                            }
-                        },
-                        "description": "Configuration for the query generator."
-                    },
-                    "max_tokens_in_context": {
-                        "type": "integer",
-                        "default": 4096,
-                        "description": "Maximum number of tokens in the context."
-                    },
-                    "max_chunks": {
-                        "type": "integer",
-                        "default": 5,
-                        "description": "Maximum number of chunks to retrieve."
-                    },
-                    "chunk_template": {
-                        "type": "string",
-                        "default": "Result {index}\nContent: {chunk.content}\nMetadata: {metadata}\n",
-                        "description": "Template for formatting each retrieved chunk in the context. Available placeholders: {index} (1-based chunk ordinal), {chunk.content} (chunk content string), {metadata} (chunk metadata dict). Default: \"Result {index}\\nContent: {chunk.content}\\nMetadata: {metadata}\\n\""
-                    },
-                    "mode": {
-                        "$ref": "#/components/schemas/RAGSearchMode",
-                        "default": "vector",
-                        "description": "Search mode for retrieval—either \"vector\", \"keyword\", or \"hybrid\". Default \"vector\"."
-                    },
-                    "ranker": {
-                        "$ref": "#/components/schemas/Ranker",
-                        "description": "Configuration for the ranker to use in hybrid search. Defaults to RRF ranker."
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "query_generator_config",
-                    "max_tokens_in_context",
-                    "max_chunks",
-                    "chunk_template"
-                ],
-                "title": "RAGQueryConfig",
-                "description": "Configuration for the RAG query generation."
-            },
-            "RAGSearchMode": {
-                "type": "string",
-                "enum": [
-                    "vector",
-                    "keyword",
-                    "hybrid"
-                ],
-                "title": "RAGSearchMode",
-                "description": "Search modes for RAG query retrieval: - VECTOR: Uses vector similarity search for semantic matching - KEYWORD: Uses keyword-based search for exact matching - HYBRID: Combines both vector and keyword search for better results"
-            },
-            "RRFRanker": {
-                "type": "object",
-                "properties": {
-                    "type": {
-                        "type": "string",
-                        "const": "rrf",
-                        "default": "rrf",
-                        "description": "The type of ranker, always \"rrf\""
-                    },
-                    "impact_factor": {
-                        "type": "number",
-                        "default": 60.0,
-                        "description": "The impact factor for RRF scoring. Higher values give more weight to higher-ranked results. Must be greater than 0"
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "type",
-                    "impact_factor"
-                ],
-                "title": "RRFRanker",
-                "description": "Reciprocal Rank Fusion (RRF) ranker configuration."
-            },
-            "Ranker": {
-                "oneOf": [
-                    {
-                        "$ref": "#/components/schemas/RRFRanker"
-                    },
-                    {
-                        "$ref": "#/components/schemas/WeightedRanker"
-                    }
-                ],
-                "discriminator": {
-                    "propertyName": "type",
-                    "mapping": {
-                        "rrf": "#/components/schemas/RRFRanker",
-                        "weighted": "#/components/schemas/WeightedRanker"
-                    }
-                }
-            },
-            "WeightedRanker": {
-                "type": "object",
-                "properties": {
-                    "type": {
-                        "type": "string",
-                        "const": "weighted",
-                        "default": "weighted",
-                        "description": "The type of ranker, always \"weighted\""
-                    },
-                    "alpha": {
-                        "type": "number",
-                        "default": 0.5,
-                        "description": "Weight factor between 0 and 1. 0 means only use keyword scores, 1 means only use vector scores, values in between blend both scores."
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "type",
-                    "alpha"
-                ],
-                "title": "WeightedRanker",
-                "description": "Weighted ranker configuration that combines vector and keyword scores."
-            },
-            "QueryRequest": {
-                "type": "object",
-                "properties": {
-                    "content": {
-                        "$ref": "#/components/schemas/InterleavedContent",
-                        "description": "The query content to search for in the indexed documents"
-                    },
-                    "vector_db_ids": {
-                        "type": "array",
-                        "items": {
-                            "type": "string"
-                        },
-                        "description": "List of vector database IDs to search within"
-                    },
-                    "query_config": {
-                        "$ref": "#/components/schemas/RAGQueryConfig",
-                        "description": "(Optional) Configuration parameters for the query operation"
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "content",
-                    "vector_db_ids"
-                ],
-                "title": "QueryRequest"
-            },
-            "RAGQueryResult": {
-                "type": "object",
-                "properties": {
-                    "content": {
-                        "$ref": "#/components/schemas/InterleavedContent",
-                        "description": "(Optional) The retrieved content from the query"
-                    },
-                    "metadata": {
-                        "type": "object",
-                        "additionalProperties": {
-                            "oneOf": [
-                                {
-                                    "type": "null"
-                                },
-                                {
-                                    "type": "boolean"
-                                },
-                                {
-                                    "type": "number"
-                                },
-                                {
-                                    "type": "string"
-                                },
-                                {
-                                    "type": "array"
-                                },
-                                {
-                                    "type": "object"
-                                }
-                            ]
-                        },
-                        "description": "Additional metadata about the query result"
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "metadata"
-                ],
-                "title": "RAGQueryResult",
-                "description": "Result of a RAG query containing retrieved content and metadata."
-            },
             "ToolGroup": {
                 "type": "object",
                 "properties": {
@@ -11732,7 +11317,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",

docs/static/llama-stack-spec.yaml

@@ -2036,69 +2036,6 @@ paths:
           schema:
             $ref: '#/components/schemas/URL'
       deprecated: false
-  /v1/tool-runtime/rag-tool/insert:
-    post:
-      responses:
-        '200':
-          description: OK
-        '400':
-          $ref: '#/components/responses/BadRequest400'
-        '429':
-          $ref: >-
-            #/components/responses/TooManyRequests429
-        '500':
-          $ref: >-
-            #/components/responses/InternalServerError500
-        default:
-          $ref: '#/components/responses/DefaultError'
-      tags:
-        - ToolRuntime
-      summary: >-
-        Index documents so they can be used by the RAG system.
-      description: >-
-        Index documents so they can be used by the RAG system.
-      parameters: []
-      requestBody:
-        content:
-          application/json:
-            schema:
-              $ref: '#/components/schemas/InsertRequest'
-        required: true
-      deprecated: false
-  /v1/tool-runtime/rag-tool/query:
-    post:
-      responses:
-        '200':
-          description: >-
-            RAGQueryResult containing the retrieved content and metadata
-          content:
-            application/json:
-              schema:
-                $ref: '#/components/schemas/RAGQueryResult'
-        '400':
-          $ref: '#/components/responses/BadRequest400'
-        '429':
-          $ref: >-
-            #/components/responses/TooManyRequests429
-        '500':
-          $ref: >-
-            #/components/responses/InternalServerError500
-        default:
-          $ref: '#/components/responses/DefaultError'
-      tags:
-        - ToolRuntime
-      summary: >-
-        Query the RAG system for context; typically invoked by the agent.
-      description: >-
-        Query the RAG system for context; typically invoked by the agent.
-      parameters: []
-      requestBody:
-        content:
-          application/json:
-            schema:
-              $ref: '#/components/schemas/QueryRequest'
-        required: true
-      deprecated: false
   /v1/toolgroups:
     get:
       responses:
@@ -5227,7 +5164,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark
@@ -5815,6 +5752,10 @@ components:
           $ref: '#/components/schemas/OpenAIResponseUsage'
           description: >-
             (Optional) Token usage information for the response
+        instructions:
+          type: string
+          description: >-
+            (Optional) System message inserted into the model's context
         input:
           type: array
           items:
@@ -6218,6 +6159,10 @@ components:
           $ref: '#/components/schemas/OpenAIResponseUsage'
           description: >-
             (Optional) Token usage information for the response
+        instructions:
+          type: string
+          description: >-
+            (Optional) System message inserted into the model's context
       additionalProperties: false
       required:
         - created_at
@@ -7911,7 +7856,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark
@@ -8219,7 +8164,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark
@@ -8700,274 +8645,6 @@ components:
       title: ListToolDefsResponse
       description: >-
         Response containing a list of tool definitions.
-    RAGDocument:
-      type: object
-      properties:
-        document_id:
-          type: string
-          description: The unique identifier for the document.
-        content:
-          oneOf:
-            - type: string
-            - $ref: '#/components/schemas/InterleavedContentItem'
-            - type: array
-              items:
-                $ref: '#/components/schemas/InterleavedContentItem'
-            - $ref: '#/components/schemas/URL'
-          description: The content of the document.
-        mime_type:
-          type: string
-          description: The MIME type of the document.
-        metadata:
-          type: object
-          additionalProperties:
-            oneOf:
-              - type: 'null'
-              - type: boolean
-              - type: number
-              - type: string
-              - type: array
-              - type: object
-          description: Additional metadata for the document.
-      additionalProperties: false
-      required:
-        - document_id
-        - content
-        - metadata
-      title: RAGDocument
-      description: >-
-        A document to be used for document ingestion in the RAG Tool.
-    InsertRequest:
-      type: object
-      properties:
-        documents:
-          type: array
-          items:
-            $ref: '#/components/schemas/RAGDocument'
-          description: >-
-            List of documents to index in the RAG system
-        vector_db_id:
-          type: string
-          description: >-
-            ID of the vector database to store the document embeddings
-        chunk_size_in_tokens:
-          type: integer
-          description: >-
-            (Optional) Size in tokens for document chunking during indexing
-      additionalProperties: false
-      required:
-        - documents
-        - vector_db_id
-        - chunk_size_in_tokens
-      title: InsertRequest
-    DefaultRAGQueryGeneratorConfig:
-      type: object
-      properties:
-        type:
-          type: string
-          const: default
-          default: default
-          description: >-
-            Type of query generator, always 'default'
-        separator:
-          type: string
-          default: ' '
-          description: >-
-            String separator used to join query terms
-      additionalProperties: false
-      required:
-        - type
-        - separator
-      title: DefaultRAGQueryGeneratorConfig
-      description: >-
-        Configuration for the default RAG query generator.
-    LLMRAGQueryGeneratorConfig:
-      type: object
-      properties:
-        type:
-          type: string
-          const: llm
-          default: llm
-          description: Type of query generator, always 'llm'
-        model:
-          type: string
-          description: >-
-            Name of the language model to use for query generation
-        template:
-          type: string
-          description: >-
-            Template string for formatting the query generation prompt
-      additionalProperties: false
-      required:
-        - type
-        - model
-        - template
-      title: LLMRAGQueryGeneratorConfig
-      description: >-
-        Configuration for the LLM-based RAG query generator.
-    RAGQueryConfig:
-      type: object
-      properties:
-        query_generator_config:
-          oneOf:
-            - $ref: '#/components/schemas/DefaultRAGQueryGeneratorConfig'
-            - $ref: '#/components/schemas/LLMRAGQueryGeneratorConfig'
-          discriminator:
-            propertyName: type
-            mapping:
-              default: '#/components/schemas/DefaultRAGQueryGeneratorConfig'
-              llm: '#/components/schemas/LLMRAGQueryGeneratorConfig'
-          description: Configuration for the query generator.
-        max_tokens_in_context:
-          type: integer
-          default: 4096
-          description: Maximum number of tokens in the context.
-        max_chunks:
-          type: integer
-          default: 5
-          description: Maximum number of chunks to retrieve.
-        chunk_template:
-          type: string
-          default: >
-            Result {index}
-
-            Content: {chunk.content}
-
-            Metadata: {metadata}
-          description: >-
-            Template for formatting each retrieved chunk in the context. Available
-            placeholders: {index} (1-based chunk ordinal), {chunk.content} (chunk
-            content string), {metadata} (chunk metadata dict). Default: "Result {index}\nContent:
-            {chunk.content}\nMetadata: {metadata}\n"
-        mode:
-          $ref: '#/components/schemas/RAGSearchMode'
-          default: vector
-          description: >-
-            Search mode for retrieval—either "vector", "keyword", or "hybrid". Default
-            "vector".
-        ranker:
-          $ref: '#/components/schemas/Ranker'
-          description: >-
-            Configuration for the ranker to use in hybrid search. Defaults to RRF
-            ranker.
-      additionalProperties: false
-      required:
-        - query_generator_config
-        - max_tokens_in_context
-        - max_chunks
-        - chunk_template
-      title: RAGQueryConfig
-      description: >-
-        Configuration for the RAG query generation.
-    RAGSearchMode:
-      type: string
-      enum:
-        - vector
-        - keyword
-        - hybrid
-      title: RAGSearchMode
-      description: >-
-        Search modes for RAG query retrieval: - VECTOR: Uses vector similarity search
-        for semantic matching - KEYWORD: Uses keyword-based search for exact matching
-        - HYBRID: Combines both vector and keyword search for better results
-    RRFRanker:
-      type: object
-      properties:
-        type:
-          type: string
-          const: rrf
-          default: rrf
-          description: The type of ranker, always "rrf"
-        impact_factor:
-          type: number
-          default: 60.0
-          description: >-
-            The impact factor for RRF scoring. Higher values give more weight to higher-ranked
-            results. Must be greater than 0
-      additionalProperties: false
-      required:
-        - type
-        - impact_factor
-      title: RRFRanker
-      description: >-
-        Reciprocal Rank Fusion (RRF) ranker configuration.
-    Ranker:
-      oneOf:
-        - $ref: '#/components/schemas/RRFRanker'
-        - $ref: '#/components/schemas/WeightedRanker'
-      discriminator:
-        propertyName: type
-        mapping:
-          rrf: '#/components/schemas/RRFRanker'
-          weighted: '#/components/schemas/WeightedRanker'
-    WeightedRanker:
-      type: object
-      properties:
-        type:
-          type: string
-          const: weighted
-          default: weighted
-          description: The type of ranker, always "weighted"
-        alpha:
-          type: number
-          default: 0.5
-          description: >-
-            Weight factor between 0 and 1. 0 means only use keyword scores, 1 means
-            only use vector scores, values in between blend both scores.
-      additionalProperties: false
-      required:
-        - type
-        - alpha
-      title: WeightedRanker
-      description: >-
-        Weighted ranker configuration that combines vector and keyword scores.
-    QueryRequest:
-      type: object
-      properties:
-        content:
-          $ref: '#/components/schemas/InterleavedContent'
-          description: >-
-            The query content to search for in the indexed documents
-        vector_db_ids:
-          type: array
-          items:
-            type: string
-          description: >-
-            List of vector database IDs to search within
-        query_config:
-          $ref: '#/components/schemas/RAGQueryConfig'
-          description: >-
-            (Optional) Configuration parameters for the query operation
-      additionalProperties: false
-      required:
-        - content
-        - vector_db_ids
-      title: QueryRequest
-    RAGQueryResult:
-      type: object
-      properties:
-        content:
-          $ref: '#/components/schemas/InterleavedContent'
-          description: >-
-            (Optional) The retrieved content from the query
-        metadata:
-          type: object
-          additionalProperties:
-            oneOf:
-              - type: 'null'
-              - type: boolean
-              - type: number
-              - type: string
-              - type: array
-              - type: object
-          description: >-
-            Additional metadata about the query result
-      additionalProperties: false
-      required:
-        - metadata
-      title: RAGQueryResult
-      description: >-
-        Result of a RAG query containing retrieved content and metadata.
     ToolGroup:
       type: object
       properties:
@@ -8982,7 +8659,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark

docs/static/stainless-llama-stack-spec.html

@@ -2624,89 +2624,6 @@
                 "deprecated": false
             }
         },
-        "/v1/tool-runtime/rag-tool/insert": {
-            "post": {
-                "responses": {
-                    "200": {
-                        "description": "OK"
-                    },
-                    "400": {
-                        "$ref": "#/components/responses/BadRequest400"
-                    },
-                    "429": {
-                        "$ref": "#/components/responses/TooManyRequests429"
-                    },
-                    "500": {
-                        "$ref": "#/components/responses/InternalServerError500"
-                    },
-                    "default": {
-                        "$ref": "#/components/responses/DefaultError"
-                    }
-                },
-                "tags": [
-                    "ToolRuntime"
-                ],
-                "summary": "Index documents so they can be used by the RAG system.",
-                "description": "Index documents so they can be used by the RAG system.",
-                "parameters": [],
-                "requestBody": {
-                    "content": {
-                        "application/json": {
-                            "schema": {
-                                "$ref": "#/components/schemas/InsertRequest"
-                            }
-                        }
-                    },
-                    "required": true
-                },
-                "deprecated": false
-            }
-        },
-        "/v1/tool-runtime/rag-tool/query": {
-            "post": {
-                "responses": {
-                    "200": {
-                        "description": "RAGQueryResult containing the retrieved content and metadata",
-                        "content": {
-                            "application/json": {
-                                "schema": {
-                                    "$ref": "#/components/schemas/RAGQueryResult"
-                                }
-                            }
-                        }
-                    },
-                    "400": {
-                        "$ref": "#/components/responses/BadRequest400"
-                    },
-                    "429": {
-                        "$ref": "#/components/responses/TooManyRequests429"
-                    },
-                    "500": {
-                        "$ref": "#/components/responses/InternalServerError500"
-                    },
-                    "default": {
-                        "$ref": "#/components/responses/DefaultError"
-                    }
-                },
-                "tags": [
-                    "ToolRuntime"
-                ],
-                "summary": "Query the RAG system for context; typically invoked by the agent.",
-                "description": "Query the RAG system for context; typically invoked by the agent.",
-                "parameters": [],
-                "requestBody": {
-                    "content": {
-                        "application/json": {
-                            "schema": {
-                                "$ref": "#/components/schemas/QueryRequest"
-                            }
-                        }
-                    },
-                    "required": true
-                },
-                "deprecated": false
-            }
-        },
         "/v1/toolgroups": {
             "get": {
                 "responses": {
@@ -8472,7 +8389,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",
@@ -9272,6 +9189,10 @@
                         "$ref": "#/components/schemas/OpenAIResponseUsage",
                         "description": "(Optional) Token usage information for the response"
                     },
+                    "instructions": {
+                        "type": "string",
+                        "description": "(Optional) System message inserted into the model's context"
+                    },
                     "input": {
                         "type": "array",
                         "items": {
@@ -9820,6 +9741,10 @@
                     "usage": {
                         "$ref": "#/components/schemas/OpenAIResponseUsage",
                         "description": "(Optional) Token usage information for the response"
+                    },
+                    "instructions": {
+                        "type": "string",
+                        "description": "(Optional) System message inserted into the model's context"
                     }
                 },
                 "additionalProperties": false,
@@ -11869,7 +11794,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",
@@ -12351,7 +12276,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",
@@ -13047,346 +12972,6 @@
                 "title": "ListToolDefsResponse",
                 "description": "Response containing a list of tool definitions."
             },
-            "RAGDocument": {
-                "type": "object",
-                "properties": {
-                    "document_id": {
-                        "type": "string",
-                        "description": "The unique identifier for the document."
-                    },
-                    "content": {
-                        "oneOf": [
-                            {
-                                "type": "string"
-                            },
-                            {
-                                "$ref": "#/components/schemas/InterleavedContentItem"
-                            },
-                            {
-                                "type": "array",
-                                "items": {
-                                    "$ref": "#/components/schemas/InterleavedContentItem"
-                                }
-                            },
-                            {
-                                "$ref": "#/components/schemas/URL"
-                            }
-                        ],
-                        "description": "The content of the document."
-                    },
-                    "mime_type": {
-                        "type": "string",
-                        "description": "The MIME type of the document."
-                    },
-                    "metadata": {
-                        "type": "object",
-                        "additionalProperties": {
-                            "oneOf": [
-                                {
-                                    "type": "null"
-                                },
-                                {
-                                    "type": "boolean"
-                                },
-                                {
-                                    "type": "number"
-                                },
-                                {
-                                    "type": "string"
-                                },
-                                {
-                                    "type": "array"
-                                },
-                                {
-                                    "type": "object"
-                                }
-                            ]
-                        },
-                        "description": "Additional metadata for the document."
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "document_id",
-                    "content",
-                    "metadata"
-                ],
-                "title": "RAGDocument",
-                "description": "A document to be used for document ingestion in the RAG Tool."
-            },
-            "InsertRequest": {
-                "type": "object",
-                "properties": {
-                    "documents": {
-                        "type": "array",
-                        "items": {
-                            "$ref": "#/components/schemas/RAGDocument"
-                        },
-                        "description": "List of documents to index in the RAG system"
-                    },
-                    "vector_db_id": {
-                        "type": "string",
-                        "description": "ID of the vector database to store the document embeddings"
-                    },
-                    "chunk_size_in_tokens": {
-                        "type": "integer",
-                        "description": "(Optional) Size in tokens for document chunking during indexing"
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "documents",
-                    "vector_db_id",
-                    "chunk_size_in_tokens"
-                ],
-                "title": "InsertRequest"
-            },
-            "DefaultRAGQueryGeneratorConfig": {
-                "type": "object",
-                "properties": {
-                    "type": {
-                        "type": "string",
-                        "const": "default",
-                        "default": "default",
-                        "description": "Type of query generator, always 'default'"
-                    },
-                    "separator": {
-                        "type": "string",
-                        "default": " ",
-                        "description": "String separator used to join query terms"
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "type",
-                    "separator"
-                ],
-                "title": "DefaultRAGQueryGeneratorConfig",
-                "description": "Configuration for the default RAG query generator."
-            },
-            "LLMRAGQueryGeneratorConfig": {
-                "type": "object",
-                "properties": {
-                    "type": {
-                        "type": "string",
-                        "const": "llm",
-                        "default": "llm",
-                        "description": "Type of query generator, always 'llm'"
-                    },
-                    "model": {
-                        "type": "string",
-                        "description": "Name of the language model to use for query generation"
-                    },
-                    "template": {
-                        "type": "string",
-                        "description": "Template string for formatting the query generation prompt"
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "type",
-                    "model",
-                    "template"
-                ],
-                "title": "LLMRAGQueryGeneratorConfig",
-                "description": "Configuration for the LLM-based RAG query generator."
-            },
-            "RAGQueryConfig": {
-                "type": "object",
-                "properties": {
-                    "query_generator_config": {
-                        "oneOf": [
-                            {
-                                "$ref": "#/components/schemas/DefaultRAGQueryGeneratorConfig"
-                            },
-                            {
-                                "$ref": "#/components/schemas/LLMRAGQueryGeneratorConfig"
-                            }
-                        ],
-                        "discriminator": {
-                            "propertyName": "type",
-                            "mapping": {
-                                "default": "#/components/schemas/DefaultRAGQueryGeneratorConfig",
-                                "llm": "#/components/schemas/LLMRAGQueryGeneratorConfig"
-                            }
-                        },
-                        "description": "Configuration for the query generator."
-                    },
-                    "max_tokens_in_context": {
-                        "type": "integer",
-                        "default": 4096,
-                        "description": "Maximum number of tokens in the context."
-                    },
-                    "max_chunks": {
-                        "type": "integer",
-                        "default": 5,
-                        "description": "Maximum number of chunks to retrieve."
-                    },
-                    "chunk_template": {
-                        "type": "string",
-                        "default": "Result {index}\nContent: {chunk.content}\nMetadata: {metadata}\n",
-                        "description": "Template for formatting each retrieved chunk in the context. Available placeholders: {index} (1-based chunk ordinal), {chunk.content} (chunk content string), {metadata} (chunk metadata dict). Default: \"Result {index}\\nContent: {chunk.content}\\nMetadata: {metadata}\\n\""
-                    },
-                    "mode": {
-                        "$ref": "#/components/schemas/RAGSearchMode",
-                        "default": "vector",
-                        "description": "Search mode for retrieval—either \"vector\", \"keyword\", or \"hybrid\". Default \"vector\"."
-                    },
-                    "ranker": {
-                        "$ref": "#/components/schemas/Ranker",
-                        "description": "Configuration for the ranker to use in hybrid search. Defaults to RRF ranker."
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "query_generator_config",
-                    "max_tokens_in_context",
-                    "max_chunks",
-                    "chunk_template"
-                ],
-                "title": "RAGQueryConfig",
-                "description": "Configuration for the RAG query generation."
-            },
-            "RAGSearchMode": {
-                "type": "string",
-                "enum": [
-                    "vector",
-                    "keyword",
-                    "hybrid"
-                ],
-                "title": "RAGSearchMode",
-                "description": "Search modes for RAG query retrieval: - VECTOR: Uses vector similarity search for semantic matching - KEYWORD: Uses keyword-based search for exact matching - HYBRID: Combines both vector and keyword search for better results"
-            },
-            "RRFRanker": {
-                "type": "object",
-                "properties": {
-                    "type": {
-                        "type": "string",
-                        "const": "rrf",
-                        "default": "rrf",
-                        "description": "The type of ranker, always \"rrf\""
-                    },
-                    "impact_factor": {
-                        "type": "number",
-                        "default": 60.0,
-                        "description": "The impact factor for RRF scoring. Higher values give more weight to higher-ranked results. Must be greater than 0"
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "type",
-                    "impact_factor"
-                ],
-                "title": "RRFRanker",
-                "description": "Reciprocal Rank Fusion (RRF) ranker configuration."
-            },
-            "Ranker": {
-                "oneOf": [
-                    {
-                        "$ref": "#/components/schemas/RRFRanker"
-                    },
-                    {
-                        "$ref": "#/components/schemas/WeightedRanker"
-                    }
-                ],
-                "discriminator": {
-                    "propertyName": "type",
-                    "mapping": {
-                        "rrf": "#/components/schemas/RRFRanker",
-                        "weighted": "#/components/schemas/WeightedRanker"
-                    }
-                }
-            },
-            "WeightedRanker": {
-                "type": "object",
-                "properties": {
-                    "type": {
-                        "type": "string",
-                        "const": "weighted",
-                        "default": "weighted",
-                        "description": "The type of ranker, always \"weighted\""
-                    },
-                    "alpha": {
-                        "type": "number",
-                        "default": 0.5,
-                        "description": "Weight factor between 0 and 1. 0 means only use keyword scores, 1 means only use vector scores, values in between blend both scores."
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "type",
-                    "alpha"
-                ],
-                "title": "WeightedRanker",
-                "description": "Weighted ranker configuration that combines vector and keyword scores."
-            },
-            "QueryRequest": {
-                "type": "object",
-                "properties": {
-                    "content": {
-                        "$ref": "#/components/schemas/InterleavedContent",
-                        "description": "The query content to search for in the indexed documents"
-                    },
-                    "vector_db_ids": {
-                        "type": "array",
-                        "items": {
-                            "type": "string"
-                        },
-                        "description": "List of vector database IDs to search within"
-                    },
-                    "query_config": {
-                        "$ref": "#/components/schemas/RAGQueryConfig",
-                        "description": "(Optional) Configuration parameters for the query operation"
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "content",
-                    "vector_db_ids"
-                ],
-                "title": "QueryRequest"
-            },
-            "RAGQueryResult": {
-                "type": "object",
-                "properties": {
-                    "content": {
-                        "$ref": "#/components/schemas/InterleavedContent",
-                        "description": "(Optional) The retrieved content from the query"
-                    },
-                    "metadata": {
-                        "type": "object",
-                        "additionalProperties": {
-                            "oneOf": [
-                                {
-                                    "type": "null"
-                                },
-                                {
-                                    "type": "boolean"
-                                },
-                                {
-                                    "type": "number"
-                                },
-                                {
-                                    "type": "string"
-                                },
-                                {
-                                    "type": "array"
-                                },
-                                {
-                                    "type": "object"
-                                }
-                            ]
-                        },
-                        "description": "Additional metadata about the query result"
-                    }
-                },
-                "additionalProperties": false,
-                "required": [
-                    "metadata"
-                ],
-                "title": "RAGQueryResult",
-                "description": "Result of a RAG query containing retrieved content and metadata."
-            },
             "ToolGroup": {
                 "type": "object",
                 "properties": {
@@ -13404,7 +12989,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",
@@ -14951,7 +14536,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",
@@ -16696,7 +16281,7 @@
                         "enum": [
                             "model",
                             "shield",
-                            "vector_db",
+                            "vector_store",
                             "dataset",
                             "scoring_function",
                             "benchmark",

docs/static/stainless-llama-stack-spec.yaml

@@ -2039,69 +2039,6 @@ paths:
           schema:
             $ref: '#/components/schemas/URL'
       deprecated: false
-  /v1/tool-runtime/rag-tool/insert:
-    post:
-      responses:
-        '200':
-          description: OK
-        '400':
-          $ref: '#/components/responses/BadRequest400'
-        '429':
-          $ref: >-
-            #/components/responses/TooManyRequests429
-        '500':
-          $ref: >-
-            #/components/responses/InternalServerError500
-        default:
-          $ref: '#/components/responses/DefaultError'
-      tags:
-        - ToolRuntime
-      summary: >-
-        Index documents so they can be used by the RAG system.
-      description: >-
-        Index documents so they can be used by the RAG system.
-      parameters: []
-      requestBody:
-        content:
-          application/json:
-            schema:
-              $ref: '#/components/schemas/InsertRequest'
-        required: true
-      deprecated: false
-  /v1/tool-runtime/rag-tool/query:
-    post:
-      responses:
-        '200':
-          description: >-
-            RAGQueryResult containing the retrieved content and metadata
-          content:
-            application/json:
-              schema:
-                $ref: '#/components/schemas/RAGQueryResult'
-        '400':
-          $ref: '#/components/responses/BadRequest400'
-        '429':
-          $ref: >-
-            #/components/responses/TooManyRequests429
-        '500':
-          $ref: >-
-            #/components/responses/InternalServerError500
-        default:
-          $ref: '#/components/responses/DefaultError'
-      tags:
-        - ToolRuntime
-      summary: >-
-        Query the RAG system for context; typically invoked by the agent.
-      description: >-
-        Query the RAG system for context; typically invoked by the agent.
-      parameters: []
-      requestBody:
-        content:
-          application/json:
-            schema:
-              $ref: '#/components/schemas/QueryRequest'
-        required: true
-      deprecated: false
   /v1/toolgroups:
     get:
       responses:
@@ -6440,7 +6377,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark
@@ -7028,6 +6965,10 @@ components:
           $ref: '#/components/schemas/OpenAIResponseUsage'
           description: >-
             (Optional) Token usage information for the response
+        instructions:
+          type: string
+          description: >-
+            (Optional) System message inserted into the model's context
         input:
           type: array
           items:
@@ -7431,6 +7372,10 @@ components:
           $ref: '#/components/schemas/OpenAIResponseUsage'
           description: >-
             (Optional) Token usage information for the response
+        instructions:
+          type: string
+          description: >-
+            (Optional) System message inserted into the model's context
       additionalProperties: false
       required:
         - created_at
@@ -9124,7 +9069,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark
@@ -9432,7 +9377,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark
@@ -9913,274 +9858,6 @@ components:
       title: ListToolDefsResponse
       description: >-
         Response containing a list of tool definitions.
-    RAGDocument:
-      type: object
-      properties:
-        document_id:
-          type: string
-          description: The unique identifier for the document.
-        content:
-          oneOf:
-            - type: string
-            - $ref: '#/components/schemas/InterleavedContentItem'
-            - type: array
-              items:
-                $ref: '#/components/schemas/InterleavedContentItem'
-            - $ref: '#/components/schemas/URL'
-          description: The content of the document.
-        mime_type:
-          type: string
-          description: The MIME type of the document.
-        metadata:
-          type: object
-          additionalProperties:
-            oneOf:
-              - type: 'null'
-              - type: boolean
-              - type: number
-              - type: string
-              - type: array
-              - type: object
-          description: Additional metadata for the document.
-      additionalProperties: false
-      required:
-        - document_id
-        - content
-        - metadata
-      title: RAGDocument
-      description: >-
-        A document to be used for document ingestion in the RAG Tool.
-    InsertRequest:
-      type: object
-      properties:
-        documents:
-          type: array
-          items:
-            $ref: '#/components/schemas/RAGDocument'
-          description: >-
-            List of documents to index in the RAG system
-        vector_db_id:
-          type: string
-          description: >-
-            ID of the vector database to store the document embeddings
-        chunk_size_in_tokens:
-          type: integer
-          description: >-
-            (Optional) Size in tokens for document chunking during indexing
-      additionalProperties: false
-      required:
-        - documents
-        - vector_db_id
-        - chunk_size_in_tokens
-      title: InsertRequest
-    DefaultRAGQueryGeneratorConfig:
-      type: object
-      properties:
-        type:
-          type: string
-          const: default
-          default: default
-          description: >-
-            Type of query generator, always 'default'
-        separator:
-          type: string
-          default: ' '
-          description: >-
-            String separator used to join query terms
-      additionalProperties: false
-      required:
-        - type
-        - separator
-      title: DefaultRAGQueryGeneratorConfig
-      description: >-
-        Configuration for the default RAG query generator.
-    LLMRAGQueryGeneratorConfig:
-      type: object
-      properties:
-        type:
-          type: string
-          const: llm
-          default: llm
-          description: Type of query generator, always 'llm'
-        model:
-          type: string
-          description: >-
-            Name of the language model to use for query generation
-        template:
-          type: string
-          description: >-
-            Template string for formatting the query generation prompt
-      additionalProperties: false
-      required:
-        - type
-        - model
-        - template
-      title: LLMRAGQueryGeneratorConfig
-      description: >-
-        Configuration for the LLM-based RAG query generator.
-    RAGQueryConfig:
-      type: object
-      properties:
-        query_generator_config:
-          oneOf:
-            - $ref: '#/components/schemas/DefaultRAGQueryGeneratorConfig'
-            - $ref: '#/components/schemas/LLMRAGQueryGeneratorConfig'
-          discriminator:
-            propertyName: type
-            mapping:
-              default: '#/components/schemas/DefaultRAGQueryGeneratorConfig'
-              llm: '#/components/schemas/LLMRAGQueryGeneratorConfig'
-          description: Configuration for the query generator.
-        max_tokens_in_context:
-          type: integer
-          default: 4096
-          description: Maximum number of tokens in the context.
-        max_chunks:
-          type: integer
-          default: 5
-          description: Maximum number of chunks to retrieve.
-        chunk_template:
-          type: string
-          default: >
-            Result {index}
-
-            Content: {chunk.content}
-
-            Metadata: {metadata}
-          description: >-
-            Template for formatting each retrieved chunk in the context. Available
-            placeholders: {index} (1-based chunk ordinal), {chunk.content} (chunk
-            content string), {metadata} (chunk metadata dict). Default: "Result {index}\nContent:
-            {chunk.content}\nMetadata: {metadata}\n"
-        mode:
-          $ref: '#/components/schemas/RAGSearchMode'
-          default: vector
-          description: >-
-            Search mode for retrieval—either "vector", "keyword", or "hybrid". Default
-            "vector".
-        ranker:
-          $ref: '#/components/schemas/Ranker'
-          description: >-
-            Configuration for the ranker to use in hybrid search. Defaults to RRF
-            ranker.
-      additionalProperties: false
-      required:
-        - query_generator_config
-        - max_tokens_in_context
-        - max_chunks
-        - chunk_template
-      title: RAGQueryConfig
-      description: >-
-        Configuration for the RAG query generation.
-    RAGSearchMode:
-      type: string
-      enum:
-        - vector
-        - keyword
-        - hybrid
-      title: RAGSearchMode
-      description: >-
-        Search modes for RAG query retrieval: - VECTOR: Uses vector similarity search
-        for semantic matching - KEYWORD: Uses keyword-based search for exact matching
-        - HYBRID: Combines both vector and keyword search for better results
-    RRFRanker:
-      type: object
-      properties:
-        type:
-          type: string
-          const: rrf
-          default: rrf
-          description: The type of ranker, always "rrf"
-        impact_factor:
-          type: number
-          default: 60.0
-          description: >-
-            The impact factor for RRF scoring. Higher values give more weight to higher-ranked
-            results. Must be greater than 0
-      additionalProperties: false
-      required:
-        - type
-        - impact_factor
-      title: RRFRanker
-      description: >-
-        Reciprocal Rank Fusion (RRF) ranker configuration.
-    Ranker:
-      oneOf:
-        - $ref: '#/components/schemas/RRFRanker'
-        - $ref: '#/components/schemas/WeightedRanker'
-      discriminator:
-        propertyName: type
-        mapping:
-          rrf: '#/components/schemas/RRFRanker'
-          weighted: '#/components/schemas/WeightedRanker'
-    WeightedRanker:
-      type: object
-      properties:
-        type:
-          type: string
-          const: weighted
-          default: weighted
-          description: The type of ranker, always "weighted"
-        alpha:
-          type: number
-          default: 0.5
-          description: >-
-            Weight factor between 0 and 1. 0 means only use keyword scores, 1 means
-            only use vector scores, values in between blend both scores.
-      additionalProperties: false
-      required:
-        - type
-        - alpha
-      title: WeightedRanker
-      description: >-
-        Weighted ranker configuration that combines vector and keyword scores.
-    QueryRequest:
-      type: object
-      properties:
-        content:
-          $ref: '#/components/schemas/InterleavedContent'
-          description: >-
-            The query content to search for in the indexed documents
-        vector_db_ids:
-          type: array
-          items:
-            type: string
-          description: >-
-            List of vector database IDs to search within
-        query_config:
-          $ref: '#/components/schemas/RAGQueryConfig'
-          description: >-
-            (Optional) Configuration parameters for the query operation
-      additionalProperties: false
-      required:
-        - content
-        - vector_db_ids
-      title: QueryRequest
-    RAGQueryResult:
-      type: object
-      properties:
-        content:
-          $ref: '#/components/schemas/InterleavedContent'
-          description: >-
-            (Optional) The retrieved content from the query
-        metadata:
-          type: object
-          additionalProperties:
-            oneOf:
-              - type: 'null'
-              - type: boolean
-              - type: number
-              - type: string
-              - type: array
-              - type: object
-          description: >-
-            Additional metadata about the query result
-      additionalProperties: false
-      required:
-        - metadata
-      title: RAGQueryResult
-      description: >-
-        Result of a RAG query containing retrieved content and metadata.
     ToolGroup:
       type: object
       properties:
@@ -10195,7 +9872,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark
@@ -11317,7 +10994,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark
@@ -12644,7 +12321,7 @@ components:
           enum:
             - model
             - shield
-            - vector_db
+            - vector_store
             - dataset
             - scoring_function
             - benchmark

llama_stack/apis/agents/openai_responses.py

@@ -545,6 +545,7 @@ class OpenAIResponseObject(BaseModel):
     :param tools: (Optional) An array of tools the model may call while generating a response.
     :param truncation: (Optional) Truncation strategy applied to the response
     :param usage: (Optional) Token usage information for the response
+    :param instructions: (Optional) System message inserted into the model's context
     """
 
     created_at: int
@@ -564,6 +565,7 @@ class OpenAIResponseObject(BaseModel):
     tools: list[OpenAIResponseTool] | None = None
     truncation: str | None = None
     usage: OpenAIResponseUsage | None = None
+    instructions: str | None = None
 
 
 @json_schema_type

llama_stack/apis/datatypes.py

@@ -121,6 +121,7 @@ class Api(Enum, metaclass=DynamicApiMeta):
 
     models = "models"
     shields = "shields"
+    vector_stores = "vector_stores"  # only used for routing table
     datasets = "datasets"
     scoring_functions = "scoring_functions"
     benchmarks = "benchmarks"

llama_stack/apis/resource.py

@@ -13,7 +13,7 @@ from pydantic import BaseModel, Field
 class ResourceType(StrEnum):
     model = "model"
     shield = "shield"
-    vector_db = "vector_db"
+    vector_store = "vector_store"
     dataset = "dataset"
     scoring_function = "scoring_function"
     benchmark = "benchmark"
@@ -34,4 +34,4 @@ class Resource(BaseModel):
 
     provider_id: str = Field(description="ID of the provider that owns this resource")
 
-    type: ResourceType = Field(description="Type of resource (e.g. 'model', 'shield', 'vector_db', etc.)")
+    type: ResourceType = Field(description="Type of resource (e.g. 'model', 'shield', 'vector_store', etc.)")

llama_stack/apis/tools/__init__.py

@@ -4,5 +4,4 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from .rag_tool import *
 from .tools import *

llama_stack/apis/tools/rag_tool.py

@@ -1,218 +0,0 @@
-# Copyright (c) Meta Platforms, Inc. and affiliates.
-# All rights reserved.
-#
-# This source code is licensed under the terms described in the LICENSE file in
-# the root directory of this source tree.
-
-from enum import Enum, StrEnum
-from typing import Annotated, Any, Literal, Protocol
-
-from pydantic import BaseModel, Field, field_validator
-from typing_extensions import runtime_checkable
-
-from llama_stack.apis.common.content_types import URL, InterleavedContent
-from llama_stack.apis.version import LLAMA_STACK_API_V1
-from llama_stack.providers.utils.telemetry.trace_protocol import trace_protocol
-from llama_stack.schema_utils import json_schema_type, register_schema, webmethod
-
-
-@json_schema_type
-class RRFRanker(BaseModel):
-    """
-    Reciprocal Rank Fusion (RRF) ranker configuration.
-
-    :param type: The type of ranker, always "rrf"
-    :param impact_factor: The impact factor for RRF scoring. Higher values give more weight to higher-ranked results.
-                         Must be greater than 0
-    """
-
-    type: Literal["rrf"] = "rrf"
-    impact_factor: float = Field(default=60.0, gt=0.0)  # default of 60 for optimal performance
-
-
-@json_schema_type
-class WeightedRanker(BaseModel):
-    """
-    Weighted ranker configuration that combines vector and keyword scores.
-
-    :param type: The type of ranker, always "weighted"
-    :param alpha: Weight factor between 0 and 1.
-                 0 means only use keyword scores,
-                 1 means only use vector scores,
-                 values in between blend both scores.
-    """
-
-    type: Literal["weighted"] = "weighted"
-    alpha: float = Field(
-        default=0.5,
-        ge=0.0,
-        le=1.0,
-        description="Weight factor between 0 and 1. 0 means only keyword scores, 1 means only vector scores.",
-    )
-
-
-Ranker = Annotated[
-    RRFRanker | WeightedRanker,
-    Field(discriminator="type"),
-]
-register_schema(Ranker, name="Ranker")
-
-
-@json_schema_type
-class RAGDocument(BaseModel):
-    """
-    A document to be used for document ingestion in the RAG Tool.
-
-    :param document_id: The unique identifier for the document.
-    :param content: The content of the document.
-    :param mime_type: The MIME type of the document.
-    :param metadata: Additional metadata for the document.
-    """
-
-    document_id: str
-    content: InterleavedContent | URL
-    mime_type: str | None = None
-    metadata: dict[str, Any] = Field(default_factory=dict)
-
-
-@json_schema_type
-class RAGQueryResult(BaseModel):
-    """Result of a RAG query containing retrieved content and metadata.
-
-    :param content: (Optional) The retrieved content from the query
-    :param metadata: Additional metadata about the query result
-    """
-
-    content: InterleavedContent | None = None
-    metadata: dict[str, Any] = Field(default_factory=dict)
-
-
-@json_schema_type
-class RAGQueryGenerator(Enum):
-    """Types of query generators for RAG systems.
-
-    :cvar default: Default query generator using simple text processing
-    :cvar llm: LLM-based query generator for enhanced query understanding
-    :cvar custom: Custom query generator implementation
-    """
-
-    default = "default"
-    llm = "llm"
-    custom = "custom"
-
-
-@json_schema_type
-class RAGSearchMode(StrEnum):
-    """
-    Search modes for RAG query retrieval:
-    - VECTOR: Uses vector similarity search for semantic matching
-    - KEYWORD: Uses keyword-based search for exact matching
-    - HYBRID: Combines both vector and keyword search for better results
-    """
-
-    VECTOR = "vector"
-    KEYWORD = "keyword"
-    HYBRID = "hybrid"
-
-
-@json_schema_type
-class DefaultRAGQueryGeneratorConfig(BaseModel):
-    """Configuration for the default RAG query generator.
-
-    :param type: Type of query generator, always 'default'
-    :param separator: String separator used to join query terms
-    """
-
-    type: Literal["default"] = "default"
-    separator: str = " "
-
-
-@json_schema_type
-class LLMRAGQueryGeneratorConfig(BaseModel):
-    """Configuration for the LLM-based RAG query generator.
-
-    :param type: Type of query generator, always 'llm'
-    :param model: Name of the language model to use for query generation
-    :param template: Template string for formatting the query generation prompt
-    """
-
-    type: Literal["llm"] = "llm"
-    model: str
-    template: str
-
-
-RAGQueryGeneratorConfig = Annotated[
-    DefaultRAGQueryGeneratorConfig | LLMRAGQueryGeneratorConfig,
-    Field(discriminator="type"),
-]
-register_schema(RAGQueryGeneratorConfig, name="RAGQueryGeneratorConfig")
-
-
-@json_schema_type
-class RAGQueryConfig(BaseModel):
-    """
-    Configuration for the RAG query generation.
-
-    :param query_generator_config: Configuration for the query generator.
-    :param max_tokens_in_context: Maximum number of tokens in the context.
-    :param max_chunks: Maximum number of chunks to retrieve.
-    :param chunk_template: Template for formatting each retrieved chunk in the context.
-        Available placeholders: {index} (1-based chunk ordinal), {chunk.content} (chunk content string), {metadata} (chunk metadata dict).
-        Default: "Result {index}\\nContent: {chunk.content}\\nMetadata: {metadata}\\n"
-    :param mode: Search mode for retrieval—either "vector", "keyword", or "hybrid". Default "vector".
-    :param ranker: Configuration for the ranker to use in hybrid search. Defaults to RRF ranker.
-    """
-
-    # This config defines how a query is generated using the messages
-    # for memory bank retrieval.
-    query_generator_config: RAGQueryGeneratorConfig = Field(default=DefaultRAGQueryGeneratorConfig())
-    max_tokens_in_context: int = 4096
-    max_chunks: int = 5
-    chunk_template: str = "Result {index}\nContent: {chunk.content}\nMetadata: {metadata}\n"
-    mode: RAGSearchMode | None = RAGSearchMode.VECTOR
-    ranker: Ranker | None = Field(default=None)  # Only used for hybrid mode
-
-    @field_validator("chunk_template")
-    def validate_chunk_template(cls, v: str) -> str:
-        if "{chunk.content}" not in v:
-            raise ValueError("chunk_template must contain {chunk.content}")
-        if "{index}" not in v:
-            raise ValueError("chunk_template must contain {index}")
-        if len(v) == 0:
-            raise ValueError("chunk_template must not be empty")
-        return v
-
-
-@runtime_checkable
-@trace_protocol
-class RAGToolRuntime(Protocol):
-    @webmethod(route="/tool-runtime/rag-tool/insert", method="POST", level=LLAMA_STACK_API_V1)
-    async def insert(
-        self,
-        documents: list[RAGDocument],
-        vector_db_id: str,
-        chunk_size_in_tokens: int = 512,
-    ) -> None:
-        """Index documents so they can be used by the RAG system.
-
-        :param documents: List of documents to index in the RAG system
-        :param vector_db_id: ID of the vector database to store the document embeddings
-        :param chunk_size_in_tokens: (Optional) Size in tokens for document chunking during indexing
-        """
-        ...
-
-    @webmethod(route="/tool-runtime/rag-tool/query", method="POST", level=LLAMA_STACK_API_V1)
-    async def query(
-        self,
-        content: InterleavedContent,
-        vector_db_ids: list[str],
-        query_config: RAGQueryConfig | None = None,
-    ) -> RAGQueryResult:
-        """Query the RAG system for context; typically invoked by the agent.
-
-        :param content: The query content to search for in the indexed documents
-        :param vector_db_ids: List of vector database IDs to search within
-        :param query_config: (Optional) Configuration parameters for the query operation
-        :returns: RAGQueryResult containing the retrieved content and metadata
-        """
-        ...

llama_stack/apis/tools/tools.py

@@ -4,7 +4,6 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from enum import Enum
 from typing import Any, Literal, Protocol
 
 from pydantic import BaseModel
@@ -16,8 +15,6 @@ from llama_stack.apis.version import LLAMA_STACK_API_V1
 from llama_stack.providers.utils.telemetry.trace_protocol import trace_protocol
 from llama_stack.schema_utils import json_schema_type, webmethod
 
-from .rag_tool import RAGToolRuntime
-
 
 @json_schema_type
 class ToolDef(BaseModel):
@@ -181,22 +178,11 @@ class ToolGroups(Protocol):
         ...
 
 
-class SpecialToolGroup(Enum):
-    """Special tool groups with predefined functionality.
-
-    :cvar rag_tool: Retrieval-Augmented Generation tool group for document search and retrieval
-    """
-
-    rag_tool = "rag_tool"
-
-
 @runtime_checkable
 @trace_protocol
 class ToolRuntime(Protocol):
     tool_store: ToolStore | None = None
 
-    rag_tool: RAGToolRuntime | None = None
-
     # TODO: This needs to be renamed once OPEN API generator name conflict issue is fixed.
     @webmethod(route="/tool-runtime/list-tools", method="GET", level=LLAMA_STACK_API_V1)
     async def list_runtime_tools(

llama_stack/apis/vector_io/vector_io.py

@@ -15,7 +15,7 @@ from fastapi import Body
 from pydantic import BaseModel, Field
 
 from llama_stack.apis.inference import InterleavedContent
-from llama_stack.apis.vector_dbs import VectorDB
+from llama_stack.apis.vector_stores import VectorStore
 from llama_stack.apis.version import LLAMA_STACK_API_V1
 from llama_stack.providers.utils.telemetry.trace_protocol import trace_protocol
 from llama_stack.providers.utils.vector_io.vector_utils import generate_chunk_id
@@ -140,6 +140,7 @@ class VectorStoreFileCounts(BaseModel):
     total: int
 
 
+# TODO: rename this as OpenAIVectorStore
 @json_schema_type
 class VectorStoreObject(BaseModel):
     """OpenAI Vector Store object.
@@ -517,17 +518,18 @@ class OpenAICreateVectorStoreFileBatchRequestWithExtraBody(BaseModel, extra="all
     chunking_strategy: VectorStoreChunkingStrategy | None = None
 
 
-class VectorDBStore(Protocol):
-    def get_vector_db(self, vector_db_id: str) -> VectorDB | None: ...
+class VectorStoreTable(Protocol):
+    def get_vector_store(self, vector_store_id: str) -> VectorStore | None: ...
 
 
 @runtime_checkable
 @trace_protocol
 class VectorIO(Protocol):
-    vector_db_store: VectorDBStore | None = None
+    vector_store_table: VectorStoreTable | None = None
 
     # this will just block now until chunks are inserted, but it should
     # probably return a Job instance which can be polled for completion
+    # TODO: rename vector_db_id to vector_store_id once Stainless is working
     @webmethod(route="/vector-io/insert", method="POST", level=LLAMA_STACK_API_V1)
     async def insert_chunks(
         self,
@@ -546,6 +548,7 @@ class VectorIO(Protocol):
         """
         ...
 
+    # TODO: rename vector_db_id to vector_store_id once Stainless is working
     @webmethod(route="/vector-io/query", method="POST", level=LLAMA_STACK_API_V1)
     async def query_chunks(
         self,

llama_stack/apis/vector_stores/__init__.py

@@ -4,4 +4,4 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from .vector_dbs import *
+from .vector_stores import *

llama_stack/apis/vector_stores/vector_stores.py

@@ -9,53 +9,43 @@ from typing import Literal
 from pydantic import BaseModel
 
 from llama_stack.apis.resource import Resource, ResourceType
-from llama_stack.schema_utils import json_schema_type
 
 
-@json_schema_type
-class VectorDB(Resource):
+# Internal resource type for storing the vector store routing and other information
+class VectorStore(Resource):
     """Vector database resource for storing and querying vector embeddings.
 
-    :param type: Type of resource, always 'vector_db' for vector databases
+    :param type: Type of resource, always 'vector_store' for vector stores
     :param embedding_model: Name of the embedding model to use for vector generation
     :param embedding_dimension: Dimension of the embedding vectors
     """
 
-    type: Literal[ResourceType.vector_db] = ResourceType.vector_db
+    type: Literal[ResourceType.vector_store] = ResourceType.vector_store
 
     embedding_model: str
     embedding_dimension: int
-    vector_db_name: str | None = None
+    vector_store_name: str | None = None
 
     @property
-    def vector_db_id(self) -> str:
+    def vector_store_id(self) -> str:
         return self.identifier
 
     @property
-    def provider_vector_db_id(self) -> str | None:
+    def provider_vector_store_id(self) -> str | None:
         return self.provider_resource_id
 
 
-class VectorDBInput(BaseModel):
+class VectorStoreInput(BaseModel):
     """Input parameters for creating or configuring a vector database.
 
-    :param vector_db_id: Unique identifier for the vector database
+    :param vector_store_id: Unique identifier for the vector store
     :param embedding_model: Name of the embedding model to use for vector generation
     :param embedding_dimension: Dimension of the embedding vectors
-    :param provider_vector_db_id: (Optional) Provider-specific identifier for the vector database
+    :param provider_vector_store_id: (Optional) Provider-specific identifier for the vector store
     """
 
-    vector_db_id: str
+    vector_store_id: str
     embedding_model: str
     embedding_dimension: int
     provider_id: str | None = None
-    provider_vector_db_id: str | None = None
-
-
-class ListVectorDBsResponse(BaseModel):
-    """Response from listing vector databases.
-
-    :param data: List of vector databases
-    """
-
-    data: list[VectorDB]
+    provider_vector_store_id: str | None = None

llama_stack/cli/stack/_build.py

@@ -1,490 +0,0 @@
-# Copyright (c) Meta Platforms, Inc. and affiliates.
-# All rights reserved.
-#
-# This source code is licensed under the terms described in the LICENSE file in
-# the root directory of this source tree.
-
-import argparse
-import importlib.resources
-import json
-import os
-import shutil
-import sys
-import textwrap
-from functools import lru_cache
-from importlib.abc import Traversable
-from pathlib import Path
-
-import yaml
-from prompt_toolkit import prompt
-from prompt_toolkit.completion import WordCompleter
-from prompt_toolkit.validation import Validator
-from termcolor import colored, cprint
-
-from llama_stack.cli.stack.utils import ImageType
-from llama_stack.cli.table import print_table
-from llama_stack.core.build import (
-    SERVER_DEPENDENCIES,
-    build_image,
-    get_provider_dependencies,
-)
-from llama_stack.core.configure import parse_and_maybe_upgrade_config
-from llama_stack.core.datatypes import (
-    BuildConfig,
-    BuildProvider,
-    DistributionSpec,
-    Provider,
-    StackRunConfig,
-)
-from llama_stack.core.distribution import get_provider_registry
-from llama_stack.core.external import load_external_apis
-from llama_stack.core.resolver import InvalidProviderError
-from llama_stack.core.stack import replace_env_vars
-from llama_stack.core.utils.config_dirs import DISTRIBS_BASE_DIR, EXTERNAL_PROVIDERS_DIR
-from llama_stack.core.utils.dynamic import instantiate_class_type
-from llama_stack.core.utils.exec import formulate_run_args, run_command
-from llama_stack.core.utils.image_types import LlamaStackImageType
-from llama_stack.providers.datatypes import Api
-from llama_stack.providers.utils.sqlstore.sqlstore import SqliteSqlStoreConfig
-
-DISTRIBS_PATH = Path(__file__).parent.parent.parent / "distributions"
-
-
-@lru_cache
-def available_distros_specs() -> dict[str, BuildConfig]:
-    import yaml
-
-    distro_specs = {}
-    for p in DISTRIBS_PATH.rglob("*build.yaml"):
-        distro_name = p.parent.name
-        with open(p) as f:
-            build_config = BuildConfig(**yaml.safe_load(f))
-            distro_specs[distro_name] = build_config
-    return distro_specs
-
-
-def run_stack_build_command(args: argparse.Namespace) -> None:
-    if args.list_distros:
-        return _run_distro_list_cmd()
-
-    if args.image_type == ImageType.VENV.value:
-        current_venv = os.environ.get("VIRTUAL_ENV")
-        image_name = args.image_name or current_venv
-    else:
-        image_name = args.image_name
-
-    if args.template:
-        cprint(
-            "The --template argument is deprecated. Please use --distro instead.",
-            color="red",
-            file=sys.stderr,
-        )
-        distro_name = args.template
-    else:
-        distro_name = args.distribution
-
-    if distro_name:
-        available_distros = available_distros_specs()
-        if distro_name not in available_distros:
-            cprint(
-                f"Could not find distribution {distro_name}. Please run `llama stack build --list-distros` to check out the available distributions",
-                color="red",
-                file=sys.stderr,
-            )
-            sys.exit(1)
-        build_config = available_distros[distro_name]
-        if args.image_type:
-            build_config.image_type = args.image_type
-        else:
-            cprint(
-                f"Please specify a image-type ({' | '.join(e.value for e in ImageType)}) for {distro_name}",
-                color="red",
-                file=sys.stderr,
-            )
-            sys.exit(1)
-    elif args.providers:
-        provider_list: dict[str, list[BuildProvider]] = dict()
-        for api_provider in args.providers.split(","):
-            if "=" not in api_provider:
-                cprint(
-                    "Could not parse `--providers`. Please ensure the list is in the format api1=provider1,api2=provider2",
-                    color="red",
-                    file=sys.stderr,
-                )
-                sys.exit(1)
-            api, provider_type = api_provider.split("=")
-            providers_for_api = get_provider_registry().get(Api(api), None)
-            if providers_for_api is None:
-                cprint(
-                    f"{api} is not a valid API.",
-                    color="red",
-                    file=sys.stderr,
-                )
-                sys.exit(1)
-            if provider_type in providers_for_api:
-                provider = BuildProvider(
-                    provider_type=provider_type,
-                    module=None,
-                )
-                provider_list.setdefault(api, []).append(provider)
-            else:
-                cprint(
-                    f"{provider} is not a valid provider for the {api} API.",
-                    color="red",
-                    file=sys.stderr,
-                )
-                sys.exit(1)
-        distribution_spec = DistributionSpec(
-            providers=provider_list,
-            description=",".join(args.providers),
-        )
-        if not args.image_type:
-            cprint(
-                f"Please specify a image-type (container | venv) for {args.template}",
-                color="red",
-                file=sys.stderr,
-            )
-            sys.exit(1)
-
-        build_config = BuildConfig(image_type=args.image_type, distribution_spec=distribution_spec)
-    elif not args.config and not distro_name:
-        name = prompt(
-            "> Enter a name for your Llama Stack (e.g. my-local-stack): ",
-            validator=Validator.from_callable(
-                lambda x: len(x) > 0,
-                error_message="Name cannot be empty, please enter a name",
-            ),
-        )
-
-        image_type = prompt(
-            "> Enter the image type you want your Llama Stack to be built as (use <TAB> to see options): ",
-            completer=WordCompleter([e.value for e in ImageType]),
-            complete_while_typing=True,
-            validator=Validator.from_callable(
-                lambda x: x in [e.value for e in ImageType],
-                error_message="Invalid image type. Use <TAB> to see options",
-            ),
-        )
-
-        image_name = f"llamastack-{name}"
-
-        cprint(
-            textwrap.dedent(
-                """
-            Llama Stack is composed of several APIs working together. Let's select
-            the provider types (implementations) you want to use for these APIs.
-            """,
-            ),
-            color="green",
-            file=sys.stderr,
-        )
-
-        cprint("Tip: use <TAB> to see options for the providers.\n", color="green", file=sys.stderr)
-
-        providers: dict[str, list[BuildProvider]] = dict()
-        for api, providers_for_api in get_provider_registry().items():
-            available_providers = [x for x in providers_for_api.keys() if x not in ("remote", "remote::sample")]
-            if not available_providers:
-                continue
-            api_provider = prompt(
-                f"> Enter provider for API {api.value}: ",
-                completer=WordCompleter(available_providers),
-                complete_while_typing=True,
-                validator=Validator.from_callable(
-                    lambda x: x in available_providers,  # noqa: B023 - see https://github.com/astral-sh/ruff/issues/7847
-                    error_message="Invalid provider, use <TAB> to see options",
-                ),
-            )
-
-            string_providers = api_provider.split(" ")
-
-            for provider in string_providers:
-                providers.setdefault(api.value, []).append(BuildProvider(provider_type=provider))
-
-        description = prompt(
-            "\n > (Optional) Enter a short description for your Llama Stack: ",
-            default="",
-        )
-
-        distribution_spec = DistributionSpec(
-            providers=providers,
-            description=description,
-        )
-
-        build_config = BuildConfig(image_type=image_type, distribution_spec=distribution_spec)
-    else:
-        with open(args.config) as f:
-            try:
-                contents = yaml.safe_load(f)
-                contents = replace_env_vars(contents)
-                build_config = BuildConfig(**contents)
-                if args.image_type:
-                    build_config.image_type = args.image_type
-            except Exception as e:
-                cprint(
-                    f"Could not parse config file {args.config}: {e}",
-                    color="red",
-                    file=sys.stderr,
-                )
-                sys.exit(1)
-
-    if args.print_deps_only:
-        print(f"# Dependencies for {distro_name or args.config or image_name}")
-        normal_deps, special_deps, external_provider_dependencies = get_provider_dependencies(build_config)
-        normal_deps += SERVER_DEPENDENCIES
-        print(f"uv pip install {' '.join(normal_deps)}")
-        for special_dep in special_deps:
-            print(f"uv pip install {special_dep}")
-        for external_dep in external_provider_dependencies:
-            print(f"uv pip install {external_dep}")
-        return
-
-    try:
-        run_config = _run_stack_build_command_from_build_config(
-            build_config,
-            image_name=image_name,
-            config_path=args.config,
-            distro_name=distro_name,
-        )
-
-    except (Exception, RuntimeError) as exc:
-        import traceback
-
-        cprint(
-            f"Error building stack: {exc}",
-            color="red",
-            file=sys.stderr,
-        )
-        cprint("Stack trace:", color="red", file=sys.stderr)
-        traceback.print_exc()
-        sys.exit(1)
-
-    if run_config is None:
-        cprint(
-            "Run config path is empty",
-            color="red",
-            file=sys.stderr,
-        )
-        sys.exit(1)
-
-    if args.run:
-        config_dict = yaml.safe_load(run_config.read_text())
-        config = parse_and_maybe_upgrade_config(config_dict)
-        if config.external_providers_dir and not config.external_providers_dir.exists():
-            config.external_providers_dir.mkdir(exist_ok=True)
-        run_args = formulate_run_args(args.image_type, image_name or config.image_name)
-        run_args.extend([str(os.getenv("LLAMA_STACK_PORT", 8321)), "--config", str(run_config)])
-        run_command(run_args)
-
-
-def _generate_run_config(
-    build_config: BuildConfig,
-    build_dir: Path,
-    image_name: str,
-) -> Path:
-    """
-    Generate a run.yaml template file for user to edit from a build.yaml file
-    """
-    apis = list(build_config.distribution_spec.providers.keys())
-    run_config = StackRunConfig(
-        container_image=(image_name if build_config.image_type == LlamaStackImageType.CONTAINER.value else None),
-        image_name=image_name,
-        apis=apis,
-        providers={},
-        external_providers_dir=build_config.external_providers_dir
-        if build_config.external_providers_dir
-        else EXTERNAL_PROVIDERS_DIR,
-    )
-    if not run_config.inference_store:
-        run_config.inference_store = SqliteSqlStoreConfig(
-            **SqliteSqlStoreConfig.sample_run_config(
-                __distro_dir__=(DISTRIBS_BASE_DIR / image_name).as_posix(), db_name="inference_store.db"
-            )
-        )
-    # build providers dict
-    provider_registry = get_provider_registry(build_config)
-    for api in apis:
-        run_config.providers[api] = []
-        providers = build_config.distribution_spec.providers[api]
-
-        for provider in providers:
-            pid = provider.provider_type.split("::")[-1]
-
-            p = provider_registry[Api(api)][provider.provider_type]
-            if p.deprecation_error:
-                raise InvalidProviderError(p.deprecation_error)
-
-            try:
-                config_type = instantiate_class_type(provider_registry[Api(api)][provider.provider_type].config_class)
-            except (ModuleNotFoundError, ValueError) as exc:
-                # HACK ALERT:
-                # This code executes after building is done, the import cannot work since the
-                # package is either available in the venv or container - not available on the host.
-                # TODO: use a "is_external" flag in ProviderSpec to check if the provider is
-                # external
-                cprint(
-                    f"Failed to import provider {provider.provider_type} for API {api} - assuming it's external, skipping: {exc}",
-                    color="yellow",
-                    file=sys.stderr,
-                )
-                # Set config_type to None to avoid UnboundLocalError
-                config_type = None
-
-            if config_type is not None and hasattr(config_type, "sample_run_config"):
-                config = config_type.sample_run_config(__distro_dir__=f"~/.llama/distributions/{image_name}")
-            else:
-                config = {}
-
-            p_spec = Provider(
-                provider_id=pid,
-                provider_type=provider.provider_type,
-                config=config,
-                module=provider.module,
-            )
-            run_config.providers[api].append(p_spec)
-
-    run_config_file = build_dir / f"{image_name}-run.yaml"
-
-    with open(run_config_file, "w") as f:
-        to_write = json.loads(run_config.model_dump_json())
-        f.write(yaml.dump(to_write, sort_keys=False))
-
-    # Only print this message for non-container builds since it will be displayed before the
-    # container is built
-    # For non-container builds, the run.yaml is generated at the very end of the build process so it
-    # makes sense to display this message
-    if build_config.image_type != LlamaStackImageType.CONTAINER.value:
-        cprint(f"You can now run your stack with `llama stack run {run_config_file}`", color="green", file=sys.stderr)
-    return run_config_file
-
-
-def _run_stack_build_command_from_build_config(
-    build_config: BuildConfig,
-    image_name: str | None = None,
-    distro_name: str | None = None,
-    config_path: str | None = None,
-) -> Path | Traversable:
-    image_name = image_name or build_config.image_name
-    if build_config.image_type == LlamaStackImageType.CONTAINER.value:
-        if distro_name:
-            image_name = f"distribution-{distro_name}"
-        else:
-            if not image_name:
-                raise ValueError("Please specify an image name when building a container image without a template")
-    else:
-        if not image_name and os.environ.get("UV_SYSTEM_PYTHON"):
-            image_name = "__system__"
-        if not image_name:
-            raise ValueError("Please specify an image name when building a venv image")
-
-    # At this point, image_name should be guaranteed to be a string
-    if image_name is None:
-        raise ValueError("image_name should not be None after validation")
-
-    if distro_name:
-        build_dir = DISTRIBS_BASE_DIR / distro_name
-        build_file_path = build_dir / f"{distro_name}-build.yaml"
-    else:
-        if image_name is None:
-            raise ValueError("image_name cannot be None")
-        build_dir = DISTRIBS_BASE_DIR / image_name
-        build_file_path = build_dir / f"{image_name}-build.yaml"
-
-    os.makedirs(build_dir, exist_ok=True)
-    run_config_file = None
-    # Generate the run.yaml so it can be included in the container image with the proper entrypoint
-    # Only do this if we're building a container image and we're not using a template
-    if build_config.image_type == LlamaStackImageType.CONTAINER.value and not distro_name and config_path:
-        cprint("Generating run.yaml file", color="yellow", file=sys.stderr)
-        run_config_file = _generate_run_config(build_config, build_dir, image_name)
-
-    with open(build_file_path, "w") as f:
-        to_write = json.loads(build_config.model_dump_json(exclude_none=True))
-        f.write(yaml.dump(to_write, sort_keys=False))
-
-    # We first install the external APIs so that the build process can use them and discover the
-    # providers dependencies
-    if build_config.external_apis_dir:
-        cprint("Installing external APIs", color="yellow", file=sys.stderr)
-        external_apis = load_external_apis(build_config)
-        if external_apis:
-            # install the external APIs
-            packages = []
-            for _, api_spec in external_apis.items():
-                if api_spec.pip_packages:
-                    packages.extend(api_spec.pip_packages)
-                    cprint(
-                        f"Installing {api_spec.name} with pip packages {api_spec.pip_packages}",
-                        color="yellow",
-                        file=sys.stderr,
-                    )
-            return_code = run_command(["uv", "pip", "install", *packages])
-            if return_code != 0:
-                packages_str = ", ".join(packages)
-                raise RuntimeError(
-                    f"Failed to install external APIs packages: {packages_str} (return code: {return_code})"
-                )
-
-    return_code = build_image(
-        build_config,
-        image_name,
-        distro_or_config=distro_name or config_path or str(build_file_path),
-        run_config=run_config_file.as_posix() if run_config_file else None,
-    )
-    if return_code != 0:
-        raise RuntimeError(f"Failed to build image {image_name}")
-
-    if distro_name:
-        # copy run.yaml from distribution to build_dir instead of generating it again
-        distro_path = importlib.resources.files("llama_stack") / f"distributions/{distro_name}/run.yaml"
-        run_config_file = build_dir / f"{distro_name}-run.yaml"
-
-        with importlib.resources.as_file(distro_path) as path:
-            shutil.copy(path, run_config_file)
-
-        cprint("Build Successful!", color="green", file=sys.stderr)
-        cprint(f"You can find the newly-built distribution here: {run_config_file}", color="blue", file=sys.stderr)
-        if build_config.image_type == LlamaStackImageType.VENV:
-            cprint(
-                "You can run the new Llama Stack distro (after activating "
-                + colored(image_name, "cyan")
-                + ") via: "
-                + colored(f"llama stack run {run_config_file}", "blue"),
-                color="green",
-                file=sys.stderr,
-            )
-        elif build_config.image_type == LlamaStackImageType.CONTAINER:
-            cprint(
-                "You can run the container with: "
-                + colored(
-                    f"docker run -p 8321:8321 -v ~/.llama:/root/.llama localhost/{image_name} --port 8321", "blue"
-                ),
-                color="green",
-                file=sys.stderr,
-            )
-        return distro_path
-    else:
-        return _generate_run_config(build_config, build_dir, image_name)
-
-
-def _run_distro_list_cmd() -> None:
-    headers = [
-        "Distribution Name",
-        # "Providers",
-        "Description",
-    ]
-
-    rows = []
-    for distro_name, spec in available_distros_specs().items():
-        rows.append(
-            [
-                distro_name,
-                # json.dumps(spec.distribution_spec.providers, indent=2),
-                spec.distribution_spec.description,
-            ]
-        )
-    print_table(
-        rows,
-        headers,
-        separate_rows=True,
-    )

llama_stack/cli/stack/build.py

@@ -1,106 +0,0 @@
-# Copyright (c) Meta Platforms, Inc. and affiliates.
-# All rights reserved.
-#
-# This source code is licensed under the terms described in the LICENSE file in
-# the root directory of this source tree.
-import argparse
-import textwrap
-
-from llama_stack.cli.stack.utils import ImageType
-from llama_stack.cli.subcommand import Subcommand
-from llama_stack.log import get_logger
-
-logger = get_logger(__name__, category="cli")
-
-
-class StackBuild(Subcommand):
-    def __init__(self, subparsers: argparse._SubParsersAction):
-        super().__init__()
-        self.parser = subparsers.add_parser(
-            "build",
-            prog="llama stack build",
-            description="[DEPRECATED] Build a Llama stack container. This command is deprecated and will be removed in a future release. Use `llama stack list-deps <distro>' instead.",
-            formatter_class=argparse.ArgumentDefaultsHelpFormatter,
-        )
-        self._add_arguments()
-        self.parser.set_defaults(func=self._run_stack_build_command)
-
-    def _add_arguments(self):
-        self.parser.add_argument(
-            "--config",
-            type=str,
-            default=None,
-            help="Path to a config file to use for the build. You can find example configs in llama_stack.cores/**/build.yaml. If this argument is not provided, you will be prompted to enter information interactively",
-        )
-
-        self.parser.add_argument(
-            "--template",
-            type=str,
-            default=None,
-            help="""(deprecated) Name of the example template config to use for build. You may use `llama stack build --list-distros` to check out the available distributions""",
-        )
-        self.parser.add_argument(
-            "--distro",
-            "--distribution",
-            dest="distribution",
-            type=str,
-            default=None,
-            help="""Name of the distribution to use for build. You may use `llama stack build --list-distros` to check out the available distributions""",
-        )
-
-        self.parser.add_argument(
-            "--list-distros",
-            "--list-distributions",
-            action="store_true",
-            dest="list_distros",
-            default=False,
-            help="Show the available distributions for building a Llama Stack distribution",
-        )
-
-        self.parser.add_argument(
-            "--image-type",
-            type=str,
-            help="Image Type to use for the build. If not specified, will use the image type from the template config.",
-            choices=[e.value for e in ImageType],
-            default=None,  # no default so we can detect if a user specified --image-type and override image_type in the config
-        )
-
-        self.parser.add_argument(
-            "--image-name",
-            type=str,
-            help=textwrap.dedent(
-                f"""[for image-type={"|".join(e.value for e in ImageType)}] Name of the virtual environment to use for
-the build. If not specified, currently active environment will be used if found.
-            """
-            ),
-            default=None,
-        )
-        self.parser.add_argument(
-            "--print-deps-only",
-            default=False,
-            action="store_true",
-            help="Print the dependencies for the stack only, without building the stack",
-        )
-
-        self.parser.add_argument(
-            "--run",
-            action="store_true",
-            default=False,
-            help="Run the stack after building using the same image type, name, and other applicable arguments",
-        )
-        self.parser.add_argument(
-            "--providers",
-            type=str,
-            default=None,
-            help="Build a config for a list of providers and only those providers. This list is formatted like: api1=provider1,api2=provider2. Where there can be multiple providers per API.",
-        )
-
-    def _run_stack_build_command(self, args: argparse.Namespace) -> None:
-        logger.warning(
-            "The 'llama stack build' command is deprecated and will be removed in a future release. Please use 'llama stack list-deps'"
-        )
-        # always keep implementation completely silo-ed away from CLI so CLI
-        # can be fast to load and reduces dependencies
-        from ._build import run_stack_build_command
-
-        return run_stack_build_command(args)

llama_stack/cli/stack/stack.py

@@ -11,7 +11,6 @@ from llama_stack.cli.stack.list_stacks import StackListBuilds
 from llama_stack.cli.stack.utils import print_subcommand_description
 from llama_stack.cli.subcommand import Subcommand
 
-from .build import StackBuild
 from .list_apis import StackListApis
 from .list_deps import StackListDeps
 from .list_providers import StackListProviders
@@ -41,7 +40,6 @@ class StackParser(Subcommand):
 
         # Add sub-commands
         StackListDeps.create(subparsers)
-        StackBuild.create(subparsers)
         StackListApis.create(subparsers)
         StackListProviders.create(subparsers)
         StackRun.create(subparsers)

llama_stack/cli/stack/utils.py

@@ -17,10 +17,19 @@ from llama_stack.core.datatypes import (
     BuildConfig,
     Provider,
     StackRunConfig,
+    StorageConfig,
 )
 from llama_stack.core.distribution import get_provider_registry
 from llama_stack.core.resolver import InvalidProviderError
-from llama_stack.core.utils.config_dirs import EXTERNAL_PROVIDERS_DIR
+from llama_stack.core.storage.datatypes import (
+    InferenceStoreReference,
+    KVStoreReference,
+    ServerStoresConfig,
+    SqliteKVStoreConfig,
+    SqliteSqlStoreConfig,
+    SqlStoreReference,
+)
+from llama_stack.core.utils.config_dirs import DISTRIBS_BASE_DIR, EXTERNAL_PROVIDERS_DIR
 from llama_stack.core.utils.dynamic import instantiate_class_type
 from llama_stack.core.utils.image_types import LlamaStackImageType
 from llama_stack.providers.datatypes import Api
@@ -51,11 +60,23 @@ def generate_run_config(
     Generate a run.yaml template file for user to edit from a build.yaml file
     """
     apis = list(build_config.distribution_spec.providers.keys())
+    distro_dir = DISTRIBS_BASE_DIR / image_name
     run_config = StackRunConfig(
         container_image=(image_name if build_config.image_type == LlamaStackImageType.CONTAINER.value else None),
         image_name=image_name,
         apis=apis,
         providers={},
+        storage=StorageConfig(
+            backends={
+                "kv_default": SqliteKVStoreConfig(db_path=str(distro_dir / "kvstore.db")),
+                "sql_default": SqliteSqlStoreConfig(db_path=str(distro_dir / "sql_store.db")),
+            },
+            stores=ServerStoresConfig(
+                metadata=KVStoreReference(backend="kv_default", namespace="registry"),
+                inference=InferenceStoreReference(backend="sql_default", table_name="inference_store"),
+                conversations=SqlStoreReference(backend="sql_default", table_name="openai_conversations"),
+            ),
+        ),
         external_providers_dir=build_config.external_providers_dir
         if build_config.external_providers_dir
         else EXTERNAL_PROVIDERS_DIR,

llama_stack/core/access_control/datatypes.py

@@ -41,7 +41,7 @@ class AccessRule(BaseModel):
     A rule defines a list of action either to permit or to forbid. It may specify a
     principal or a resource that must match for the rule to take effect. The resource
     to match should be specified in the form of a type qualified identifier, e.g.
-    model::my-model or vector_db::some-db, or a wildcard for all resources of a type,
+    model::my-model or vector_store::some-db, or a wildcard for all resources of a type,
     e.g. model::*. If the principal or resource are not specified, they will match all
     requests.
 
@@ -79,9 +79,9 @@ class AccessRule(BaseModel):
       description: any user has read access to any resource created by a member of their team
     - forbid:
         actions: [create, read, delete]
-        resource: vector_db::*
+        resource: vector_store::*
       unless: user with admin in roles
-      description: only user with admin role can use vector_db resources
+      description: only user with admin role can use vector_store resources
 
     """
 

llama_stack/core/build_container.sh

@@ -1,410 +0,0 @@
-#!/usr/bin/env bash
-
-# Copyright (c) Meta Platforms, Inc. and affiliates.
-# All rights reserved.
-#
-# This source code is licensed under the terms described in the LICENSE file in
-# the root directory of this source tree.
-
-LLAMA_STACK_DIR=${LLAMA_STACK_DIR:-}
-LLAMA_STACK_CLIENT_DIR=${LLAMA_STACK_CLIENT_DIR:-}
-
-TEST_PYPI_VERSION=${TEST_PYPI_VERSION:-}
-PYPI_VERSION=${PYPI_VERSION:-}
-BUILD_PLATFORM=${BUILD_PLATFORM:-}
-# This timeout (in seconds) is necessary when installing PyTorch via uv since it's likely to time out
-# Reference: https://github.com/astral-sh/uv/pull/1694
-UV_HTTP_TIMEOUT=${UV_HTTP_TIMEOUT:-500}
-
-# mounting is not supported by docker buildx, so we use COPY instead
-USE_COPY_NOT_MOUNT=${USE_COPY_NOT_MOUNT:-}
-# Path to the run.yaml file in the container
-RUN_CONFIG_PATH=/app/run.yaml
-
-BUILD_CONTEXT_DIR=$(pwd)
-
-set -euo pipefail
-
-# Define color codes
-RED='\033[0;31m'
-NC='\033[0m' # No Color
-
-# Usage function
-usage() {
-  echo "Usage: $0 --image-name <image_name> --container-base <container_base> --normal-deps <pip_dependencies> [--run-config <run_config>] [--external-provider-deps <external_provider_deps>] [--optional-deps <special_pip_deps>]"
-  echo "Example: $0 --image-name llama-stack-img --container-base python:3.12-slim --normal-deps 'numpy pandas' --run-config ./run.yaml --external-provider-deps 'foo' --optional-deps 'bar'"
-  exit 1
-}
-
-# Parse arguments
-image_name=""
-container_base=""
-normal_deps=""
-external_provider_deps=""
-optional_deps=""
-run_config=""
-distro_or_config=""
-
-while [[ $# -gt 0 ]]; do
-  key="$1"
-  case "$key" in
-    --image-name)
-      if [[ -z "$2" || "$2" == --* ]]; then
-        echo "Error: --image-name requires a string value" >&2
-        usage
-      fi
-      image_name="$2"
-      shift 2
-      ;;
-    --container-base)
-      if [[ -z "$2" || "$2" == --* ]]; then
-        echo "Error: --container-base requires a string value" >&2
-        usage
-      fi
-      container_base="$2"
-      shift 2
-      ;;
-    --normal-deps)
-      if [[ -z "$2" || "$2" == --* ]]; then
-        echo "Error: --normal-deps requires a string value" >&2
-        usage
-      fi
-      normal_deps="$2"
-      shift 2
-      ;;
-    --external-provider-deps)
-      if [[ -z "$2" || "$2" == --* ]]; then
-        echo "Error: --external-provider-deps requires a string value" >&2
-        usage
-      fi
-      external_provider_deps="$2"
-      shift 2
-      ;;
-    --optional-deps)
-      if [[ -z "$2" || "$2" == --* ]]; then
-        echo "Error: --optional-deps requires a string value" >&2
-        usage
-      fi
-      optional_deps="$2"
-      shift 2
-      ;;
-    --run-config)
-      if [[ -z "$2" || "$2" == --* ]]; then
-        echo "Error: --run-config requires a string value" >&2
-        usage
-      fi
-      run_config="$2"
-      shift 2
-      ;;
-    --distro-or-config)
-      if [[ -z "$2" || "$2" == --* ]]; then
-        echo "Error: --distro-or-config requires a string value" >&2
-        usage
-      fi
-      distro_or_config="$2"
-      shift 2
-      ;;
-    *)
-      echo "Unknown option: $1" >&2
-      usage
-      ;;
-  esac
-done
-
-# Check required arguments
-if [[ -z "$image_name" || -z "$container_base" || -z "$normal_deps" ]]; then
-  echo "Error: --image-name, --container-base, and --normal-deps are required." >&2
-  usage
-fi
-
-CONTAINER_BINARY=${CONTAINER_BINARY:-docker}
-CONTAINER_OPTS=${CONTAINER_OPTS:---progress=plain}
-TEMP_DIR=$(mktemp -d)
-SCRIPT_DIR=$(dirname "$(readlink -f "$0")")
-source "$SCRIPT_DIR/common.sh"
-
-add_to_container() {
-  output_file="$TEMP_DIR/Containerfile"
-  if [ -t 0 ]; then
-    printf '%s\n' "$1" >>"$output_file"
-  else
-    cat >>"$output_file"
-  fi
-}
-
-if ! is_command_available "$CONTAINER_BINARY"; then
-  printf "${RED}Error: ${CONTAINER_BINARY} command not found. Is ${CONTAINER_BINARY} installed and in your PATH?${NC}" >&2
-  exit 1
-fi
-
-if [[ $container_base == *"registry.access.redhat.com/ubi9"* ]]; then
-  add_to_container << EOF
-FROM $container_base
-WORKDIR /app
-
-# We install the Python 3.12 dev headers and build tools so that any
-# C-extension wheels (e.g. polyleven, faiss-cpu) can compile successfully.
-
-RUN dnf -y update && dnf install -y iputils git net-tools wget \
-    vim-minimal python3.12 python3.12-pip python3.12-wheel \
-    python3.12-setuptools python3.12-devel gcc gcc-c++ make && \
-    ln -s /bin/pip3.12 /bin/pip && ln -s /bin/python3.12 /bin/python && dnf clean all
-
-ENV UV_SYSTEM_PYTHON=1
-RUN pip install uv
-EOF
-else
-  add_to_container << EOF
-FROM $container_base
-WORKDIR /app
-
-RUN apt-get update && apt-get install -y \
-       iputils-ping net-tools iproute2 dnsutils telnet \
-       curl wget telnet git\
-       procps psmisc lsof \
-       traceroute \
-       bubblewrap \
-       gcc g++ \
-       && rm -rf /var/lib/apt/lists/*
-
-ENV UV_SYSTEM_PYTHON=1
-RUN pip install uv
-EOF
-fi
-
-# Add pip dependencies first since llama-stack is what will change most often
-# so we can reuse layers.
-if [ -n "$normal_deps" ]; then
-  read -ra pip_args <<<  "$normal_deps"
-  quoted_deps=$(printf " %q" "${pip_args[@]}")
-  add_to_container << EOF
-RUN uv pip install --no-cache $quoted_deps
-EOF
-fi
-
-if [ -n "$optional_deps" ]; then
-  IFS='#' read -ra parts <<<"$optional_deps"
-  for part in "${parts[@]}"; do
-    read -ra pip_args <<< "$part"
-    quoted_deps=$(printf " %q" "${pip_args[@]}")
-    add_to_container <<EOF
-RUN uv pip install --no-cache $quoted_deps
-EOF
-  done
-fi
-
-if [ -n "$external_provider_deps" ]; then
-  IFS='#' read -ra parts <<<"$external_provider_deps"
-  for part in "${parts[@]}"; do
-    read -ra pip_args <<< "$part"
-    quoted_deps=$(printf " %q" "${pip_args[@]}")
-    add_to_container <<EOF
-RUN uv pip install --no-cache $quoted_deps
-EOF
-    add_to_container <<EOF
-RUN python3 - <<PYTHON | uv pip install --no-cache -r -
-import importlib
-import sys
-
-try:
-    package_name = '$part'.split('==')[0].split('>=')[0].split('<=')[0].split('!=')[0].split('<')[0].split('>')[0]
-    module = importlib.import_module(f'{package_name}.provider')
-    spec = module.get_provider_spec()
-    if hasattr(spec, 'pip_packages') and spec.pip_packages:
-        if isinstance(spec.pip_packages, (list, tuple)):
-            print('\n'.join(spec.pip_packages))
-except Exception as e:
-    print(f'Error getting provider spec for {package_name}: {e}', file=sys.stderr)
-PYTHON
-EOF
-  done
-fi
-
-get_python_cmd() {
-    if is_command_available python; then
-        echo "python"
-    elif is_command_available python3; then
-        echo "python3"
-    else
-        echo "Error: Neither python nor python3 is installed. Please install Python to continue." >&2
-        exit 1
-    fi
-}
-
-if [ -n "$run_config" ]; then
-  # Copy the run config to the build context since it's an absolute path
-  cp "$run_config" "$BUILD_CONTEXT_DIR/run.yaml"
-
-  # Parse the run.yaml configuration to identify external provider directories
-  # If external providers are specified, copy their directory to the container
-  # and update the configuration to reference the new container path
-  python_cmd=$(get_python_cmd)
-  external_providers_dir=$($python_cmd -c "import yaml; config = yaml.safe_load(open('$run_config')); print(config.get('external_providers_dir') or '')")
-  external_providers_dir=$(eval echo "$external_providers_dir")
-  if [ -n "$external_providers_dir" ]; then
-    if [ -d "$external_providers_dir" ]; then
-    echo "Copying external providers directory: $external_providers_dir"
-    cp -r "$external_providers_dir" "$BUILD_CONTEXT_DIR/providers.d"
-    add_to_container << EOF
-COPY providers.d /.llama/providers.d
-EOF
-    fi
-
-    # Edit the run.yaml file to change the external_providers_dir to /.llama/providers.d
-    if [ "$(uname)" = "Darwin" ]; then
-      sed -i.bak -e 's|external_providers_dir:.*|external_providers_dir: /.llama/providers.d|' "$BUILD_CONTEXT_DIR/run.yaml"
-      rm -f "$BUILD_CONTEXT_DIR/run.yaml.bak"
-    else
-      sed -i 's|external_providers_dir:.*|external_providers_dir: /.llama/providers.d|' "$BUILD_CONTEXT_DIR/run.yaml"
-    fi
-  fi
-
-  # Copy run config into docker image
-  add_to_container << EOF
-COPY run.yaml $RUN_CONFIG_PATH
-EOF
-fi
-
-stack_mount="/app/llama-stack-source"
-client_mount="/app/llama-stack-client-source"
-
-install_local_package() {
-  local dir="$1"
-  local mount_point="$2"
-  local name="$3"
-
-  if [ ! -d "$dir" ]; then
-    echo "${RED}Warning: $name is set but directory does not exist: $dir${NC}" >&2
-    exit 1
-  fi
-
-  if [ "$USE_COPY_NOT_MOUNT" = "true" ]; then
-    add_to_container << EOF
-COPY $dir $mount_point
-EOF
-  fi
-  add_to_container << EOF
-RUN uv pip install --no-cache -e $mount_point
-EOF
-}
-
-
-if [ -n "$LLAMA_STACK_CLIENT_DIR" ]; then
-  install_local_package "$LLAMA_STACK_CLIENT_DIR" "$client_mount" "LLAMA_STACK_CLIENT_DIR"
-fi
-
-if [ -n "$LLAMA_STACK_DIR" ]; then
-  install_local_package "$LLAMA_STACK_DIR" "$stack_mount" "LLAMA_STACK_DIR"
-else
-  if [ -n "$TEST_PYPI_VERSION" ]; then
-    # these packages are damaged in test-pypi, so install them first
-    add_to_container << EOF
-RUN uv pip install --no-cache fastapi libcst
-EOF
-    add_to_container << EOF
-RUN uv pip install --no-cache --extra-index-url https://test.pypi.org/simple/ \
-  --index-strategy unsafe-best-match \
-  llama-stack==$TEST_PYPI_VERSION
-
-EOF
-  else
-    if [ -n "$PYPI_VERSION" ]; then
-      SPEC_VERSION="llama-stack==${PYPI_VERSION}"
-    else
-      SPEC_VERSION="llama-stack"
-    fi
-    add_to_container << EOF
-RUN uv pip install --no-cache $SPEC_VERSION
-EOF
-  fi
-fi
-
-# remove uv after installation
-  add_to_container << EOF
-RUN pip uninstall -y uv
-EOF
-
-# If a run config is provided, we use the llama stack CLI
-if [[ -n "$run_config" ]]; then
-  add_to_container << EOF
-ENTRYPOINT ["llama", "stack", "run", "$RUN_CONFIG_PATH"]
-EOF
-elif [[ "$distro_or_config" != *.yaml ]]; then
-  add_to_container << EOF
-ENTRYPOINT ["llama", "stack", "run", "$distro_or_config"]
-EOF
-fi
-
-# Add other require item commands genearic to all containers
-add_to_container << EOF
-
-RUN mkdir -p /.llama /.cache && chmod -R g+rw /.llama /.cache && (chmod -R g+rw /app 2>/dev/null || true)
-EOF
-
-printf "Containerfile created successfully in %s/Containerfile\n\n" "$TEMP_DIR"
-cat "$TEMP_DIR"/Containerfile
-printf "\n"
-
-# Start building the CLI arguments
-CLI_ARGS=()
-
-# Read CONTAINER_OPTS and put it in an array
-read -ra CLI_ARGS <<< "$CONTAINER_OPTS"
-
-if [ "$USE_COPY_NOT_MOUNT" != "true" ]; then
-  if [ -n "$LLAMA_STACK_DIR" ]; then
-    CLI_ARGS+=("-v" "$(readlink -f "$LLAMA_STACK_DIR"):$stack_mount")
-  fi
-  if [ -n "$LLAMA_STACK_CLIENT_DIR" ]; then
-    CLI_ARGS+=("-v" "$(readlink -f "$LLAMA_STACK_CLIENT_DIR"):$client_mount")
-  fi
-fi
-
-if is_command_available selinuxenabled && selinuxenabled; then
-  # Disable SELinux labels -- we don't want to relabel the llama-stack source dir
-  CLI_ARGS+=("--security-opt" "label=disable")
-fi
-
-# Set version tag based on PyPI version
-if [ -n "$PYPI_VERSION" ]; then
-  version_tag="$PYPI_VERSION"
-elif [ -n "$TEST_PYPI_VERSION" ]; then
-  version_tag="test-$TEST_PYPI_VERSION"
-elif [[ -n "$LLAMA_STACK_DIR" || -n "$LLAMA_STACK_CLIENT_DIR" ]]; then
-  version_tag="dev"
-else
-  URL="https://pypi.org/pypi/llama-stack/json"
-  version_tag=$(curl -s $URL | jq -r '.info.version')
-fi
-
-# Add version tag to image name
-image_tag="$image_name:$version_tag"
-
-# Detect platform architecture
-ARCH=$(uname -m)
-if [ -n "$BUILD_PLATFORM" ]; then
-  CLI_ARGS+=("--platform" "$BUILD_PLATFORM")
-elif [ "$ARCH" = "arm64" ] || [ "$ARCH" = "aarch64" ]; then
-  CLI_ARGS+=("--platform" "linux/arm64")
-elif [ "$ARCH" = "x86_64" ]; then
-  CLI_ARGS+=("--platform" "linux/amd64")
-else
-  echo "Unsupported architecture: $ARCH"
-  exit 1
-fi
-
-echo "PWD: $(pwd)"
-echo "Containerfile: $TEMP_DIR/Containerfile"
-set -x
-
-$CONTAINER_BINARY build \
-  "${CLI_ARGS[@]}" \
-  -t "$image_tag" \
-  -f "$TEMP_DIR/Containerfile" \
-  "$BUILD_CONTEXT_DIR"
-
-# clean up tmp/configs
-rm -rf "$BUILD_CONTEXT_DIR/run.yaml" "$TEMP_DIR"
-set +x
-
-echo "Success!"

llama_stack/core/build_venv.sh

@@ -1,220 +0,0 @@
-#!/bin/bash
-
-# Copyright (c) Meta Platforms, Inc. and affiliates.
-# All rights reserved.
-#
-# This source code is licensed under the terms described in the LICENSE file in
-# the root directory of this source tree.
-
-LLAMA_STACK_DIR=${LLAMA_STACK_DIR:-}
-LLAMA_STACK_CLIENT_DIR=${LLAMA_STACK_CLIENT_DIR:-}
-TEST_PYPI_VERSION=${TEST_PYPI_VERSION:-}
-# This timeout (in seconds) is necessary when installing PyTorch via uv since it's likely to time out
-# Reference: https://github.com/astral-sh/uv/pull/1694
-UV_HTTP_TIMEOUT=${UV_HTTP_TIMEOUT:-500}
-UV_SYSTEM_PYTHON=${UV_SYSTEM_PYTHON:-}
-VIRTUAL_ENV=${VIRTUAL_ENV:-}
-
-set -euo pipefail
-
-# Define color codes
-RED='\033[0;31m'
-NC='\033[0m' # No Color
-
-SCRIPT_DIR=$(dirname "$(readlink -f "$0")")
-source "$SCRIPT_DIR/common.sh"
-
-# Usage function
-usage() {
-  echo "Usage: $0 --env-name <env_name> --normal-deps <pip_dependencies> [--external-provider-deps <external_provider_deps>] [--optional-deps <special_pip_deps>]"
-  echo "Example: $0 --env-name mybuild --normal-deps 'numpy pandas scipy' --external-provider-deps 'foo' --optional-deps 'bar'"
-  exit 1
-}
-
-# Parse arguments
-env_name=""
-normal_deps=""
-external_provider_deps=""
-optional_deps=""
-
-while [[ $# -gt 0 ]]; do
-  key="$1"
-  case "$key" in
-    --env-name)
-      if [[ -z "$2" || "$2" == --* ]]; then
-        echo "Error: --env-name requires a string value" >&2
-        usage
-      fi
-      env_name="$2"
-      shift 2
-      ;;
-    --normal-deps)
-      if [[ -z "$2" || "$2" == --* ]]; then
-        echo "Error: --normal-deps requires a string value" >&2
-        usage
-      fi
-      normal_deps="$2"
-      shift 2
-      ;;
-    --external-provider-deps)
-      if [[ -z "$2" || "$2" == --* ]]; then
-        echo "Error: --external-provider-deps requires a string value" >&2
-        usage
-      fi
-      external_provider_deps="$2"
-      shift 2
-      ;;
-    --optional-deps)
-      if [[ -z "$2" || "$2" == --* ]]; then
-        echo "Error: --optional-deps requires a string value" >&2
-        usage
-      fi
-      optional_deps="$2"
-      shift 2
-      ;;
-    *)
-      echo "Unknown option: $1" >&2
-      usage
-      ;;
-  esac
-done
-
-# Check required arguments
-if [[ -z "$env_name" || -z "$normal_deps" ]]; then
-  echo "Error: --env-name and --normal-deps are required." >&2
-  usage
-fi
-
-if [ -n "$LLAMA_STACK_DIR" ]; then
-  echo "Using llama-stack-dir=$LLAMA_STACK_DIR"
-fi
-if [ -n "$LLAMA_STACK_CLIENT_DIR" ]; then
-  echo "Using llama-stack-client-dir=$LLAMA_STACK_CLIENT_DIR"
-fi
-
-ENVNAME=""
-
-# pre-run checks to make sure we can proceed with the installation
-pre_run_checks() {
-  local env_name="$1"
-
-  if ! is_command_available uv; then
-    echo "uv is not installed, trying to install it."
-    if ! is_command_available pip; then
-      echo "pip is not installed, cannot automatically install 'uv'."
-      echo "Follow this link to install it:"
-      echo "https://docs.astral.sh/uv/getting-started/installation/"
-      exit 1
-    else
-      pip install uv
-    fi
-  fi
-
-  # checking if an environment with the same name already exists
-  if [ -d "$env_name" ]; then
-    echo "Environment '$env_name' already exists, re-using it."
-  fi
-}
-
-run() {
-  # Use only global variables set by flag parser
-  if [ -n "$UV_SYSTEM_PYTHON" ] || [ "$env_name" == "__system__" ]; then
-    echo "Installing dependencies in system Python environment"
-    export UV_SYSTEM_PYTHON=1
-  elif [ "$VIRTUAL_ENV" == "$env_name" ]; then
-    echo "Virtual environment $env_name is already active"
-  else
-    echo "Using virtual environment $env_name"
-    uv venv "$env_name"
-    source "$env_name/bin/activate"
-  fi
-
-  if [ -n "$TEST_PYPI_VERSION" ]; then
-    uv pip install fastapi libcst
-    uv pip install --extra-index-url https://test.pypi.org/simple/ \
-      --index-strategy unsafe-best-match \
-      llama-stack=="$TEST_PYPI_VERSION" \
-      $normal_deps
-    if [ -n "$optional_deps" ]; then
-      IFS='#' read -ra parts <<<"$optional_deps"
-      for part in "${parts[@]}"; do
-        echo "$part"
-        uv pip install $part
-      done
-    fi
-    if [ -n "$external_provider_deps" ]; then
-      IFS='#' read -ra parts <<<"$external_provider_deps"
-      for part in "${parts[@]}"; do
-        echo "$part"
-        uv pip install "$part"
-      done
-    fi
-  else
-    if [ -n "$LLAMA_STACK_DIR" ]; then
-      # only warn if DIR does not start with "git+"
-      if [ ! -d "$LLAMA_STACK_DIR" ] && [[ "$LLAMA_STACK_DIR" != git+* ]]; then
-        printf "${RED}Warning: LLAMA_STACK_DIR is set but directory does not exist: %s${NC}\n" "$LLAMA_STACK_DIR" >&2
-        exit 1
-      fi
-      printf "Installing from LLAMA_STACK_DIR: %s\n"  "$LLAMA_STACK_DIR"
-      # editable only if LLAMA_STACK_DIR does not start with "git+"
-      if [[ "$LLAMA_STACK_DIR" != git+* ]]; then
-        EDITABLE="-e"
-      else
-        EDITABLE=""
-      fi
-      uv pip install --no-cache-dir $EDITABLE "$LLAMA_STACK_DIR"
-    else
-      uv pip install --no-cache-dir llama-stack
-    fi
-
-    if [ -n "$LLAMA_STACK_CLIENT_DIR" ]; then
-      # only warn if DIR does not start with "git+"
-      if [ ! -d "$LLAMA_STACK_CLIENT_DIR" ] && [[ "$LLAMA_STACK_CLIENT_DIR" != git+* ]]; then
-        printf "${RED}Warning: LLAMA_STACK_CLIENT_DIR is set but directory does not exist: %s${NC}\n" "$LLAMA_STACK_CLIENT_DIR" >&2
-        exit 1
-      fi
-      printf "Installing from LLAMA_STACK_CLIENT_DIR: %s\n" "$LLAMA_STACK_CLIENT_DIR"
-      # editable only if LLAMA_STACK_CLIENT_DIR does not start with "git+"
-      if [[ "$LLAMA_STACK_CLIENT_DIR" != git+* ]]; then
-        EDITABLE="-e"
-      else
-        EDITABLE=""
-      fi
-      uv pip install --no-cache-dir $EDITABLE "$LLAMA_STACK_CLIENT_DIR"
-    fi
-
-    printf "Installing pip dependencies\n"
-    uv pip install $normal_deps
-    if [ -n "$optional_deps" ]; then
-      IFS='#' read -ra parts <<<"$optional_deps"
-      for part in "${parts[@]}"; do
-        echo "Installing special provider module: $part"
-        uv pip install $part
-      done
-    fi
-    if [ -n "$external_provider_deps" ]; then
-      IFS='#' read -ra parts <<<"$external_provider_deps"
-      for part in "${parts[@]}"; do
-        echo "Installing external provider module: $part"
-        uv pip install "$part"
-        echo "Getting provider spec for module: $part and installing dependencies"
-        package_name=$(echo "$part" | sed 's/[<>=!].*//')
-        python3 -c "
-import importlib
-import sys
-try:
-    module = importlib.import_module(f'$package_name.provider')
-    spec = module.get_provider_spec()
-    if hasattr(spec, 'pip_packages') and spec.pip_packages:
-        print('\\n'.join(spec.pip_packages))
-except Exception as e:
-    print(f'Error getting provider spec for $package_name: {e}', file=sys.stderr)
-" | uv pip install -r -
-      done
-    fi
-  fi
-}
-
-pre_run_checks "$env_name"
-run

llama_stack/core/configure.py

@@ -159,6 +159,37 @@ def upgrade_from_routing_table(
     config_dict["apis"] = config_dict["apis_to_serve"]
     config_dict.pop("apis_to_serve", None)
 
+    # Add default storage config if not present
+    if "storage" not in config_dict:
+        config_dict["storage"] = {
+            "backends": {
+                "kv_default": {
+                    "type": "kv_sqlite",
+                    "db_path": "~/.llama/kvstore.db",
+                },
+                "sql_default": {
+                    "type": "sql_sqlite",
+                    "db_path": "~/.llama/sql_store.db",
+                },
+            },
+            "stores": {
+                "metadata": {
+                    "namespace": "registry",
+                    "backend": "kv_default",
+                },
+                "inference": {
+                    "table_name": "inference_store",
+                    "backend": "sql_default",
+                    "max_write_queue_size": 10000,
+                    "num_writers": 4,
+                },
+                "conversations": {
+                    "table_name": "openai_conversations",
+                    "backend": "sql_default",
+                },
+            },
+        }
+
     return config_dict
 
 

llama_stack/core/conversations/conversations.py

@@ -4,7 +4,6 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-import os
 import secrets
 import time
 from typing import Any
@@ -21,16 +20,11 @@ from llama_stack.apis.conversations.conversations import (
     Conversations,
     Metadata,
 )
-from llama_stack.core.datatypes import AccessRule
-from llama_stack.core.utils.config_dirs import DISTRIBS_BASE_DIR
+from llama_stack.core.datatypes import AccessRule, StackRunConfig
 from llama_stack.log import get_logger
 from llama_stack.providers.utils.sqlstore.api import ColumnDefinition, ColumnType
 from llama_stack.providers.utils.sqlstore.authorized_sqlstore import AuthorizedSqlStore
-from llama_stack.providers.utils.sqlstore.sqlstore import (
-    SqliteSqlStoreConfig,
-    SqlStoreConfig,
-    sqlstore_impl,
-)
+from llama_stack.providers.utils.sqlstore.sqlstore import sqlstore_impl
 
 logger = get_logger(name=__name__, category="openai_conversations")
 
@@ -38,13 +32,11 @@ logger = get_logger(name=__name__, category="openai_conversations")
 class ConversationServiceConfig(BaseModel):
     """Configuration for the built-in conversation service.
 
-    :param conversations_store: SQL store configuration for conversations (defaults to SQLite)
+    :param run_config: Stack run configuration for resolving persistence
     :param policy: Access control rules
     """
 
-    conversations_store: SqlStoreConfig = SqliteSqlStoreConfig(
-        db_path=(DISTRIBS_BASE_DIR / "conversations.db").as_posix()
-    )
+    run_config: StackRunConfig
     policy: list[AccessRule] = []
 
 
@@ -63,14 +55,16 @@ class ConversationServiceImpl(Conversations):
         self.deps = deps
         self.policy = config.policy
 
-        base_sql_store = sqlstore_impl(config.conversations_store)
+        # Use conversations store reference from run config
+        conversations_ref = config.run_config.storage.stores.conversations
+        if not conversations_ref:
+            raise ValueError("storage.stores.conversations must be configured in run config")
+
+        base_sql_store = sqlstore_impl(conversations_ref)
         self.sql_store = AuthorizedSqlStore(base_sql_store, self.policy)
 
     async def initialize(self) -> None:
         """Initialize the store and create tables."""
-        if isinstance(self.config.conversations_store, SqliteSqlStoreConfig):
-            os.makedirs(os.path.dirname(self.config.conversations_store.db_path), exist_ok=True)
-
         await self.sql_store.create_table(
             "openai_conversations",
             {

llama_stack/core/datatypes.py

@@ -23,12 +23,15 @@ from llama_stack.apis.scoring import Scoring
 from llama_stack.apis.scoring_functions import ScoringFn, ScoringFnInput
 from llama_stack.apis.shields import Shield, ShieldInput
 from llama_stack.apis.tools import ToolGroup, ToolGroupInput, ToolRuntime
-from llama_stack.apis.vector_dbs import VectorDB, VectorDBInput
 from llama_stack.apis.vector_io import VectorIO
+from llama_stack.apis.vector_stores import VectorStore, VectorStoreInput
 from llama_stack.core.access_control.datatypes import AccessRule
+from llama_stack.core.storage.datatypes import (
+    KVStoreReference,
+    StorageBackendType,
+    StorageConfig,
+)
 from llama_stack.providers.datatypes import Api, ProviderSpec
-from llama_stack.providers.utils.kvstore.config import KVStoreConfig, SqliteKVStoreConfig
-from llama_stack.providers.utils.sqlstore.sqlstore import SqlStoreConfig
 
 LLAMA_STACK_BUILD_CONFIG_VERSION = 2
 LLAMA_STACK_RUN_CONFIG_VERSION = 2
@@ -68,7 +71,7 @@ class ShieldWithOwner(Shield, ResourceWithOwner):
     pass
 
 
-class VectorDBWithOwner(VectorDB, ResourceWithOwner):
+class VectorStoreWithOwner(VectorStore, ResourceWithOwner):
     pass
 
 
@@ -88,12 +91,12 @@ class ToolGroupWithOwner(ToolGroup, ResourceWithOwner):
     pass
 
 
-RoutableObject = Model | Shield | VectorDB | Dataset | ScoringFn | Benchmark | ToolGroup
+RoutableObject = Model | Shield | VectorStore | Dataset | ScoringFn | Benchmark | ToolGroup
 
 RoutableObjectWithProvider = Annotated[
     ModelWithOwner
     | ShieldWithOwner
-    | VectorDBWithOwner
+    | VectorStoreWithOwner
     | DatasetWithOwner
     | ScoringFnWithOwner
     | BenchmarkWithOwner
@@ -351,12 +354,32 @@ class AuthenticationRequiredError(Exception):
     pass
 
 
+class QualifiedModel(BaseModel):
+    """A qualified model identifier, consisting of a provider ID and a model ID."""
+
+    provider_id: str
+    model_id: str
+
+
+class VectorStoresConfig(BaseModel):
+    """Configuration for vector stores in the stack."""
+
+    default_provider_id: str | None = Field(
+        default=None,
+        description="ID of the vector_io provider to use as default when multiple providers are available and none is specified.",
+    )
+    default_embedding_model: QualifiedModel | None = Field(
+        default=None,
+        description="Default embedding model configuration for vector stores.",
+    )
+
+
 class QuotaPeriod(StrEnum):
     DAY = "day"
 
 
 class QuotaConfig(BaseModel):
-    kvstore: SqliteKVStoreConfig = Field(description="Config for KV store backend (SQLite only for now)")
+    kvstore: KVStoreReference = Field(description="Config for KV store backend (SQLite only for now)")
     anonymous_max_requests: int = Field(default=100, description="Max requests for unauthenticated clients per period")
     authenticated_max_requests: int = Field(
         default=1000, description="Max requests for authenticated clients per period"
@@ -399,6 +422,18 @@ def process_cors_config(cors_config: bool | CORSConfig | None) -> CORSConfig | N
     raise ValueError(f"Expected bool or CORSConfig, got {type(cors_config).__name__}")
 
 
+class RegisteredResources(BaseModel):
+    """Registry of resources available in the distribution."""
+
+    models: list[ModelInput] = Field(default_factory=list)
+    shields: list[ShieldInput] = Field(default_factory=list)
+    vector_stores: list[VectorStoreInput] = Field(default_factory=list)
+    datasets: list[DatasetInput] = Field(default_factory=list)
+    scoring_fns: list[ScoringFnInput] = Field(default_factory=list)
+    benchmarks: list[BenchmarkInput] = Field(default_factory=list)
+    tool_groups: list[ToolGroupInput] = Field(default_factory=list)
+
+
 class ServerConfig(BaseModel):
     port: int = Field(
         default=8321,
@@ -438,18 +473,6 @@ class ServerConfig(BaseModel):
     )
 
 
-class InferenceStoreConfig(BaseModel):
-    sql_store_config: SqlStoreConfig
-    max_write_queue_size: int = Field(default=10000, description="Max queued writes for inference store")
-    num_writers: int = Field(default=4, description="Number of concurrent background writers")
-
-
-class ResponsesStoreConfig(BaseModel):
-    sql_store_config: SqlStoreConfig
-    max_write_queue_size: int = Field(default=10000, description="Max queued writes for responses store")
-    num_writers: int = Field(default=4, description="Number of concurrent background writers")
-
-
 class StackRunConfig(BaseModel):
     version: int = LLAMA_STACK_RUN_CONFIG_VERSION
 
@@ -476,37 +499,15 @@ One or more providers to use for each API. The same provider_type (e.g., meta-re
 can be instantiated multiple times (with different configs) if necessary.
 """,
     )
-    metadata_store: KVStoreConfig | None = Field(
-        default=None,
-        description="""
-Configuration for the persistence store used by the distribution registry. If not specified,
-a default SQLite store will be used.""",
+    storage: StorageConfig = Field(
+        description="Catalog of named storage backends and references available to the stack",
     )
 
-    inference_store: InferenceStoreConfig | SqlStoreConfig | None = Field(
-        default=None,
-        description="""
-Configuration for the persistence store used by the inference API. Can be either a
-InferenceStoreConfig (with queue tuning parameters) or a SqlStoreConfig (deprecated).
-If not specified, a default SQLite store will be used.""",
+    registered_resources: RegisteredResources = Field(
+        default_factory=RegisteredResources,
+        description="Registry of resources available in the distribution",
     )
 
-    conversations_store: SqlStoreConfig | None = Field(
-        default=None,
-        description="""
-Configuration for the persistence store used by the conversations API.
-If not specified, a default SQLite store will be used.""",
-    )
-
-    # registry of "resources" in the distribution
-    models: list[ModelInput] = Field(default_factory=list)
-    shields: list[ShieldInput] = Field(default_factory=list)
-    vector_dbs: list[VectorDBInput] = Field(default_factory=list)
-    datasets: list[DatasetInput] = Field(default_factory=list)
-    scoring_fns: list[ScoringFnInput] = Field(default_factory=list)
-    benchmarks: list[BenchmarkInput] = Field(default_factory=list)
-    tool_groups: list[ToolGroupInput] = Field(default_factory=list)
-
     logging: LoggingConfig | None = Field(default=None, description="Configuration for Llama Stack Logging")
 
     telemetry: TelemetryConfig = Field(default_factory=TelemetryConfig, description="Configuration for telemetry")
@@ -526,6 +527,11 @@ If not specified, a default SQLite store will be used.""",
         description="Path to directory containing external API implementations. The APIs code and dependencies must be installed on the system.",
     )
 
+    vector_stores: VectorStoresConfig | None = Field(
+        default=None,
+        description="Configuration for vector stores, including default embedding model",
+    )
+
     @field_validator("external_providers_dir")
     @classmethod
     def validate_external_providers_dir(cls, v):
@@ -535,6 +541,49 @@ If not specified, a default SQLite store will be used.""",
             return Path(v)
         return v
 
+    @model_validator(mode="after")
+    def validate_server_stores(self) -> "StackRunConfig":
+        backend_map = self.storage.backends
+        stores = self.storage.stores
+        kv_backends = {
+            name
+            for name, cfg in backend_map.items()
+            if cfg.type
+            in {
+                StorageBackendType.KV_REDIS,
+                StorageBackendType.KV_SQLITE,
+                StorageBackendType.KV_POSTGRES,
+                StorageBackendType.KV_MONGODB,
+            }
+        }
+        sql_backends = {
+            name
+            for name, cfg in backend_map.items()
+            if cfg.type in {StorageBackendType.SQL_SQLITE, StorageBackendType.SQL_POSTGRES}
+        }
+
+        def _ensure_backend(reference, expected_set, store_name: str) -> None:
+            if reference is None:
+                return
+            backend_name = reference.backend
+            if backend_name not in backend_map:
+                raise ValueError(
+                    f"{store_name} references unknown backend '{backend_name}'. "
+                    f"Available backends: {sorted(backend_map)}"
+                )
+            if backend_name not in expected_set:
+                raise ValueError(
+                    f"{store_name} references backend '{backend_name}' of type "
+                    f"'{backend_map[backend_name].type.value}', but a backend of type "
+                    f"{'kv_*' if expected_set is kv_backends else 'sql_*'} is required."
+                )
+
+        _ensure_backend(stores.metadata, kv_backends, "storage.stores.metadata")
+        _ensure_backend(stores.inference, sql_backends, "storage.stores.inference")
+        _ensure_backend(stores.conversations, sql_backends, "storage.stores.conversations")
+        _ensure_backend(stores.responses, sql_backends, "storage.stores.responses")
+        return self
+
 
 class BuildConfig(BaseModel):
     version: int = LLAMA_STACK_BUILD_CONFIG_VERSION

llama_stack/core/distribution.py

@@ -63,6 +63,10 @@ def builtin_automatically_routed_apis() -> list[AutoRoutedApiInfo]:
             routing_table_api=Api.tool_groups,
             router_api=Api.tool_runtime,
         ),
+        AutoRoutedApiInfo(
+            routing_table_api=Api.vector_stores,
+            router_api=Api.vector_io,
+        ),
     ]
 
 

llama_stack/core/library_client.py

@@ -278,7 +278,7 @@ class AsyncLlamaStackAsLibraryClient(AsyncLlamaStackClient):
             else:
                 prefix = "!" if in_notebook() else ""
                 cprint(
-                    f"Please run:\n\n{prefix}llama stack build --distro {self.config_path_or_distro_name} --image-type venv\n\n",
+                    f"Please run:\n\n{prefix}llama stack list-deps {self.config_path_or_distro_name} | xargs -L1 uv pip install\n\n",
                     "yellow",
                     file=sys.stderr,
                 )

llama_stack/core/prompts/prompts.py

@@ -11,9 +11,8 @@ from pydantic import BaseModel
 
 from llama_stack.apis.prompts import ListPromptsResponse, Prompt, Prompts
 from llama_stack.core.datatypes import StackRunConfig
-from llama_stack.core.utils.config_dirs import DISTRIBS_BASE_DIR
+from llama_stack.core.storage.datatypes import KVStoreReference
 from llama_stack.providers.utils.kvstore import KVStore, kvstore_impl
-from llama_stack.providers.utils.kvstore.config import SqliteKVStoreConfig
 
 
 class PromptServiceConfig(BaseModel):
@@ -41,10 +40,12 @@ class PromptServiceImpl(Prompts):
         self.kvstore: KVStore
 
     async def initialize(self) -> None:
-        kvstore_config = SqliteKVStoreConfig(
-            db_path=(DISTRIBS_BASE_DIR / self.config.run_config.image_name / "prompts.db").as_posix()
-        )
-        self.kvstore = await kvstore_impl(kvstore_config)
+        # Use metadata store backend with prompts-specific namespace
+        metadata_ref = self.config.run_config.storage.stores.metadata
+        if not metadata_ref:
+            raise ValueError("storage.stores.metadata must be configured in run config")
+        prompts_ref = KVStoreReference(namespace="prompts", backend=metadata_ref.backend)
+        self.kvstore = await kvstore_impl(prompts_ref)
 
     def _get_default_key(self, prompt_id: str) -> str:
         """Get the KVStore key that stores the default version number."""

llama_stack/core/resolver.py

@@ -30,6 +30,7 @@ from llama_stack.apis.shields import Shields
 from llama_stack.apis.telemetry import Telemetry
 from llama_stack.apis.tools import ToolGroups, ToolRuntime
 from llama_stack.apis.vector_io import VectorIO
+from llama_stack.apis.vector_stores import VectorStore
 from llama_stack.apis.version import LLAMA_STACK_API_V1ALPHA
 from llama_stack.core.client import get_client_impl
 from llama_stack.core.datatypes import (
@@ -81,6 +82,7 @@ def api_protocol_map(external_apis: dict[Api, ExternalApiSpec] | None = None) ->
         Api.inspect: Inspect,
         Api.batches: Batches,
         Api.vector_io: VectorIO,
+        Api.vector_stores: VectorStore,
         Api.models: Models,
         Api.safety: Safety,
         Api.shields: Shields,

llama_stack/core/routers/__init__.py

@@ -6,7 +6,10 @@
 
 from typing import Any
 
-from llama_stack.core.datatypes import AccessRule, RoutedProtocol
+from llama_stack.core.datatypes import (
+    AccessRule,
+    RoutedProtocol,
+)
 from llama_stack.core.stack import StackRunConfig
 from llama_stack.core.store import DistributionRegistry
 from llama_stack.providers.datatypes import Api, RoutingTable
@@ -26,6 +29,7 @@ async def get_routing_table_impl(
     from ..routing_tables.scoring_functions import ScoringFunctionsRoutingTable
     from ..routing_tables.shields import ShieldsRoutingTable
     from ..routing_tables.toolgroups import ToolGroupsRoutingTable
+    from ..routing_tables.vector_stores import VectorStoresRoutingTable
 
     api_to_tables = {
         "models": ModelsRoutingTable,
@@ -34,6 +38,7 @@ async def get_routing_table_impl(
         "scoring_functions": ScoringFunctionsRoutingTable,
         "benchmarks": BenchmarksRoutingTable,
         "tool_groups": ToolGroupsRoutingTable,
+        "vector_stores": VectorStoresRoutingTable,
     }
 
     if api.value not in api_to_tables:
@@ -76,14 +81,21 @@ async def get_auto_router_impl(
                 api_to_dep_impl[dep_name] = deps[dep_api]
 
     # TODO: move pass configs to routers instead
-    if api == Api.inference and run_config.inference_store:
+    if api == Api.inference:
+        inference_ref = run_config.storage.stores.inference
+        if not inference_ref:
+            raise ValueError("storage.stores.inference must be configured in run config")
+
         inference_store = InferenceStore(
-            config=run_config.inference_store,
+            reference=inference_ref,
             policy=policy,
         )
         await inference_store.initialize()
         api_to_dep_impl["store"] = inference_store
 
+    elif api == Api.vector_io:
+        api_to_dep_impl["vector_stores_config"] = run_config.vector_stores
+
     impl = api_to_routers[api.value](routing_table, **api_to_dep_impl)
     await impl.initialize()
     return impl

llama_stack/core/routers/tool_runtime.py

@@ -8,16 +8,8 @@ from typing import Any
 
 from llama_stack.apis.common.content_types import (
     URL,
-    InterleavedContent,
-)
-from llama_stack.apis.tools import (
-    ListToolDefsResponse,
-    RAGDocument,
-    RAGQueryConfig,
-    RAGQueryResult,
-    RAGToolRuntime,
-    ToolRuntime,
 )
+from llama_stack.apis.tools import ListToolDefsResponse, ToolRuntime
 from llama_stack.log import get_logger
 
 from ..routing_tables.toolgroups import ToolGroupsRoutingTable
@@ -26,36 +18,6 @@ logger = get_logger(name=__name__, category="core::routers")
 
 
 class ToolRuntimeRouter(ToolRuntime):
-    class RagToolImpl(RAGToolRuntime):
-        def __init__(
-            self,
-            routing_table: ToolGroupsRoutingTable,
-        ) -> None:
-            logger.debug("Initializing ToolRuntimeRouter.RagToolImpl")
-            self.routing_table = routing_table
-
-        async def query(
-            self,
-            content: InterleavedContent,
-            vector_db_ids: list[str],
-            query_config: RAGQueryConfig | None = None,
-        ) -> RAGQueryResult:
-            logger.debug(f"ToolRuntimeRouter.RagToolImpl.query: {vector_db_ids}")
-            provider = await self.routing_table.get_provider_impl("knowledge_search")
-            return await provider.query(content, vector_db_ids, query_config)
-
-        async def insert(
-            self,
-            documents: list[RAGDocument],
-            vector_db_id: str,
-            chunk_size_in_tokens: int = 512,
-        ) -> None:
-            logger.debug(
-                f"ToolRuntimeRouter.RagToolImpl.insert: {vector_db_id}, {len(documents)} documents, chunk_size={chunk_size_in_tokens}"
-            )
-            provider = await self.routing_table.get_provider_impl("insert_into_memory")
-            return await provider.insert(documents, vector_db_id, chunk_size_in_tokens)
-
     def __init__(
         self,
         routing_table: ToolGroupsRoutingTable,
@@ -63,11 +25,6 @@ class ToolRuntimeRouter(ToolRuntime):
         logger.debug("Initializing ToolRuntimeRouter")
         self.routing_table = routing_table
 
-        # HACK ALERT this should be in sync with "get_all_api_endpoints()"
-        self.rag_tool = self.RagToolImpl(routing_table)
-        for method in ("query", "insert"):
-            setattr(self, f"rag_tool.{method}", getattr(self.rag_tool, method))
-
     async def initialize(self) -> None:
         logger.debug("ToolRuntimeRouter.initialize")
         pass

llama_stack/core/routers/vector_io.py

@@ -31,6 +31,7 @@ from llama_stack.apis.vector_io import (
     VectorStoreObject,
     VectorStoreSearchResponsePage,
 )
+from llama_stack.core.datatypes import VectorStoresConfig
 from llama_stack.log import get_logger
 from llama_stack.providers.datatypes import HealthResponse, HealthStatus, RoutingTable
 
@@ -43,9 +44,11 @@ class VectorIORouter(VectorIO):
     def __init__(
         self,
         routing_table: RoutingTable,
+        vector_stores_config: VectorStoresConfig | None = None,
     ) -> None:
         logger.debug("Initializing VectorIORouter")
         self.routing_table = routing_table
+        self.vector_stores_config = vector_stores_config
 
     async def initialize(self) -> None:
         logger.debug("VectorIORouter.initialize")
@@ -68,25 +71,6 @@ class VectorIORouter(VectorIO):
 
         raise ValueError(f"Embedding model '{embedding_model_id}' not found or not an embedding model")
 
-    async def register_vector_db(
-        self,
-        vector_db_id: str,
-        embedding_model: str,
-        embedding_dimension: int | None = 384,
-        provider_id: str | None = None,
-        vector_db_name: str | None = None,
-        provider_vector_db_id: str | None = None,
-    ) -> None:
-        logger.debug(f"VectorIORouter.register_vector_db: {vector_db_id}, {embedding_model}")
-        await self.routing_table.register_vector_db(
-            vector_db_id,
-            embedding_model,
-            embedding_dimension,
-            provider_id,
-            vector_db_name,
-            provider_vector_db_id,
-        )
-
     async def insert_chunks(
         self,
         vector_db_id: str,
@@ -122,6 +106,17 @@ class VectorIORouter(VectorIO):
         embedding_dimension = extra.get("embedding_dimension")
         provider_id = extra.get("provider_id")
 
+        # Use default embedding model if not specified
+        if (
+            embedding_model is None
+            and self.vector_stores_config
+            and self.vector_stores_config.default_embedding_model is not None
+        ):
+            # Construct the full model ID with provider prefix
+            embedding_provider_id = self.vector_stores_config.default_embedding_model.provider_id
+            model_id = self.vector_stores_config.default_embedding_model.model_id
+            embedding_model = f"{embedding_provider_id}/{model_id}"
+
         if embedding_model is not None and embedding_dimension is None:
             embedding_dimension = await self._get_embedding_model_dimension(embedding_model)
 
@@ -132,28 +127,41 @@ class VectorIORouter(VectorIO):
                 raise ValueError("No vector_io providers available")
             if num_providers > 1:
                 available_providers = list(self.routing_table.impls_by_provider_id.keys())
-                raise ValueError(
-                    f"Multiple vector_io providers available. Please specify provider_id in extra_body. "
-                    f"Available providers: {available_providers}"
-                )
-            provider_id = list(self.routing_table.impls_by_provider_id.keys())[0]
+                # Use default configured provider
+                if self.vector_stores_config and self.vector_stores_config.default_provider_id:
+                    default_provider = self.vector_stores_config.default_provider_id
+                    if default_provider in available_providers:
+                        provider_id = default_provider
+                        logger.debug(f"Using configured default vector store provider: {provider_id}")
+                    else:
+                        raise ValueError(
+                            f"Configured default vector store provider '{default_provider}' not found. "
+                            f"Available providers: {available_providers}"
+                        )
+                else:
+                    raise ValueError(
+                        f"Multiple vector_io providers available. Please specify provider_id in extra_body. "
+                        f"Available providers: {available_providers}"
+                    )
+            else:
+                provider_id = list(self.routing_table.impls_by_provider_id.keys())[0]
 
-        vector_db_id = f"vs_{uuid.uuid4()}"
-        registered_vector_db = await self.routing_table.register_vector_db(
-            vector_db_id=vector_db_id,
+        vector_store_id = f"vs_{uuid.uuid4()}"
+        registered_vector_store = await self.routing_table.register_vector_store(
+            vector_store_id=vector_store_id,
             embedding_model=embedding_model,
             embedding_dimension=embedding_dimension,
             provider_id=provider_id,
-            provider_vector_db_id=vector_db_id,
-            vector_db_name=params.name,
+            provider_vector_store_id=vector_store_id,
+            vector_store_name=params.name,
         )
-        provider = await self.routing_table.get_provider_impl(registered_vector_db.identifier)
+        provider = await self.routing_table.get_provider_impl(registered_vector_store.identifier)
 
-        # Update model_extra with registered values so provider uses the already-registered vector_db
+        # Update model_extra with registered values so provider uses the already-registered vector_store
         if params.model_extra is None:
             params.model_extra = {}
-        params.model_extra["provider_vector_db_id"] = registered_vector_db.provider_resource_id
-        params.model_extra["provider_id"] = registered_vector_db.provider_id
+        params.model_extra["provider_vector_store_id"] = registered_vector_store.provider_resource_id
+        params.model_extra["provider_id"] = registered_vector_store.provider_id
         if embedding_model is not None:
             params.model_extra["embedding_model"] = embedding_model
         if embedding_dimension is not None:
@@ -171,15 +179,15 @@ class VectorIORouter(VectorIO):
         logger.debug(f"VectorIORouter.openai_list_vector_stores: limit={limit}")
         # Route to default provider for now - could aggregate from all providers in the future
         # call retrieve on each vector dbs to get list of vector stores
-        vector_dbs = await self.routing_table.get_all_with_type("vector_db")
+        vector_stores = await self.routing_table.get_all_with_type("vector_store")
         all_stores = []
-        for vector_db in vector_dbs:
+        for vector_store in vector_stores:
             try:
-                provider = await self.routing_table.get_provider_impl(vector_db.identifier)
-                vector_store = await provider.openai_retrieve_vector_store(vector_db.identifier)
+                provider = await self.routing_table.get_provider_impl(vector_store.identifier)
+                vector_store = await provider.openai_retrieve_vector_store(vector_store.identifier)
                 all_stores.append(vector_store)
             except Exception as e:
-                logger.error(f"Error retrieving vector store {vector_db.identifier}: {e}")
+                logger.error(f"Error retrieving vector store {vector_store.identifier}: {e}")
                 continue
 
         # Sort by created_at
@@ -243,8 +251,7 @@ class VectorIORouter(VectorIO):
         vector_store_id: str,
     ) -> VectorStoreDeleteResponse:
         logger.debug(f"VectorIORouter.openai_delete_vector_store: {vector_store_id}")
-        provider = await self.routing_table.get_provider_impl(vector_store_id)
-        return await provider.openai_delete_vector_store(vector_store_id)
+        return await self.routing_table.openai_delete_vector_store(vector_store_id)
 
     async def openai_search_vector_store(
         self,

llama_stack/core/routing_tables/common.py

@@ -41,7 +41,7 @@ async def register_object_with_provider(obj: RoutableObject, p: Any) -> Routable
     elif api == Api.safety:
         return await p.register_shield(obj)
     elif api == Api.vector_io:
-        return await p.register_vector_db(obj)
+        return await p.register_vector_store(obj)
     elif api == Api.datasetio:
         return await p.register_dataset(obj)
     elif api == Api.scoring:
@@ -57,7 +57,7 @@ async def register_object_with_provider(obj: RoutableObject, p: Any) -> Routable
 async def unregister_object_from_provider(obj: RoutableObject, p: Any) -> None:
     api = get_impl_api(p)
     if api == Api.vector_io:
-        return await p.unregister_vector_db(obj.identifier)
+        return await p.unregister_vector_store(obj.identifier)
     elif api == Api.inference:
         return await p.unregister_model(obj.identifier)
     elif api == Api.safety:
@@ -108,7 +108,7 @@ class CommonRoutingTableImpl(RoutingTable):
             elif api == Api.safety:
                 p.shield_store = self
             elif api == Api.vector_io:
-                p.vector_db_store = self
+                p.vector_store_store = self
             elif api == Api.datasetio:
                 p.dataset_store = self
             elif api == Api.scoring:
@@ -134,12 +134,15 @@ class CommonRoutingTableImpl(RoutingTable):
         from .scoring_functions import ScoringFunctionsRoutingTable
         from .shields import ShieldsRoutingTable
         from .toolgroups import ToolGroupsRoutingTable
+        from .vector_stores import VectorStoresRoutingTable
 
         def apiname_object():
             if isinstance(self, ModelsRoutingTable):
                 return ("Inference", "model")
             elif isinstance(self, ShieldsRoutingTable):
                 return ("Safety", "shield")
+            elif isinstance(self, VectorStoresRoutingTable):
+                return ("VectorIO", "vector_store")
             elif isinstance(self, DatasetsRoutingTable):
                 return ("DatasetIO", "dataset")
             elif isinstance(self, ScoringFunctionsRoutingTable):

llama_stack/core/routing_tables/vector_stores.py

@@ -0,0 +1,292 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from typing import Any
+
+from llama_stack.apis.common.errors import ModelNotFoundError, ModelTypeError
+from llama_stack.apis.models import ModelType
+from llama_stack.apis.resource import ResourceType
+
+# Removed VectorStores import to avoid exposing public API
+from llama_stack.apis.vector_io.vector_io import (
+    SearchRankingOptions,
+    VectorStoreChunkingStrategy,
+    VectorStoreDeleteResponse,
+    VectorStoreFileContentsResponse,
+    VectorStoreFileDeleteResponse,
+    VectorStoreFileObject,
+    VectorStoreFileStatus,
+    VectorStoreObject,
+    VectorStoreSearchResponsePage,
+)
+from llama_stack.core.datatypes import (
+    VectorStoreWithOwner,
+)
+from llama_stack.log import get_logger
+
+from .common import CommonRoutingTableImpl, lookup_model
+
+logger = get_logger(name=__name__, category="core::routing_tables")
+
+
+class VectorStoresRoutingTable(CommonRoutingTableImpl):
+    """Internal routing table for vector_store operations.
+
+    Does not inherit from VectorStores to avoid exposing public API endpoints.
+    Only provides internal routing functionality for VectorIORouter.
+    """
+
+    # Internal methods only - no public API exposure
+
+    async def register_vector_store(
+        self,
+        vector_store_id: str,
+        embedding_model: str,
+        embedding_dimension: int | None = 384,
+        provider_id: str | None = None,
+        provider_vector_store_id: str | None = None,
+        vector_store_name: str | None = None,
+    ) -> Any:
+        if provider_id is None:
+            if len(self.impls_by_provider_id) > 0:
+                provider_id = list(self.impls_by_provider_id.keys())[0]
+                if len(self.impls_by_provider_id) > 1:
+                    logger.warning(
+                        f"No provider specified and multiple providers available. Arbitrarily selected the first provider {provider_id}."
+                    )
+            else:
+                raise ValueError("No provider available. Please configure a vector_io provider.")
+        model = await lookup_model(self, embedding_model)
+        if model is None:
+            raise ModelNotFoundError(embedding_model)
+        if model.model_type != ModelType.embedding:
+            raise ModelTypeError(embedding_model, model.model_type, ModelType.embedding)
+
+        vector_store = VectorStoreWithOwner(
+            identifier=vector_store_id,
+            type=ResourceType.vector_store.value,
+            provider_id=provider_id,
+            provider_resource_id=provider_vector_store_id,
+            embedding_model=embedding_model,
+            embedding_dimension=embedding_dimension,
+            vector_store_name=vector_store_name,
+        )
+        await self.register_object(vector_store)
+        return vector_store
+
+    async def openai_retrieve_vector_store(
+        self,
+        vector_store_id: str,
+    ) -> VectorStoreObject:
+        await self.assert_action_allowed("read", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_retrieve_vector_store(vector_store_id)
+
+    async def openai_update_vector_store(
+        self,
+        vector_store_id: str,
+        name: str | None = None,
+        expires_after: dict[str, Any] | None = None,
+        metadata: dict[str, Any] | None = None,
+    ) -> VectorStoreObject:
+        await self.assert_action_allowed("update", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_update_vector_store(
+            vector_store_id=vector_store_id,
+            name=name,
+            expires_after=expires_after,
+            metadata=metadata,
+        )
+
+    async def openai_delete_vector_store(
+        self,
+        vector_store_id: str,
+    ) -> VectorStoreDeleteResponse:
+        await self.assert_action_allowed("delete", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        result = await provider.openai_delete_vector_store(vector_store_id)
+        await self.unregister_vector_store(vector_store_id)
+        return result
+
+    async def unregister_vector_store(self, vector_store_id: str) -> None:
+        """Remove the vector store from the routing table registry."""
+        try:
+            vector_store_obj = await self.get_object_by_identifier("vector_store", vector_store_id)
+            if vector_store_obj:
+                await self.unregister_object(vector_store_obj)
+        except Exception as e:
+            # Log the error but don't fail the operation
+            logger.warning(f"Failed to unregister vector store {vector_store_id} from routing table: {e}")
+
+    async def openai_search_vector_store(
+        self,
+        vector_store_id: str,
+        query: str | list[str],
+        filters: dict[str, Any] | None = None,
+        max_num_results: int | None = 10,
+        ranking_options: SearchRankingOptions | None = None,
+        rewrite_query: bool | None = False,
+        search_mode: str | None = "vector",
+    ) -> VectorStoreSearchResponsePage:
+        await self.assert_action_allowed("read", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_search_vector_store(
+            vector_store_id=vector_store_id,
+            query=query,
+            filters=filters,
+            max_num_results=max_num_results,
+            ranking_options=ranking_options,
+            rewrite_query=rewrite_query,
+            search_mode=search_mode,
+        )
+
+    async def openai_attach_file_to_vector_store(
+        self,
+        vector_store_id: str,
+        file_id: str,
+        attributes: dict[str, Any] | None = None,
+        chunking_strategy: VectorStoreChunkingStrategy | None = None,
+    ) -> VectorStoreFileObject:
+        await self.assert_action_allowed("update", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_attach_file_to_vector_store(
+            vector_store_id=vector_store_id,
+            file_id=file_id,
+            attributes=attributes,
+            chunking_strategy=chunking_strategy,
+        )
+
+    async def openai_list_files_in_vector_store(
+        self,
+        vector_store_id: str,
+        limit: int | None = 20,
+        order: str | None = "desc",
+        after: str | None = None,
+        before: str | None = None,
+        filter: VectorStoreFileStatus | None = None,
+    ) -> list[VectorStoreFileObject]:
+        await self.assert_action_allowed("read", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_list_files_in_vector_store(
+            vector_store_id=vector_store_id,
+            limit=limit,
+            order=order,
+            after=after,
+            before=before,
+            filter=filter,
+        )
+
+    async def openai_retrieve_vector_store_file(
+        self,
+        vector_store_id: str,
+        file_id: str,
+    ) -> VectorStoreFileObject:
+        await self.assert_action_allowed("read", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_retrieve_vector_store_file(
+            vector_store_id=vector_store_id,
+            file_id=file_id,
+        )
+
+    async def openai_retrieve_vector_store_file_contents(
+        self,
+        vector_store_id: str,
+        file_id: str,
+    ) -> VectorStoreFileContentsResponse:
+        await self.assert_action_allowed("read", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_retrieve_vector_store_file_contents(
+            vector_store_id=vector_store_id,
+            file_id=file_id,
+        )
+
+    async def openai_update_vector_store_file(
+        self,
+        vector_store_id: str,
+        file_id: str,
+        attributes: dict[str, Any],
+    ) -> VectorStoreFileObject:
+        await self.assert_action_allowed("update", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_update_vector_store_file(
+            vector_store_id=vector_store_id,
+            file_id=file_id,
+            attributes=attributes,
+        )
+
+    async def openai_delete_vector_store_file(
+        self,
+        vector_store_id: str,
+        file_id: str,
+    ) -> VectorStoreFileDeleteResponse:
+        await self.assert_action_allowed("delete", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_delete_vector_store_file(
+            vector_store_id=vector_store_id,
+            file_id=file_id,
+        )
+
+    async def openai_create_vector_store_file_batch(
+        self,
+        vector_store_id: str,
+        file_ids: list[str],
+        attributes: dict[str, Any] | None = None,
+        chunking_strategy: Any | None = None,
+    ):
+        await self.assert_action_allowed("update", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_create_vector_store_file_batch(
+            vector_store_id=vector_store_id,
+            file_ids=file_ids,
+            attributes=attributes,
+            chunking_strategy=chunking_strategy,
+        )
+
+    async def openai_retrieve_vector_store_file_batch(
+        self,
+        batch_id: str,
+        vector_store_id: str,
+    ):
+        await self.assert_action_allowed("read", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_retrieve_vector_store_file_batch(
+            batch_id=batch_id,
+            vector_store_id=vector_store_id,
+        )
+
+    async def openai_list_files_in_vector_store_file_batch(
+        self,
+        batch_id: str,
+        vector_store_id: str,
+        after: str | None = None,
+        before: str | None = None,
+        filter: str | None = None,
+        limit: int | None = 20,
+        order: str | None = "desc",
+    ):
+        await self.assert_action_allowed("read", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_list_files_in_vector_store_file_batch(
+            batch_id=batch_id,
+            vector_store_id=vector_store_id,
+            after=after,
+            before=before,
+            filter=filter,
+            limit=limit,
+            order=order,
+        )
+
+    async def openai_cancel_vector_store_file_batch(
+        self,
+        batch_id: str,
+        vector_store_id: str,
+    ):
+        await self.assert_action_allowed("update", "vector_store", vector_store_id)
+        provider = await self.get_provider_impl(vector_store_id)
+        return await provider.openai_cancel_vector_store_file_batch(
+            batch_id=batch_id,
+            vector_store_id=vector_store_id,
+        )

llama_stack/core/server/auth_providers.py

@@ -72,13 +72,30 @@ class AuthProvider(ABC):
 def get_attributes_from_claims(claims: dict[str, str], mapping: dict[str, str]) -> dict[str, list[str]]:
     attributes: dict[str, list[str]] = {}
     for claim_key, attribute_key in mapping.items():
-        if claim_key not in claims:
+        # First try dot notation for nested traversal (e.g., "resource_access.llamastack.roles")
+        # Then fall back to literal key with dots (e.g., "my.dotted.key")
+        claim: object = claims
+        keys = claim_key.split(".")
+        for key in keys:
+            if isinstance(claim, dict) and key in claim:
+                claim = claim[key]
+            else:
+                claim = None
+                break
+
+        if claim is None and claim_key in claims:
+            # Fall back to checking if claim_key exists as a literal key
+            claim = claims[claim_key]
+
+        if claim is None:
             continue
-        claim = claims[claim_key]
+
         if isinstance(claim, list):
             values = claim
-        else:
+        elif isinstance(claim, str):
             values = claim.split()
+        else:
+            continue
 
         if attribute_key in attributes:
             attributes[attribute_key].extend(values)

llama_stack/core/server/quota.py

@@ -10,10 +10,10 @@ from datetime import UTC, datetime, timedelta
 
 from starlette.types import ASGIApp, Receive, Scope, Send
 
+from llama_stack.core.storage.datatypes import KVStoreReference, StorageBackendType
 from llama_stack.log import get_logger
 from llama_stack.providers.utils.kvstore.api import KVStore
-from llama_stack.providers.utils.kvstore.config import KVStoreConfig, SqliteKVStoreConfig
-from llama_stack.providers.utils.kvstore.kvstore import kvstore_impl
+from llama_stack.providers.utils.kvstore.kvstore import _KVSTORE_BACKENDS, kvstore_impl
 
 logger = get_logger(name=__name__, category="core::server")
 
@@ -33,7 +33,7 @@ class QuotaMiddleware:
     def __init__(
         self,
         app: ASGIApp,
-        kv_config: KVStoreConfig,
+        kv_config: KVStoreReference,
         anonymous_max_requests: int,
         authenticated_max_requests: int,
         window_seconds: int = 86400,
@@ -45,15 +45,15 @@ class QuotaMiddleware:
         self.authenticated_max_requests = authenticated_max_requests
         self.window_seconds = window_seconds
 
-        if isinstance(self.kv_config, SqliteKVStoreConfig):
-            logger.warning(
-                "QuotaMiddleware: Using SQLite backend. Expiry/TTL is not enforced; cleanup is manual. "
-                f"window_seconds={self.window_seconds}"
-            )
-
     async def _get_kv(self) -> KVStore:
         if self.kv is None:
             self.kv = await kvstore_impl(self.kv_config)
+            backend_config = _KVSTORE_BACKENDS.get(self.kv_config.backend)
+            if backend_config and backend_config.type == StorageBackendType.KV_SQLITE:
+                logger.warning(
+                    "QuotaMiddleware: Using SQLite backend. Expiry/TTL is not enforced; cleanup is manual. "
+                    f"window_seconds={self.window_seconds}"
+                )
         return self.kv
 
     async def __call__(self, scope: Scope, receive: Receive, send: Send):

llama_stack/core/server/routes.py

@@ -13,7 +13,6 @@ from aiohttp import hdrs
 from starlette.routing import Route
 
 from llama_stack.apis.datatypes import Api, ExternalApiSpec
-from llama_stack.apis.tools import RAGToolRuntime, SpecialToolGroup
 from llama_stack.core.resolver import api_protocol_map
 from llama_stack.schema_utils import WebMethod
 
@@ -25,33 +24,16 @@ RouteImpls = dict[str, PathImpl]
 RouteMatch = tuple[EndpointFunc, PathParams, str, WebMethod]
 
 
-def toolgroup_protocol_map():
-    return {
-        SpecialToolGroup.rag_tool: RAGToolRuntime,
-    }
-
-
 def get_all_api_routes(
     external_apis: dict[Api, ExternalApiSpec] | None = None,
 ) -> dict[Api, list[tuple[Route, WebMethod]]]:
     apis = {}
 
     protocols = api_protocol_map(external_apis)
-    toolgroup_protocols = toolgroup_protocol_map()
     for api, protocol in protocols.items():
         routes = []
         protocol_methods = inspect.getmembers(protocol, predicate=inspect.isfunction)
 
-        # HACK ALERT
-        if api == Api.tool_runtime:
-            for tool_group in SpecialToolGroup:
-                sub_protocol = toolgroup_protocols[tool_group]
-                sub_protocol_methods = inspect.getmembers(sub_protocol, predicate=inspect.isfunction)
-                for name, method in sub_protocol_methods:
-                    if not hasattr(method, "__webmethod__"):
-                        continue
-                    protocol_methods.append((f"{tool_group.value}.{name}", method))
-
         for name, method in protocol_methods:
             # Get all webmethods for this method (supports multiple decorators)
             webmethods = getattr(method, "__webmethods__", [])

llama_stack/core/stack.py

@@ -32,16 +32,26 @@ from llama_stack.apis.scoring_functions import ScoringFunctions
 from llama_stack.apis.shields import Shields
 from llama_stack.apis.synthetic_data_generation import SyntheticDataGeneration
 from llama_stack.apis.telemetry import Telemetry
-from llama_stack.apis.tools import RAGToolRuntime, ToolGroups, ToolRuntime
+from llama_stack.apis.tools import ToolGroups, ToolRuntime
 from llama_stack.apis.vector_io import VectorIO
 from llama_stack.core.conversations.conversations import ConversationServiceConfig, ConversationServiceImpl
-from llama_stack.core.datatypes import Provider, StackRunConfig
+from llama_stack.core.datatypes import Provider, StackRunConfig, VectorStoresConfig
 from llama_stack.core.distribution import get_provider_registry
 from llama_stack.core.inspect import DistributionInspectConfig, DistributionInspectImpl
 from llama_stack.core.prompts.prompts import PromptServiceConfig, PromptServiceImpl
 from llama_stack.core.providers import ProviderImpl, ProviderImplConfig
 from llama_stack.core.resolver import ProviderRegistry, resolve_impls
 from llama_stack.core.routing_tables.common import CommonRoutingTableImpl
+from llama_stack.core.storage.datatypes import (
+    InferenceStoreReference,
+    KVStoreReference,
+    ServerStoresConfig,
+    SqliteKVStoreConfig,
+    SqliteSqlStoreConfig,
+    SqlStoreReference,
+    StorageBackendConfig,
+    StorageConfig,
+)
 from llama_stack.core.store.registry import create_dist_registry
 from llama_stack.core.utils.dynamic import instantiate_class_type
 from llama_stack.log import get_logger
@@ -70,7 +80,6 @@ class LlamaStack(
     Inspect,
     ToolGroups,
     ToolRuntime,
-    RAGToolRuntime,
     Files,
     Prompts,
     Conversations,
@@ -98,33 +107,9 @@ REGISTRY_REFRESH_TASK = None
 TEST_RECORDING_CONTEXT = None
 
 
-async def validate_default_embedding_model(impls: dict[Api, Any]):
-    """Validate that at most one embedding model is marked as default."""
-    if Api.models not in impls:
-        return
-
-    models_impl = impls[Api.models]
-    response = await models_impl.list_models()
-    models_list = response.data if hasattr(response, "data") else response
-
-    default_embedding_models = []
-    for model in models_list:
-        if model.model_type == "embedding" and model.metadata.get("default_configured") is True:
-            default_embedding_models.append(model.identifier)
-
-    if len(default_embedding_models) > 1:
-        raise ValueError(
-            f"Multiple embedding models marked as default_configured=True: {default_embedding_models}. "
-            "Only one embedding model can be marked as default."
-        )
-
-    if default_embedding_models:
-        logger.info(f"Default embedding model configured: {default_embedding_models[0]}")
-
-
 async def register_resources(run_config: StackRunConfig, impls: dict[Api, Any]):
     for rsrc, api, register_method, list_method in RESOURCES:
-        objects = getattr(run_config, rsrc)
+        objects = getattr(run_config.registered_resources, rsrc)
         if api not in impls:
             continue
 
@@ -152,7 +137,41 @@ async def register_resources(run_config: StackRunConfig, impls: dict[Api, Any]):
                 f"{rsrc.capitalize()}: {obj.identifier} served by {obj.provider_id}",
             )
 
-    await validate_default_embedding_model(impls)
+
+async def validate_vector_stores_config(vector_stores_config: VectorStoresConfig | None, impls: dict[Api, Any]):
+    """Validate vector stores configuration."""
+    if vector_stores_config is None:
+        return
+
+    default_embedding_model = vector_stores_config.default_embedding_model
+    if default_embedding_model is None:
+        return
+
+    provider_id = default_embedding_model.provider_id
+    model_id = default_embedding_model.model_id
+    default_model_id = f"{provider_id}/{model_id}"
+
+    if Api.models not in impls:
+        raise ValueError(f"Models API is not available but vector_stores config requires model '{default_model_id}'")
+
+    models_impl = impls[Api.models]
+    response = await models_impl.list_models()
+    models_list = {m.identifier: m for m in response.data if m.model_type == "embedding"}
+
+    default_model = models_list.get(default_model_id)
+    if default_model is None:
+        raise ValueError(f"Embedding model '{default_model_id}' not found. Available embedding models: {models_list}")
+
+    embedding_dimension = default_model.metadata.get("embedding_dimension")
+    if embedding_dimension is None:
+        raise ValueError(f"Embedding model '{default_model_id}' is missing 'embedding_dimension' in metadata")
+
+    try:
+        int(embedding_dimension)
+    except ValueError as err:
+        raise ValueError(f"Embedding dimension '{embedding_dimension}' cannot be converted to an integer") from err
+
+    logger.debug(f"Validated default embedding model: {default_model_id} (dimension: {embedding_dimension})")
 
 
 class EnvVarError(Exception):
@@ -329,6 +348,25 @@ def add_internal_implementations(impls: dict[Api, Any], run_config: StackRunConf
     impls[Api.conversations] = conversations_impl
 
 
+def _initialize_storage(run_config: StackRunConfig):
+    kv_backends: dict[str, StorageBackendConfig] = {}
+    sql_backends: dict[str, StorageBackendConfig] = {}
+    for backend_name, backend_config in run_config.storage.backends.items():
+        type = backend_config.type.value
+        if type.startswith("kv_"):
+            kv_backends[backend_name] = backend_config
+        elif type.startswith("sql_"):
+            sql_backends[backend_name] = backend_config
+        else:
+            raise ValueError(f"Unknown storage backend type: {type}")
+
+    from llama_stack.providers.utils.kvstore.kvstore import register_kvstore_backends
+    from llama_stack.providers.utils.sqlstore.sqlstore import register_sqlstore_backends
+
+    register_kvstore_backends(kv_backends)
+    register_sqlstore_backends(sql_backends)
+
+
 class Stack:
     def __init__(self, run_config: StackRunConfig, provider_registry: ProviderRegistry | None = None):
         self.run_config = run_config
@@ -347,7 +385,11 @@ class Stack:
                 TEST_RECORDING_CONTEXT.__enter__()
                 logger.info(f"API recording enabled: mode={os.environ.get('LLAMA_STACK_TEST_INFERENCE_MODE')}")
 
-        dist_registry, _ = await create_dist_registry(self.run_config.metadata_store, self.run_config.image_name)
+        _initialize_storage(self.run_config)
+        stores = self.run_config.storage.stores
+        if not stores.metadata:
+            raise ValueError("storage.stores.metadata must be configured with a kv_* backend")
+        dist_registry, _ = await create_dist_registry(stores.metadata, self.run_config.image_name)
         policy = self.run_config.server.auth.access_policy if self.run_config.server.auth else []
 
         internal_impls = {}
@@ -367,8 +409,8 @@ class Stack:
             await impls[Api.conversations].initialize()
 
         await register_resources(self.run_config, impls)
-
         await refresh_registry_once(impls)
+        await validate_vector_stores_config(self.run_config.vector_stores, impls)
         self.impls = impls
 
     def create_registry_refresh_task(self):
@@ -488,5 +530,16 @@ def run_config_from_adhoc_config_spec(
         image_name="distro-test",
         apis=list(provider_configs_by_api.keys()),
         providers=provider_configs_by_api,
+        storage=StorageConfig(
+            backends={
+                "kv_default": SqliteKVStoreConfig(db_path=f"{distro_dir}/kvstore.db"),
+                "sql_default": SqliteSqlStoreConfig(db_path=f"{distro_dir}/sql_store.db"),
+            },
+            stores=ServerStoresConfig(
+                metadata=KVStoreReference(backend="kv_default", namespace="registry"),
+                inference=InferenceStoreReference(backend="sql_default", table_name="inference_store"),
+                conversations=SqlStoreReference(backend="sql_default", table_name="openai_conversations"),
+            ),
+        ),
     )
     return config

llama_stack/core/storage/datatypes.py

@@ -0,0 +1,283 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+import re
+from abc import abstractmethod
+from enum import StrEnum
+from pathlib import Path
+from typing import Annotated, Literal
+
+from pydantic import BaseModel, Field, field_validator
+
+
+class StorageBackendType(StrEnum):
+    KV_REDIS = "kv_redis"
+    KV_SQLITE = "kv_sqlite"
+    KV_POSTGRES = "kv_postgres"
+    KV_MONGODB = "kv_mongodb"
+    SQL_SQLITE = "sql_sqlite"
+    SQL_POSTGRES = "sql_postgres"
+
+
+class CommonConfig(BaseModel):
+    namespace: str | None = Field(
+        default=None,
+        description="All keys will be prefixed with this namespace",
+    )
+
+
+class RedisKVStoreConfig(CommonConfig):
+    type: Literal[StorageBackendType.KV_REDIS] = StorageBackendType.KV_REDIS
+    host: str = "localhost"
+    port: int = 6379
+
+    @property
+    def url(self) -> str:
+        return f"redis://{self.host}:{self.port}"
+
+    @classmethod
+    def pip_packages(cls) -> list[str]:
+        return ["redis"]
+
+    @classmethod
+    def sample_run_config(cls):
+        return {
+            "type": StorageBackendType.KV_REDIS.value,
+            "host": "${env.REDIS_HOST:=localhost}",
+            "port": "${env.REDIS_PORT:=6379}",
+        }
+
+
+class SqliteKVStoreConfig(CommonConfig):
+    type: Literal[StorageBackendType.KV_SQLITE] = StorageBackendType.KV_SQLITE
+    db_path: str = Field(
+        description="File path for the sqlite database",
+    )
+
+    @classmethod
+    def pip_packages(cls) -> list[str]:
+        return ["aiosqlite"]
+
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, db_name: str = "kvstore.db"):
+        return {
+            "type": StorageBackendType.KV_SQLITE.value,
+            "db_path": "${env.SQLITE_STORE_DIR:=" + __distro_dir__ + "}/" + db_name,
+        }
+
+
+class PostgresKVStoreConfig(CommonConfig):
+    type: Literal[StorageBackendType.KV_POSTGRES] = StorageBackendType.KV_POSTGRES
+    host: str = "localhost"
+    port: int | str = 5432
+    db: str = "llamastack"
+    user: str
+    password: str | None = None
+    ssl_mode: str | None = None
+    ca_cert_path: str | None = None
+    table_name: str = "llamastack_kvstore"
+
+    @classmethod
+    def sample_run_config(cls, table_name: str = "llamastack_kvstore", **kwargs):
+        return {
+            "type": StorageBackendType.KV_POSTGRES.value,
+            "host": "${env.POSTGRES_HOST:=localhost}",
+            "port": "${env.POSTGRES_PORT:=5432}",
+            "db": "${env.POSTGRES_DB:=llamastack}",
+            "user": "${env.POSTGRES_USER:=llamastack}",
+            "password": "${env.POSTGRES_PASSWORD:=llamastack}",
+            "table_name": "${env.POSTGRES_TABLE_NAME:=" + table_name + "}",
+        }
+
+    @classmethod
+    @field_validator("table_name")
+    def validate_table_name(cls, v: str) -> str:
+        # PostgreSQL identifiers rules:
+        # - Must start with a letter or underscore
+        # - Can contain letters, numbers, and underscores
+        # - Maximum length is 63 bytes
+        pattern = r"^[a-zA-Z_][a-zA-Z0-9_]*$"
+        if not re.match(pattern, v):
+            raise ValueError(
+                "Invalid table name. Must start with letter or underscore and contain only letters, numbers, and underscores"
+            )
+        if len(v) > 63:
+            raise ValueError("Table name must be less than 63 characters")
+        return v
+
+    @classmethod
+    def pip_packages(cls) -> list[str]:
+        return ["psycopg2-binary"]
+
+
+class MongoDBKVStoreConfig(CommonConfig):
+    type: Literal[StorageBackendType.KV_MONGODB] = StorageBackendType.KV_MONGODB
+    host: str = "localhost"
+    port: int = 27017
+    db: str = "llamastack"
+    user: str | None = None
+    password: str | None = None
+    collection_name: str = "llamastack_kvstore"
+
+    @classmethod
+    def pip_packages(cls) -> list[str]:
+        return ["pymongo"]
+
+    @classmethod
+    def sample_run_config(cls, collection_name: str = "llamastack_kvstore"):
+        return {
+            "type": StorageBackendType.KV_MONGODB.value,
+            "host": "${env.MONGODB_HOST:=localhost}",
+            "port": "${env.MONGODB_PORT:=5432}",
+            "db": "${env.MONGODB_DB}",
+            "user": "${env.MONGODB_USER}",
+            "password": "${env.MONGODB_PASSWORD}",
+            "collection_name": "${env.MONGODB_COLLECTION_NAME:=" + collection_name + "}",
+        }
+
+
+class SqlAlchemySqlStoreConfig(BaseModel):
+    @property
+    @abstractmethod
+    def engine_str(self) -> str: ...
+
+    # TODO: move this when we have a better way to specify dependencies with internal APIs
+    @classmethod
+    def pip_packages(cls) -> list[str]:
+        return ["sqlalchemy[asyncio]"]
+
+
+class SqliteSqlStoreConfig(SqlAlchemySqlStoreConfig):
+    type: Literal[StorageBackendType.SQL_SQLITE] = StorageBackendType.SQL_SQLITE
+    db_path: str = Field(
+        description="Database path, e.g. ~/.llama/distributions/ollama/sqlstore.db",
+    )
+
+    @property
+    def engine_str(self) -> str:
+        return "sqlite+aiosqlite:///" + Path(self.db_path).expanduser().as_posix()
+
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str, db_name: str = "sqlstore.db"):
+        return {
+            "type": StorageBackendType.SQL_SQLITE.value,
+            "db_path": "${env.SQLITE_STORE_DIR:=" + __distro_dir__ + "}/" + db_name,
+        }
+
+    @classmethod
+    def pip_packages(cls) -> list[str]:
+        return super().pip_packages() + ["aiosqlite"]
+
+
+class PostgresSqlStoreConfig(SqlAlchemySqlStoreConfig):
+    type: Literal[StorageBackendType.SQL_POSTGRES] = StorageBackendType.SQL_POSTGRES
+    host: str = "localhost"
+    port: int | str = 5432
+    db: str = "llamastack"
+    user: str
+    password: str | None = None
+
+    @property
+    def engine_str(self) -> str:
+        return f"postgresql+asyncpg://{self.user}:{self.password}@{self.host}:{self.port}/{self.db}"
+
+    @classmethod
+    def pip_packages(cls) -> list[str]:
+        return super().pip_packages() + ["asyncpg"]
+
+    @classmethod
+    def sample_run_config(cls, **kwargs):
+        return {
+            "type": StorageBackendType.SQL_POSTGRES.value,
+            "host": "${env.POSTGRES_HOST:=localhost}",
+            "port": "${env.POSTGRES_PORT:=5432}",
+            "db": "${env.POSTGRES_DB:=llamastack}",
+            "user": "${env.POSTGRES_USER:=llamastack}",
+            "password": "${env.POSTGRES_PASSWORD:=llamastack}",
+        }
+
+
+# reference = (backend_name, table_name)
+class SqlStoreReference(BaseModel):
+    """A reference to a 'SQL-like' persistent store. A table name must be provided."""
+
+    table_name: str = Field(
+        description="Name of the table to use for the SqlStore",
+    )
+
+    backend: str = Field(
+        description="Name of backend from storage.backends",
+    )
+
+
+# reference = (backend_name, namespace)
+class KVStoreReference(BaseModel):
+    """A reference to a 'key-value' persistent store. A namespace must be provided."""
+
+    namespace: str = Field(
+        description="Key prefix for KVStore backends",
+    )
+
+    backend: str = Field(
+        description="Name of backend from storage.backends",
+    )
+
+
+StorageBackendConfig = Annotated[
+    RedisKVStoreConfig
+    | SqliteKVStoreConfig
+    | PostgresKVStoreConfig
+    | MongoDBKVStoreConfig
+    | SqliteSqlStoreConfig
+    | PostgresSqlStoreConfig,
+    Field(discriminator="type"),
+]
+
+
+class InferenceStoreReference(SqlStoreReference):
+    """Inference store configuration with queue tuning."""
+
+    max_write_queue_size: int = Field(
+        default=10000,
+        description="Max queued writes for inference store",
+    )
+    num_writers: int = Field(
+        default=4,
+        description="Number of concurrent background writers",
+    )
+
+
+class ResponsesStoreReference(InferenceStoreReference):
+    """Responses store configuration with queue tuning."""
+
+
+class ServerStoresConfig(BaseModel):
+    metadata: KVStoreReference | None = Field(
+        default=None,
+        description="Metadata store configuration (uses KV backend)",
+    )
+    inference: InferenceStoreReference | None = Field(
+        default=None,
+        description="Inference store configuration (uses SQL backend)",
+    )
+    conversations: SqlStoreReference | None = Field(
+        default=None,
+        description="Conversations store configuration (uses SQL backend)",
+    )
+    responses: ResponsesStoreReference | None = Field(
+        default=None,
+        description="Responses store configuration (uses SQL backend)",
+    )
+
+
+class StorageConfig(BaseModel):
+    backends: dict[str, StorageBackendConfig] = Field(
+        description="Named backend configurations (e.g., 'default', 'cache')",
+    )
+    stores: ServerStoresConfig = Field(
+        default_factory=lambda: ServerStoresConfig(),
+        description="Named references to storage backends used by the stack core",
+    )

llama_stack/core/store/registry.py

@@ -11,10 +11,9 @@ from typing import Protocol
 import pydantic
 
 from llama_stack.core.datatypes import RoutableObjectWithProvider
-from llama_stack.core.utils.config_dirs import DISTRIBS_BASE_DIR
+from llama_stack.core.storage.datatypes import KVStoreReference
 from llama_stack.log import get_logger
 from llama_stack.providers.utils.kvstore import KVStore, kvstore_impl
-from llama_stack.providers.utils.kvstore.config import KVStoreConfig, SqliteKVStoreConfig
 
 logger = get_logger(__name__, category="core::registry")
 
@@ -191,16 +190,10 @@ class CachedDiskDistributionRegistry(DiskDistributionRegistry):
 
 
 async def create_dist_registry(
-    metadata_store: KVStoreConfig | None,
-    image_name: str,
+    metadata_store: KVStoreReference, image_name: str
 ) -> tuple[CachedDiskDistributionRegistry, KVStore]:
     # instantiate kvstore for storing and retrieving distribution metadata
-    if metadata_store:
-        dist_kvstore = await kvstore_impl(metadata_store)
-    else:
-        dist_kvstore = await kvstore_impl(
-            SqliteKVStoreConfig(db_path=(DISTRIBS_BASE_DIR / image_name / "kvstore.db").as_posix())
-        )
+    dist_kvstore = await kvstore_impl(metadata_store)
     dist_registry = CachedDiskDistributionRegistry(dist_kvstore)
     await dist_registry.initialize()
     return dist_registry, dist_kvstore

llama_stack/core/ui/README.md

@@ -9,7 +9,7 @@
 1. Start up Llama Stack API server. More details [here](https://llamastack.github.io/latest/getting_started/index.htmll).
 
 ```
-llama stack build --distro together --image-type venv
+llama stack list-deps together | xargs -L1 uv pip install
 
 llama stack run together
 ```

llama_stack/core/ui/page/playground/tools.py

@@ -32,7 +32,7 @@ def tool_chat_page():
     tool_groups_list = [tool_group.identifier for tool_group in tool_groups]
     mcp_tools_list = [tool for tool in tool_groups_list if tool.startswith("mcp::")]
     builtin_tools_list = [tool for tool in tool_groups_list if not tool.startswith("mcp::")]
-    selected_vector_dbs = []
+    selected_vector_stores = []
 
     def reset_agent():
         st.session_state.clear()
@@ -55,13 +55,13 @@ def tool_chat_page():
         )
 
         if "builtin::rag" in toolgroup_selection:
-            vector_dbs = llama_stack_api.client.vector_dbs.list() or []
-            if not vector_dbs:
+            vector_stores = llama_stack_api.client.vector_stores.list() or []
+            if not vector_stores:
                 st.info("No vector databases available for selection.")
-            vector_dbs = [vector_db.identifier for vector_db in vector_dbs]
-            selected_vector_dbs = st.multiselect(
+            vector_stores = [vector_store.identifier for vector_store in vector_stores]
+            selected_vector_stores = st.multiselect(
                 label="Select Document Collections to use in RAG queries",
-                options=vector_dbs,
+                options=vector_stores,
                 on_change=reset_agent,
             )
 
@@ -119,7 +119,7 @@ def tool_chat_page():
             tool_dict = dict(
                 name="builtin::rag",
                 args={
-                    "vector_db_ids": list(selected_vector_dbs),
+                    "vector_store_ids": list(selected_vector_stores),
                 },
             )
             toolgroup_selection[i] = tool_dict

llama_stack/distributions/ci-tests/build.yaml

@@ -25,6 +25,8 @@ distribution_spec:
     - provider_type: inline::milvus
     - provider_type: remote::chromadb
     - provider_type: remote::pgvector
+    - provider_type: remote::qdrant
+    - provider_type: remote::weaviate
     files:
     - provider_type: inline::localfs
     safety:
@@ -46,7 +48,6 @@ distribution_spec:
     tool_runtime:
     - provider_type: remote::brave-search
     - provider_type: remote::tavily-search
-    - provider_type: inline::rag-runtime
     - provider_type: remote::model-context-protocol
     batches:
     - provider_type: inline::reference

llama_stack/distributions/ci-tests/run.yaml

@@ -93,30 +93,30 @@ providers:
   - provider_id: faiss
     provider_type: inline::faiss
     config:
-      kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/faiss_store.db
+      persistence:
+        namespace: vector_io::faiss
+        backend: kv_default
   - provider_id: sqlite-vec
     provider_type: inline::sqlite-vec
     config:
       db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/sqlite_vec.db
-      kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/sqlite_vec_registry.db
+      persistence:
+        namespace: vector_io::sqlite_vec
+        backend: kv_default
   - provider_id: ${env.MILVUS_URL:+milvus}
     provider_type: inline::milvus
     config:
       db_path: ${env.MILVUS_DB_PATH:=~/.llama/distributions/ci-tests}/milvus.db
-      kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/milvus_registry.db
+      persistence:
+        namespace: vector_io::milvus
+        backend: kv_default
   - provider_id: ${env.CHROMADB_URL:+chromadb}
     provider_type: remote::chromadb
     config:
       url: ${env.CHROMADB_URL:=}
-      kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests/}/chroma_remote_registry.db
+      persistence:
+        namespace: vector_io::chroma_remote
+        backend: kv_default
   - provider_id: ${env.PGVECTOR_DB:+pgvector}
     provider_type: remote::pgvector
     config:
@@ -125,17 +125,32 @@ providers:
       db: ${env.PGVECTOR_DB:=}
       user: ${env.PGVECTOR_USER:=}
       password: ${env.PGVECTOR_PASSWORD:=}
-      kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/pgvector_registry.db
+      persistence:
+        namespace: vector_io::pgvector
+        backend: kv_default
+  - provider_id: ${env.QDRANT_URL:+qdrant}
+    provider_type: remote::qdrant
+    config:
+      api_key: ${env.QDRANT_API_KEY:=}
+      persistence:
+        namespace: vector_io::qdrant_remote
+        backend: kv_default
+  - provider_id: ${env.WEAVIATE_CLUSTER_URL:+weaviate}
+    provider_type: remote::weaviate
+    config:
+      weaviate_api_key: null
+      weaviate_cluster_url: ${env.WEAVIATE_CLUSTER_URL:=localhost:8080}
+      persistence:
+        namespace: vector_io::weaviate
+        backend: kv_default
   files:
   - provider_id: meta-reference-files
     provider_type: inline::localfs
     config:
       storage_dir: ${env.FILES_STORAGE_DIR:=~/.llama/distributions/ci-tests/files}
       metadata_store:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/files_metadata.db
+        table_name: files_metadata
+        backend: sql_default
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
@@ -147,12 +162,15 @@ providers:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
     config:
-      persistence_store:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/agents_store.db
-      responses_store:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/responses_store.db
+      persistence:
+        agent_state:
+          namespace: agents
+          backend: kv_default
+        responses:
+          table_name: responses
+          backend: sql_default
+          max_write_queue_size: 10000
+          num_writers: 4
   post_training:
   - provider_id: torchtune-cpu
     provider_type: inline::torchtune-cpu
@@ -163,21 +181,21 @@ providers:
     provider_type: inline::meta-reference
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/meta_reference_eval.db
+        namespace: eval
+        backend: kv_default
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/huggingface_datasetio.db
+        namespace: datasetio::huggingface
+        backend: kv_default
   - provider_id: localfs
     provider_type: inline::localfs
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/localfs_datasetio.db
+        namespace: datasetio::localfs
+        backend: kv_default
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -198,8 +216,6 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
-  - provider_id: rag-runtime
-    provider_type: inline::rag-runtime
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
   batches:
@@ -207,35 +223,50 @@ providers:
     provider_type: inline::reference
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/batches.db
-metadata_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/registry.db
-inference_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/inference_store.db
-conversations_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/conversations.db
-models: []
-shields:
-- shield_id: llama-guard
-  provider_id: ${env.SAFETY_MODEL:+llama-guard}
-  provider_shield_id: ${env.SAFETY_MODEL:=}
-- shield_id: code-scanner
-  provider_id: ${env.CODE_SCANNER_MODEL:+code-scanner}
-  provider_shield_id: ${env.CODE_SCANNER_MODEL:=}
-vector_dbs: []
-datasets: []
-scoring_fns: []
-benchmarks: []
-tool_groups:
-- toolgroup_id: builtin::websearch
-  provider_id: tavily-search
-- toolgroup_id: builtin::rag
-  provider_id: rag-runtime
+        namespace: batches
+        backend: kv_default
+storage:
+  backends:
+    kv_default:
+      type: kv_sqlite
+      db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/kvstore.db
+    sql_default:
+      type: sql_sqlite
+      db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ci-tests}/sql_store.db
+  stores:
+    metadata:
+      namespace: registry
+      backend: kv_default
+    inference:
+      table_name: inference_store
+      backend: sql_default
+      max_write_queue_size: 10000
+      num_writers: 4
+    conversations:
+      table_name: openai_conversations
+      backend: sql_default
+registered_resources:
+  models: []
+  shields:
+  - shield_id: llama-guard
+    provider_id: ${env.SAFETY_MODEL:+llama-guard}
+    provider_shield_id: ${env.SAFETY_MODEL:=}
+  - shield_id: code-scanner
+    provider_id: ${env.CODE_SCANNER_MODEL:+code-scanner}
+    provider_shield_id: ${env.CODE_SCANNER_MODEL:=}
+  vector_dbs: []
+  datasets: []
+  scoring_fns: []
+  benchmarks: []
+  tool_groups:
+  - toolgroup_id: builtin::websearch
+    provider_id: tavily-search
 server:
   port: 8321
 telemetry:
   enabled: true
+vector_stores:
+  default_provider_id: faiss
+  default_embedding_model:
+    provider_id: sentence-transformers
+    model_id: nomic-ai/nomic-embed-text-v1.5

llama_stack/distributions/dell/build.yaml

@@ -26,7 +26,6 @@ distribution_spec:
     tool_runtime:
     - provider_type: remote::brave-search
     - provider_type: remote::tavily-search
-    - provider_type: inline::rag-runtime
 image_type: venv
 additional_pip_packages:
 - aiosqlite

llama_stack/distributions/dell/dell.py

@@ -45,7 +45,6 @@ def get_distribution_template() -> DistributionTemplate:
         "tool_runtime": [
             BuildProvider(provider_type="remote::brave-search"),
             BuildProvider(provider_type="remote::tavily-search"),
-            BuildProvider(provider_type="inline::rag-runtime"),
         ],
     }
     name = "dell"
@@ -98,10 +97,6 @@ def get_distribution_template() -> DistributionTemplate:
             toolgroup_id="builtin::websearch",
             provider_id="brave-search",
         ),
-        ToolGroupInput(
-            toolgroup_id="builtin::rag",
-            provider_id="rag-runtime",
-        ),
     ]
 
     return DistributionTemplate(

llama_stack/distributions/dell/doc_template.md

@@ -157,7 +157,7 @@ docker run \
 Make sure you have done `pip install llama-stack` and have the Llama Stack CLI available.
 
 ```bash
-llama stack build --distro {{ name }} --image-type conda
+llama stack list-deps {{ name }} | xargs -L1 pip install
 INFERENCE_MODEL=$INFERENCE_MODEL \
 DEH_URL=$DEH_URL \
 CHROMA_URL=$CHROMA_URL \

llama_stack/distributions/dell/run-with-safety.yaml

@@ -26,9 +26,9 @@ providers:
     provider_type: remote::chromadb
     config:
       url: ${env.CHROMADB_URL:=}
-      kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell/}/chroma_remote_registry.db
+      persistence:
+        namespace: vector_io::chroma_remote
+        backend: kv_default
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
@@ -38,32 +38,35 @@ providers:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
     config:
-      persistence_store:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/agents_store.db
-      responses_store:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/responses_store.db
+      persistence:
+        agent_state:
+          namespace: agents
+          backend: kv_default
+        responses:
+          table_name: responses
+          backend: sql_default
+          max_write_queue_size: 10000
+          num_writers: 4
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/meta_reference_eval.db
+        namespace: eval
+        backend: kv_default
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/huggingface_datasetio.db
+        namespace: datasetio::huggingface
+        backend: kv_default
   - provider_id: localfs
     provider_type: inline::localfs
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/localfs_datasetio.db
+        namespace: datasetio::localfs
+        backend: kv_default
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -84,42 +87,50 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
-  - provider_id: rag-runtime
-    provider_type: inline::rag-runtime
-metadata_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/registry.db
-inference_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/inference_store.db
-conversations_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/conversations.db
-models:
-- metadata: {}
-  model_id: ${env.INFERENCE_MODEL}
-  provider_id: tgi0
-  model_type: llm
-- metadata: {}
-  model_id: ${env.SAFETY_MODEL}
-  provider_id: tgi1
-  model_type: llm
-- metadata:
-    embedding_dimension: 768
-  model_id: nomic-embed-text-v1.5
-  provider_id: sentence-transformers
-  model_type: embedding
-shields:
-- shield_id: ${env.SAFETY_MODEL}
-vector_dbs: []
-datasets: []
-scoring_fns: []
-benchmarks: []
-tool_groups:
-- toolgroup_id: builtin::websearch
-  provider_id: brave-search
-- toolgroup_id: builtin::rag
-  provider_id: rag-runtime
+storage:
+  backends:
+    kv_default:
+      type: kv_sqlite
+      db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/kvstore.db
+    sql_default:
+      type: sql_sqlite
+      db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/sql_store.db
+  stores:
+    metadata:
+      namespace: registry
+      backend: kv_default
+    inference:
+      table_name: inference_store
+      backend: sql_default
+      max_write_queue_size: 10000
+      num_writers: 4
+    conversations:
+      table_name: openai_conversations
+      backend: sql_default
+registered_resources:
+  models:
+  - metadata: {}
+    model_id: ${env.INFERENCE_MODEL}
+    provider_id: tgi0
+    model_type: llm
+  - metadata: {}
+    model_id: ${env.SAFETY_MODEL}
+    provider_id: tgi1
+    model_type: llm
+  - metadata:
+      embedding_dimension: 768
+    model_id: nomic-embed-text-v1.5
+    provider_id: sentence-transformers
+    model_type: embedding
+  shields:
+  - shield_id: ${env.SAFETY_MODEL}
+  vector_dbs: []
+  datasets: []
+  scoring_fns: []
+  benchmarks: []
+  tool_groups:
+  - toolgroup_id: builtin::websearch
+    provider_id: brave-search
 server:
   port: 8321
 telemetry:

llama_stack/distributions/dell/run.yaml

@@ -22,9 +22,9 @@ providers:
     provider_type: remote::chromadb
     config:
       url: ${env.CHROMADB_URL:=}
-      kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell/}/chroma_remote_registry.db
+      persistence:
+        namespace: vector_io::chroma_remote
+        backend: kv_default
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
@@ -34,32 +34,35 @@ providers:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
     config:
-      persistence_store:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/agents_store.db
-      responses_store:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/responses_store.db
+      persistence:
+        agent_state:
+          namespace: agents
+          backend: kv_default
+        responses:
+          table_name: responses
+          backend: sql_default
+          max_write_queue_size: 10000
+          num_writers: 4
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/meta_reference_eval.db
+        namespace: eval
+        backend: kv_default
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/huggingface_datasetio.db
+        namespace: datasetio::huggingface
+        backend: kv_default
   - provider_id: localfs
     provider_type: inline::localfs
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/localfs_datasetio.db
+        namespace: datasetio::localfs
+        backend: kv_default
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -80,37 +83,45 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
-  - provider_id: rag-runtime
-    provider_type: inline::rag-runtime
-metadata_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/registry.db
-inference_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/inference_store.db
-conversations_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/conversations.db
-models:
-- metadata: {}
-  model_id: ${env.INFERENCE_MODEL}
-  provider_id: tgi0
-  model_type: llm
-- metadata:
-    embedding_dimension: 768
-  model_id: nomic-embed-text-v1.5
-  provider_id: sentence-transformers
-  model_type: embedding
-shields: []
-vector_dbs: []
-datasets: []
-scoring_fns: []
-benchmarks: []
-tool_groups:
-- toolgroup_id: builtin::websearch
-  provider_id: brave-search
-- toolgroup_id: builtin::rag
-  provider_id: rag-runtime
+storage:
+  backends:
+    kv_default:
+      type: kv_sqlite
+      db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/kvstore.db
+    sql_default:
+      type: sql_sqlite
+      db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/dell}/sql_store.db
+  stores:
+    metadata:
+      namespace: registry
+      backend: kv_default
+    inference:
+      table_name: inference_store
+      backend: sql_default
+      max_write_queue_size: 10000
+      num_writers: 4
+    conversations:
+      table_name: openai_conversations
+      backend: sql_default
+registered_resources:
+  models:
+  - metadata: {}
+    model_id: ${env.INFERENCE_MODEL}
+    provider_id: tgi0
+    model_type: llm
+  - metadata:
+      embedding_dimension: 768
+    model_id: nomic-embed-text-v1.5
+    provider_id: sentence-transformers
+    model_type: embedding
+  shields: []
+  vector_dbs: []
+  datasets: []
+  scoring_fns: []
+  benchmarks: []
+  tool_groups:
+  - toolgroup_id: builtin::websearch
+    provider_id: brave-search
 server:
   port: 8321
 telemetry:

llama_stack/distributions/meta-reference-gpu/build.yaml

@@ -24,7 +24,6 @@ distribution_spec:
     tool_runtime:
     - provider_type: remote::brave-search
     - provider_type: remote::tavily-search
-    - provider_type: inline::rag-runtime
     - provider_type: remote::model-context-protocol
 image_type: venv
 additional_pip_packages:

llama_stack/distributions/meta-reference-gpu/meta_reference.py

@@ -47,7 +47,6 @@ def get_distribution_template() -> DistributionTemplate:
         "tool_runtime": [
             BuildProvider(provider_type="remote::brave-search"),
             BuildProvider(provider_type="remote::tavily-search"),
-            BuildProvider(provider_type="inline::rag-runtime"),
             BuildProvider(provider_type="remote::model-context-protocol"),
         ],
     }
@@ -92,10 +91,6 @@ def get_distribution_template() -> DistributionTemplate:
             toolgroup_id="builtin::websearch",
             provider_id="tavily-search",
         ),
-        ToolGroupInput(
-            toolgroup_id="builtin::rag",
-            provider_id="rag-runtime",
-        ),
     ]
 
     return DistributionTemplate(

llama_stack/distributions/meta-reference-gpu/run-with-safety.yaml

@@ -37,9 +37,9 @@ providers:
   - provider_id: faiss
     provider_type: inline::faiss
     config:
-      kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/faiss_store.db
+      persistence:
+        namespace: vector_io::faiss
+        backend: kv_default
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
@@ -49,32 +49,35 @@ providers:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
     config:
-      persistence_store:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/agents_store.db
-      responses_store:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/responses_store.db
+      persistence:
+        agent_state:
+          namespace: agents
+          backend: kv_default
+        responses:
+          table_name: responses
+          backend: sql_default
+          max_write_queue_size: 10000
+          num_writers: 4
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/meta_reference_eval.db
+        namespace: eval
+        backend: kv_default
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/huggingface_datasetio.db
+        namespace: datasetio::huggingface
+        backend: kv_default
   - provider_id: localfs
     provider_type: inline::localfs
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/localfs_datasetio.db
+        namespace: datasetio::localfs
+        backend: kv_default
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -95,44 +98,52 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
-  - provider_id: rag-runtime
-    provider_type: inline::rag-runtime
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
-metadata_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/registry.db
-inference_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/inference_store.db
-conversations_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/conversations.db
-models:
-- metadata: {}
-  model_id: ${env.INFERENCE_MODEL}
-  provider_id: meta-reference-inference
-  model_type: llm
-- metadata: {}
-  model_id: ${env.SAFETY_MODEL}
-  provider_id: meta-reference-safety
-  model_type: llm
-- metadata:
-    embedding_dimension: 768
-  model_id: nomic-embed-text-v1.5
-  provider_id: sentence-transformers
-  model_type: embedding
-shields:
-- shield_id: ${env.SAFETY_MODEL}
-vector_dbs: []
-datasets: []
-scoring_fns: []
-benchmarks: []
-tool_groups:
-- toolgroup_id: builtin::websearch
-  provider_id: tavily-search
-- toolgroup_id: builtin::rag
-  provider_id: rag-runtime
+storage:
+  backends:
+    kv_default:
+      type: kv_sqlite
+      db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/kvstore.db
+    sql_default:
+      type: sql_sqlite
+      db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/sql_store.db
+  stores:
+    metadata:
+      namespace: registry
+      backend: kv_default
+    inference:
+      table_name: inference_store
+      backend: sql_default
+      max_write_queue_size: 10000
+      num_writers: 4
+    conversations:
+      table_name: openai_conversations
+      backend: sql_default
+registered_resources:
+  models:
+  - metadata: {}
+    model_id: ${env.INFERENCE_MODEL}
+    provider_id: meta-reference-inference
+    model_type: llm
+  - metadata: {}
+    model_id: ${env.SAFETY_MODEL}
+    provider_id: meta-reference-safety
+    model_type: llm
+  - metadata:
+      embedding_dimension: 768
+    model_id: nomic-embed-text-v1.5
+    provider_id: sentence-transformers
+    model_type: embedding
+  shields:
+  - shield_id: ${env.SAFETY_MODEL}
+  vector_dbs: []
+  datasets: []
+  scoring_fns: []
+  benchmarks: []
+  tool_groups:
+  - toolgroup_id: builtin::websearch
+    provider_id: tavily-search
 server:
   port: 8321
 telemetry:

llama_stack/distributions/meta-reference-gpu/run.yaml

@@ -27,9 +27,9 @@ providers:
   - provider_id: faiss
     provider_type: inline::faiss
     config:
-      kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/faiss_store.db
+      persistence:
+        namespace: vector_io::faiss
+        backend: kv_default
   safety:
   - provider_id: llama-guard
     provider_type: inline::llama-guard
@@ -39,32 +39,35 @@ providers:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
     config:
-      persistence_store:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/agents_store.db
-      responses_store:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/responses_store.db
+      persistence:
+        agent_state:
+          namespace: agents
+          backend: kv_default
+        responses:
+          table_name: responses
+          backend: sql_default
+          max_write_queue_size: 10000
+          num_writers: 4
   eval:
   - provider_id: meta-reference
     provider_type: inline::meta-reference
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/meta_reference_eval.db
+        namespace: eval
+        backend: kv_default
   datasetio:
   - provider_id: huggingface
     provider_type: remote::huggingface
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/huggingface_datasetio.db
+        namespace: datasetio::huggingface
+        backend: kv_default
   - provider_id: localfs
     provider_type: inline::localfs
     config:
       kvstore:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/localfs_datasetio.db
+        namespace: datasetio::localfs
+        backend: kv_default
   scoring:
   - provider_id: basic
     provider_type: inline::basic
@@ -85,39 +88,47 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
-  - provider_id: rag-runtime
-    provider_type: inline::rag-runtime
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
-metadata_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/registry.db
-inference_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/inference_store.db
-conversations_store:
-  type: sqlite
-  db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/conversations.db
-models:
-- metadata: {}
-  model_id: ${env.INFERENCE_MODEL}
-  provider_id: meta-reference-inference
-  model_type: llm
-- metadata:
-    embedding_dimension: 768
-  model_id: nomic-embed-text-v1.5
-  provider_id: sentence-transformers
-  model_type: embedding
-shields: []
-vector_dbs: []
-datasets: []
-scoring_fns: []
-benchmarks: []
-tool_groups:
-- toolgroup_id: builtin::websearch
-  provider_id: tavily-search
-- toolgroup_id: builtin::rag
-  provider_id: rag-runtime
+storage:
+  backends:
+    kv_default:
+      type: kv_sqlite
+      db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/kvstore.db
+    sql_default:
+      type: sql_sqlite
+      db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/meta-reference-gpu}/sql_store.db
+  stores:
+    metadata:
+      namespace: registry
+      backend: kv_default
+    inference:
+      table_name: inference_store
+      backend: sql_default
+      max_write_queue_size: 10000
+      num_writers: 4
+    conversations:
+      table_name: openai_conversations
+      backend: sql_default
+registered_resources:
+  models:
+  - metadata: {}
+    model_id: ${env.INFERENCE_MODEL}
+    provider_id: meta-reference-inference
+    model_type: llm
+  - metadata:
+      embedding_dimension: 768
+    model_id: nomic-embed-text-v1.5
+    provider_id: sentence-transformers
+    model_type: embedding
+  shields: []
+  vector_dbs: []
+  datasets: []
+  scoring_fns: []
+  benchmarks: []
+  tool_groups:
+  - toolgroup_id: builtin::websearch
+    provider_id: tavily-search
 server:
   port: 8321
 telemetry:

llama_stack/distributions/nvidia/build.yaml

@@ -19,8 +19,7 @@ distribution_spec:
     - provider_type: remote::nvidia
     scoring:
     - provider_type: inline::basic
-    tool_runtime:
-    - provider_type: inline::rag-runtime
+    tool_runtime: []
     files:
     - provider_type: inline::localfs
 image_type: venv

llama_stack/distributions/nvidia/nvidia.py

@@ -28,7 +28,7 @@ def get_distribution_template(name: str = "nvidia") -> DistributionTemplate:
             BuildProvider(provider_type="remote::nvidia"),
         ],
         "scoring": [BuildProvider(provider_type="inline::basic")],
-        "tool_runtime": [BuildProvider(provider_type="inline::rag-runtime")],
+        "tool_runtime": [],
         "files": [BuildProvider(provider_type="inline::localfs")],
     }
 
@@ -66,12 +66,7 @@ def get_distribution_template(name: str = "nvidia") -> DistributionTemplate:
         provider_id="nvidia",
     )
 
-    default_tool_groups = [
-        ToolGroupInput(
-            toolgroup_id="builtin::rag",
-            provider_id="rag-runtime",
-        ),
-    ]
+    default_tool_groups: list[ToolGroupInput] = []
 
     return DistributionTemplate(
         name=name,