Merge branch 'main' into custom-collection-name-vectordb

.github/CODEOWNERS

@@ -2,4 +2,4 @@
 
 # These owners will be the default owners for everything in
 # the repo. Unless a later match takes precedence,
-* @ashwinb @yanxi0830 @hardikjshah @raghotham @ehhuang @leseb @bbrowning @reluctantfuturist @mattf @slekkala1 @franciscojavierarceo
+* @ashwinb @raghotham @ehhuang @leseb @bbrowning @mattf @franciscojavierarceo

.github/actions/setup-typescript-client/action.yml

@@ -0,0 +1,35 @@
+name: Setup TypeScript client
+description: Conditionally checkout and link llama-stack-client-typescript based on client-version
+inputs:
+  client-version:
+    description: 'Client version (latest or published)'
+    required: true
+
+outputs:
+  ts-client-path:
+    description: 'Path or version to use for TypeScript client'
+    value: ${{ steps.set-path.outputs.ts-client-path }}
+
+runs:
+  using: "composite"
+  steps:
+    - name: Checkout TypeScript client (latest)
+      if: ${{ inputs.client-version == 'latest' }}
+      uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
+      with:
+        repository: llamastack/llama-stack-client-typescript
+        ref: main
+        path: .ts-client-checkout
+
+    - name: Set TS_CLIENT_PATH
+      id: set-path
+      shell: bash
+      run: |
+        if [ "${{ inputs.client-version }}" = "latest" ]; then
+          echo "ts-client-path=${{ github.workspace }}/.ts-client-checkout" >> $GITHUB_OUTPUT
+        elif [ "${{ inputs.client-version }}" = "published" ]; then
+          echo "ts-client-path=^0.3.2" >> $GITHUB_OUTPUT
+        else
+          echo "::error::Invalid client-version: ${{ inputs.client-version }}"
+          exit 1
+        fi

.github/workflows/integration-tests.yml

@@ -93,11 +93,27 @@ jobs:
           suite: ${{ matrix.config.suite }}
           inference-mode: 'replay'
 
+      - name: Setup Node.js for TypeScript client tests
+        if: ${{ matrix.client == 'server' }}
+        uses: actions/setup-node@39370e3970a6d050c480ffad4ff0ed4d3fdee5af # v4.1.0
+        with:
+          node-version: '20'
+          cache: 'npm'
+          cache-dependency-path: tests/integration/client-typescript/package-lock.json
+
+      - name: Setup TypeScript client
+        if: ${{ matrix.client == 'server' }}
+        id: setup-ts-client
+        uses: ./.github/actions/setup-typescript-client
+        with:
+          client-version: ${{ matrix.client-version }}
+
       - name: Run tests
         if: ${{ matrix.config.allowed_clients == null || contains(matrix.config.allowed_clients, matrix.client) }}
         uses: ./.github/actions/run-and-record-tests
         env:
           OPENAI_API_KEY: dummy
+          TS_CLIENT_PATH: ${{ steps.setup-ts-client.outputs.ts-client-path || '' }}
         with:
           stack-config: >-
             ${{ matrix.config.stack_config

.github/workflows/stainless-builds.yml

@@ -43,7 +43,41 @@ env:
   #   Stainless organization dashboard
 
 jobs:
+  compute-branch:
+    runs-on: ubuntu-latest
+    outputs:
+      preview_branch: ${{ steps.compute.outputs.preview_branch }}
+      base_branch: ${{ steps.compute.outputs.base_branch }}
+      merge_branch: ${{ steps.compute.outputs.merge_branch }}
+    steps:
+      - name: Compute branch names
+        id: compute
+        run: |
+          HEAD_REPO="${{ github.event.pull_request.head.repo.full_name }}"
+          BASE_REPO="${{ github.repository }}"
+          BRANCH_NAME="${{ github.event.pull_request.head.ref }}"
+          FORK_OWNER="${{ github.event.pull_request.head.repo.owner.login }}"
+
+          if [ "$HEAD_REPO" != "$BASE_REPO" ]; then
+            # Fork PR: prefix with fork owner for isolation
+            if [ -z "$FORK_OWNER" ]; then
+              echo "Error: Fork PR detected but fork owner is empty" >&2
+              exit 1
+            fi
+            PREVIEW_BRANCH="preview/${FORK_OWNER}/${BRANCH_NAME}"
+            BASE_BRANCH="preview/base/${FORK_OWNER}/${BRANCH_NAME}"
+          else
+            # Same-repo PR
+            PREVIEW_BRANCH="preview/${BRANCH_NAME}"
+            BASE_BRANCH="preview/base/${BRANCH_NAME}"
+          fi
+
+          echo "preview_branch=${PREVIEW_BRANCH}" >> $GITHUB_OUTPUT
+          echo "base_branch=${BASE_BRANCH}" >> $GITHUB_OUTPUT
+          echo "merge_branch=${PREVIEW_BRANCH}" >> $GITHUB_OUTPUT
+
   preview:
+    needs: compute-branch
     if: github.event.action != 'closed'
     runs-on: ubuntu-latest
     permissions:
@@ -59,8 +93,6 @@ jobs:
           ref: ${{ github.event.pull_request.head.sha }}
           fetch-depth: 2
 
-      # This action builds preview SDKs from the OpenAPI spec changes and
-      # posts/updates a comment on the PR with build results and links to the preview.
       - name: Run preview builds
         uses: stainless-api/upload-openapi-spec-action/preview@32823b096b4319c53ee948d702d9052873af485f # 1.6.0
         with:
@@ -73,8 +105,11 @@ jobs:
           base_sha: ${{ github.event.pull_request.base.sha }}
           base_ref: ${{ github.event.pull_request.base.ref }}
           head_sha: ${{ github.event.pull_request.head.sha }}
+          branch: ${{ needs.compute-branch.outputs.preview_branch }}
+          base_branch: ${{ needs.compute-branch.outputs.base_branch }}
 
   merge:
+    needs: compute-branch
     if: github.event.action == 'closed' && github.event.pull_request.merged == true
     runs-on: ubuntu-latest
     permissions:
@@ -91,11 +126,11 @@ jobs:
           fetch-depth: 2
 
       # Note that this only merges in changes that happened on the last build on
-      # preview/${{ github.head_ref }}. It's possible that there are OAS/config
+      # the computed preview branch. It's possible that there are OAS/config
-      # changes that haven't been built, if the preview-sdk job didn't finish
+      # changes that haven't been built, if the preview job didn't finish
       # before this step starts. In theory we want to wait for all builds
-      # against preview/${{ github.head_ref }} to complete, but assuming that
+      # against the preview branch to complete, but assuming that
-      # the preview-sdk job happens before the PR merge, it should be fine.
+      # the preview job happens before the PR merge, it should be fine.
       - name: Run merge build
         uses: stainless-api/upload-openapi-spec-action/merge@32823b096b4319c53ee948d702d9052873af485f # 1.6.0
         with:
@@ -108,3 +143,4 @@ jobs:
           base_sha: ${{ github.event.pull_request.base.sha }}
           base_ref: ${{ github.event.pull_request.base.ref }}
           head_sha: ${{ github.event.pull_request.head.sha }}
+          merge_branch: ${{ needs.compute-branch.outputs.merge_branch }}

.gitignore

@@ -35,3 +35,5 @@ docs/static/imported-files/
 docs/docs/api-deprecated/
 docs/docs/api-experimental/
 docs/docs/api/
+tests/integration/client-typescript/node_modules/
+.ts-client-checkout/

client-sdks/stainless/openapi.yml

@@ -9862,9 +9862,21 @@ components:
           title: Object
           default: vector_store.file
         attributes:
-          additionalProperties: true
+          additionalProperties:
+            anyOf:
+            - type: string
+              maxLength: 512
+            - type: number
+            - type: boolean
+            title: string | number | boolean
+          propertyNames:
+            type: string
+            maxLength: 64
           type: object
+          maxProperties: 16
           title: Attributes
+          description: Set of 16 key-value pairs that can be attached to an object. This can be useful for storing additional information about the object in a structured format, and querying for objects via API or the dashboard. Keys are strings with a maximum length of 64 characters. Values are strings with a maximum length of 512 characters, booleans, or numbers.
+          x-oaiTypeLabel: map
         chunking_strategy:
           oneOf:
           - $ref: '#/components/schemas/VectorStoreChunkingStrategyAuto'

docs/docs/providers/inference/remote_azure.mdx

@@ -24,7 +24,7 @@ https://learn.microsoft.com/en-us/azure/ai-foundry/openai/overview
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_key` | `SecretStr \| None` | No |  | Authentication credential for the provider |
-| `api_base` | `HttpUrl` | No |  | Azure API base for Azure (e.g., https://your-resource-name.openai.azure.com) |
+| `base_url` | `HttpUrl \| None` | No |  | Azure API base for Azure (e.g., https://your-resource-name.openai.azure.com/openai/v1) |
 | `api_version` | `str \| None` | No |  | Azure API version for Azure (e.g., 2024-12-01-preview) |
 | `api_type` | `str \| None` | No | azure | Azure API type for Azure (e.g., azure) |
 
@@ -32,7 +32,7 @@ https://learn.microsoft.com/en-us/azure/ai-foundry/openai/overview
 
 ```yaml
 api_key: ${env.AZURE_API_KEY:=}
-api_base: ${env.AZURE_API_BASE:=}
+base_url: ${env.AZURE_API_BASE:=}
 api_version: ${env.AZURE_API_VERSION:=}
 api_type: ${env.AZURE_API_TYPE:=}
 ```

docs/docs/providers/inference/remote_cerebras.mdx

@@ -17,11 +17,11 @@ Cerebras inference provider for running models on Cerebras Cloud platform.
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_key` | `SecretStr \| None` | No |  | Authentication credential for the provider |
-| `base_url` | `str` | No | https://api.cerebras.ai | Base URL for the Cerebras API |
+| `base_url` | `HttpUrl \| None` | No | https://api.cerebras.ai/v1 | Base URL for the Cerebras API |
 
 ## Sample Configuration
 
 ```yaml
-base_url: https://api.cerebras.ai
+base_url: https://api.cerebras.ai/v1
 api_key: ${env.CEREBRAS_API_KEY:=}
 ```

docs/docs/providers/inference/remote_databricks.mdx

@@ -17,11 +17,11 @@ Databricks inference provider for running models on Databricks' unified analytic
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_token` | `SecretStr \| None` | No |  | The Databricks API token |
-| `url` | `str \| None` | No |  | The URL for the Databricks model serving endpoint |
+| `base_url` | `HttpUrl \| None` | No |  | The URL for the Databricks model serving endpoint (should include /serving-endpoints path) |
 
 ## Sample Configuration
 
 ```yaml
-url: ${env.DATABRICKS_HOST:=}
+base_url: ${env.DATABRICKS_HOST:=}
 api_token: ${env.DATABRICKS_TOKEN:=}
 ```

docs/docs/providers/inference/remote_fireworks.mdx

@@ -17,11 +17,11 @@ Fireworks AI inference provider for Llama models and other AI models on the Fire
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_key` | `SecretStr \| None` | No |  | Authentication credential for the provider |
-| `url` | `str` | No | https://api.fireworks.ai/inference/v1 | The URL for the Fireworks server |
+| `base_url` | `HttpUrl \| None` | No | https://api.fireworks.ai/inference/v1 | The URL for the Fireworks server |
 
 ## Sample Configuration
 
 ```yaml
-url: https://api.fireworks.ai/inference/v1
+base_url: https://api.fireworks.ai/inference/v1
 api_key: ${env.FIREWORKS_API_KEY:=}
 ```

docs/docs/providers/inference/remote_groq.mdx

@@ -17,11 +17,11 @@ Groq inference provider for ultra-fast inference using Groq's LPU technology.
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_key` | `SecretStr \| None` | No |  | Authentication credential for the provider |
-| `url` | `str` | No | https://api.groq.com | The URL for the Groq AI server |
+| `base_url` | `HttpUrl \| None` | No | https://api.groq.com/openai/v1 | The URL for the Groq AI server |
 
 ## Sample Configuration
 
 ```yaml
-url: https://api.groq.com
+base_url: https://api.groq.com/openai/v1
 api_key: ${env.GROQ_API_KEY:=}
 ```

docs/docs/providers/inference/remote_llama-openai-compat.mdx

@@ -17,11 +17,11 @@ Llama OpenAI-compatible provider for using Llama models with OpenAI API format.
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_key` | `SecretStr \| None` | No |  | Authentication credential for the provider |
-| `openai_compat_api_base` | `str` | No | https://api.llama.com/compat/v1/ | The URL for the Llama API server |
+| `base_url` | `HttpUrl \| None` | No | https://api.llama.com/compat/v1/ | The URL for the Llama API server |
 
 ## Sample Configuration
 
 ```yaml
-openai_compat_api_base: https://api.llama.com/compat/v1/
+base_url: https://api.llama.com/compat/v1/
 api_key: ${env.LLAMA_API_KEY}
 ```

docs/docs/providers/inference/remote_nvidia.mdx

@@ -17,15 +17,13 @@ NVIDIA inference provider for accessing NVIDIA NIM models and AI services.
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_key` | `SecretStr \| None` | No |  | Authentication credential for the provider |
-| `url` | `str` | No | https://integrate.api.nvidia.com | A base url for accessing the NVIDIA NIM |
+| `base_url` | `HttpUrl \| None` | No | https://integrate.api.nvidia.com/v1 | A base url for accessing the NVIDIA NIM |
 | `timeout` | `int` | No | 60 | Timeout for the HTTP requests |
-| `append_api_version` | `bool` | No | True | When set to false, the API version will not be appended to the base_url. By default, it is true. |
 | `rerank_model_to_url` | `dict[str, str]` | No | `{'nv-rerank-qa-mistral-4b:1': 'https://ai.api.nvidia.com/v1/retrieval/nvidia/reranking', 'nvidia/nv-rerankqa-mistral-4b-v3': 'https://ai.api.nvidia.com/v1/retrieval/nvidia/nv-rerankqa-mistral-4b-v3/reranking', 'nvidia/llama-3.2-nv-rerankqa-1b-v2': 'https://ai.api.nvidia.com/v1/retrieval/nvidia/llama-3_2-nv-rerankqa-1b-v2/reranking'}` | Mapping of rerank model identifiers to their API endpoints.  |
 
 ## Sample Configuration
 
 ```yaml
-url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com}
+base_url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com/v1}
 api_key: ${env.NVIDIA_API_KEY:=}
-append_api_version: ${env.NVIDIA_APPEND_API_VERSION:=True}
 ```

docs/docs/providers/inference/remote_ollama.mdx

@@ -16,10 +16,10 @@ Ollama inference provider for running local models through the Ollama runtime.
 |-------|------|----------|---------|-------------|
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
-| `url` | `str` | No | http://localhost:11434 |  |
+| `base_url` | `HttpUrl \| None` | No | http://localhost:11434/v1 |  |
 
 ## Sample Configuration
 
 ```yaml
-url: ${env.OLLAMA_URL:=http://localhost:11434}
+base_url: ${env.OLLAMA_URL:=http://localhost:11434/v1}
 ```

docs/docs/providers/inference/remote_openai.mdx

@@ -17,7 +17,7 @@ OpenAI inference provider for accessing GPT models and other OpenAI services.
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_key` | `SecretStr \| None` | No |  | Authentication credential for the provider |
-| `base_url` | `str` | No | https://api.openai.com/v1 | Base URL for OpenAI API |
+| `base_url` | `HttpUrl \| None` | No | https://api.openai.com/v1 | Base URL for OpenAI API |
 
 ## Sample Configuration
 

docs/docs/providers/inference/remote_passthrough.mdx

@@ -17,11 +17,11 @@ Passthrough inference provider for connecting to any external inference service
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_key` | `SecretStr \| None` | No |  | Authentication credential for the provider |
-| `url` | `str` | No |  | The URL for the passthrough endpoint |
+| `base_url` | `HttpUrl \| None` | No |  | The URL for the passthrough endpoint |
 
 ## Sample Configuration
 
 ```yaml
-url: ${env.PASSTHROUGH_URL}
+base_url: ${env.PASSTHROUGH_URL}
 api_key: ${env.PASSTHROUGH_API_KEY}
 ```

docs/docs/providers/inference/remote_runpod.mdx

@@ -17,11 +17,11 @@ RunPod inference provider for running models on RunPod's cloud GPU platform.
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_token` | `SecretStr \| None` | No |  | The API token |
-| `url` | `str \| None` | No |  | The URL for the Runpod model serving endpoint |
+| `base_url` | `HttpUrl \| None` | No |  | The URL for the Runpod model serving endpoint |
 
 ## Sample Configuration
 
 ```yaml
-url: ${env.RUNPOD_URL:=}
+base_url: ${env.RUNPOD_URL:=}
 api_token: ${env.RUNPOD_API_TOKEN}
 ```

docs/docs/providers/inference/remote_sambanova.mdx

@@ -17,11 +17,11 @@ SambaNova inference provider for running models on SambaNova's dataflow architec
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_key` | `SecretStr \| None` | No |  | Authentication credential for the provider |
-| `url` | `str` | No | https://api.sambanova.ai/v1 | The URL for the SambaNova AI server |
+| `base_url` | `HttpUrl \| None` | No | https://api.sambanova.ai/v1 | The URL for the SambaNova AI server |
 
 ## Sample Configuration
 
 ```yaml
-url: https://api.sambanova.ai/v1
+base_url: https://api.sambanova.ai/v1
 api_key: ${env.SAMBANOVA_API_KEY:=}
 ```

docs/docs/providers/inference/remote_tgi.mdx

@@ -16,10 +16,10 @@ Text Generation Inference (TGI) provider for HuggingFace model serving.
 |-------|------|----------|---------|-------------|
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
-| `url` | `str` | No |  | The URL for the TGI serving endpoint |
+| `base_url` | `HttpUrl \| None` | No |  | The URL for the TGI serving endpoint (should include /v1 path) |
 
 ## Sample Configuration
 
 ```yaml
-url: ${env.TGI_URL:=}
+base_url: ${env.TGI_URL:=}
 ```

docs/docs/providers/inference/remote_together.mdx

@@ -17,11 +17,11 @@ Together AI inference provider for open-source models and collaborative AI devel
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_key` | `SecretStr \| None` | No |  | Authentication credential for the provider |
-| `url` | `str` | No | https://api.together.xyz/v1 | The URL for the Together AI server |
+| `base_url` | `HttpUrl \| None` | No | https://api.together.xyz/v1 | The URL for the Together AI server |
 
 ## Sample Configuration
 
 ```yaml
-url: https://api.together.xyz/v1
+base_url: https://api.together.xyz/v1
 api_key: ${env.TOGETHER_API_KEY:=}
 ```

docs/docs/providers/inference/remote_vllm.mdx

@@ -17,14 +17,14 @@ Remote vLLM inference provider for connecting to vLLM servers.
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_token` | `SecretStr \| None` | No |  | The API token |
-| `url` | `str \| None` | No |  | The URL for the vLLM model serving endpoint |
+| `base_url` | `HttpUrl \| None` | No |  | The URL for the vLLM model serving endpoint |
 | `max_tokens` | `int` | No | 4096 | Maximum number of tokens to generate. |
 | `tls_verify` | `bool \| str` | No | True | Whether to verify TLS certificates. Can be a boolean or a path to a CA certificate file. |
 
 ## Sample Configuration
 
 ```yaml
-url: ${env.VLLM_URL:=}
+base_url: ${env.VLLM_URL:=}
 max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
 api_token: ${env.VLLM_API_TOKEN:=fake}
 tls_verify: ${env.VLLM_TLS_VERIFY:=true}

docs/docs/providers/inference/remote_watsonx.mdx

@@ -17,14 +17,14 @@ IBM WatsonX inference provider for accessing AI models on IBM's WatsonX platform
 | `allowed_models` | `list[str] \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `refresh_models` | `bool` | No | False | Whether to refresh models periodically from the provider |
 | `api_key` | `SecretStr \| None` | No |  | Authentication credential for the provider |
-| `url` | `str` | No | https://us-south.ml.cloud.ibm.com | A base url for accessing the watsonx.ai |
+| `base_url` | `HttpUrl \| None` | No | https://us-south.ml.cloud.ibm.com | A base url for accessing the watsonx.ai |
 | `project_id` | `str \| None` | No |  | The watsonx.ai project ID |
 | `timeout` | `int` | No | 60 | Timeout for the HTTP requests |
 
 ## Sample Configuration
 
 ```yaml
-url: ${env.WATSONX_BASE_URL:=https://us-south.ml.cloud.ibm.com}
+base_url: ${env.WATSONX_BASE_URL:=https://us-south.ml.cloud.ibm.com}
 api_key: ${env.WATSONX_API_KEY:=}
 project_id: ${env.WATSONX_PROJECT_ID:=}
 ```

docs/package-lock.json

@@ -10712,12 +10712,6 @@
       "integrity": "sha512-QMUezzXWII9EV5aTFXW1UBVUO77wYPpjqIF8/AviUCThNeSYZykpoTixUeaNNBwmCev0AMDWMAni+f8Hxb1IFw==",
       "license": "Unlicense"
     },
-    "node_modules/fs.realpath": {
-      "version": "1.0.0",
-      "resolved": "https://registry.npmjs.org/fs.realpath/-/fs.realpath-1.0.0.tgz",
-      "integrity": "sha512-OO0pH2lK6a0hZnAdau5ItzHPI6pUlvI7jMVnxUQRtw4owF2wk8lOSabtGDCTP4Ggrg2MbGnWO9X8K1t4+fGMDw==",
-      "license": "ISC"
-    },
     "node_modules/fsevents": {
       "version": "2.3.3",
       "resolved": "https://registry.npmjs.org/fsevents/-/fsevents-2.3.3.tgz",
@@ -10821,21 +10815,20 @@
       "license": "ISC"
     },
     "node_modules/glob": {
-      "version": "7.2.3",
+      "version": "10.5.0",
-      "resolved": "https://registry.npmjs.org/glob/-/glob-7.2.3.tgz",
+      "resolved": "https://registry.npmjs.org/glob/-/glob-10.5.0.tgz",
-      "integrity": "sha512-nFR0zLpU2YCaRxwoCJvL6UvCH2JFyFVIvwTLsIf21AuHlMskA1hhTdk+LlYJtOlYt9v6dvszD2BGRqBL+iQK9Q==",
+      "integrity": "sha512-DfXN8DfhJ7NH3Oe7cFmu3NCu1wKbkReJ8TorzSAFbSKrlNaQSKfIzqYqVY8zlbs2NLBbWpRiU52GX2PbaBVNkg==",
-      "deprecated": "Glob versions prior to v9 are no longer supported",
       "license": "ISC",
       "dependencies": {
-        "fs.realpath": "^1.0.0",
+        "foreground-child": "^3.1.0",
-        "inflight": "^1.0.4",
+        "jackspeak": "^3.1.2",
-        "inherits": "2",
+        "minimatch": "^9.0.4",
-        "minimatch": "^3.1.1",
+        "minipass": "^7.1.2",
-        "once": "^1.3.0",
+        "package-json-from-dist": "^1.0.0",
-        "path-is-absolute": "^1.0.0"
+        "path-scurry": "^1.11.1"
       },
-      "engines": {
+      "bin": {
-        "node": "*"
+        "glob": "dist/esm/bin.mjs"
       },
       "funding": {
         "url": "https://github.com/sponsors/isaacs"
@@ -10859,26 +10852,19 @@
       "integrity": "sha512-lkX1HJXwyMcprw/5YUZc2s7DrpAiHB21/V+E1rHUrVNokkvB6bqMzT0VfV6/86ZNabt1k14YOIaT7nDvOX3Iiw==",
       "license": "BSD-2-Clause"
     },
-    "node_modules/glob/node_modules/brace-expansion": {
-      "version": "1.1.12",
-      "resolved": "https://registry.npmjs.org/brace-expansion/-/brace-expansion-1.1.12.tgz",
-      "integrity": "sha512-9T9UjW3r0UW5c1Q7GTwllptXwhvYmEzFhzMfZ9H7FQWt+uZePjZPjBP/W1ZEyZ1twGWom5/56TF4lPcqjnDHcg==",
-      "license": "MIT",
-      "dependencies": {
-        "balanced-match": "^1.0.0",
-        "concat-map": "0.0.1"
-      }
-    },
     "node_modules/glob/node_modules/minimatch": {
-      "version": "3.1.2",
+      "version": "9.0.5",
-      "resolved": "https://registry.npmjs.org/minimatch/-/minimatch-3.1.2.tgz",
+      "resolved": "https://registry.npmjs.org/minimatch/-/minimatch-9.0.5.tgz",
-      "integrity": "sha512-J7p63hRiAjw1NDEww1W7i37+ByIrOWO5XQQAzZ3VOcL0PNybwpfmV/N05zFAzwQ9USyEcX6t3UO+K5aqBQOIHw==",
+      "integrity": "sha512-G6T0ZX48xgozx7587koeX9Ys2NYy6Gmv//P89sEte9V9whIapMNF4idKxnW2QtCcLiTWlb/wfCabAtAFWhhBow==",
       "license": "ISC",
       "dependencies": {
-        "brace-expansion": "^1.1.7"
+        "brace-expansion": "^2.0.1"
       },
       "engines": {
-        "node": "*"
+        "node": ">=16 || 14 >=14.17"
+      },
+      "funding": {
+        "url": "https://github.com/sponsors/isaacs"
       }
     },
     "node_modules/global-dirs": {
@@ -11792,17 +11778,6 @@
         "node": ">=12"
       }
     },
-    "node_modules/inflight": {
-      "version": "1.0.6",
-      "resolved": "https://registry.npmjs.org/inflight/-/inflight-1.0.6.tgz",
-      "integrity": "sha512-k92I/b08q4wvFscXCLvqfsHCrjrF7yiXsQuIVvVE7N82W3+aqpzuUdBbfhWcy/FZR3/4IgflMgKLOsvPDrGCJA==",
-      "deprecated": "This module is not supported, and leaks memory. Do not use it. Check out lru-cache if you want a good and tested way to coalesce async requests by a key value, which is much more comprehensive and powerful.",
-      "license": "ISC",
-      "dependencies": {
-        "once": "^1.3.0",
-        "wrappy": "1"
-      }
-    },
     "node_modules/inherits": {
       "version": "2.0.4",
       "resolved": "https://registry.npmjs.org/inherits/-/inherits-2.0.4.tgz",
@@ -15570,15 +15545,6 @@
         "node": ">= 0.8"
       }
     },
-    "node_modules/once": {
-      "version": "1.4.0",
-      "resolved": "https://registry.npmjs.org/once/-/once-1.4.0.tgz",
-      "integrity": "sha512-lNaJgI+2Q5URQBkccEKHTQOPaXdUxnZZElQTZY0MFUAuaEqe1E+Nyvgdz/aIyNi6Z9MzO5dv1H8n58/GELp3+w==",
-      "license": "ISC",
-      "dependencies": {
-        "wrappy": "1"
-      }
-    },
     "node_modules/onetime": {
       "version": "5.1.2",
       "resolved": "https://registry.npmjs.org/onetime/-/onetime-5.1.2.tgz",
@@ -15955,15 +15921,6 @@
         "node": "^12.20.0 || ^14.13.1 || >=16.0.0"
       }
     },
-    "node_modules/path-is-absolute": {
-      "version": "1.0.1",
-      "resolved": "https://registry.npmjs.org/path-is-absolute/-/path-is-absolute-1.0.1.tgz",
-      "integrity": "sha512-AVbw3UJ2e9bq64vSaS9Am0fje1Pa8pbGqTTsmXfaIiMpnr5DlDhfJOuLj9Sf95ZPVDAUerDfEk88MPmPe7UCQg==",
-      "license": "MIT",
-      "engines": {
-        "node": ">=0.10.0"
-      }
-    },
     "node_modules/path-is-inside": {
       "version": "1.0.2",
       "resolved": "https://registry.npmjs.org/path-is-inside/-/path-is-inside-1.0.2.tgz",
@@ -20038,41 +19995,6 @@
         "node": ">= 6"
       }
     },
-    "node_modules/sucrase/node_modules/glob": {
-      "version": "10.4.5",
-      "resolved": "https://registry.npmjs.org/glob/-/glob-10.4.5.tgz",
-      "integrity": "sha512-7Bv8RF0k6xjo7d4A/PxYLbUCfb6c+Vpd2/mB2yRDlew7Jb5hEXiCD9ibfO7wpk8i4sevK6DFny9h7EYbM3/sHg==",
-      "license": "ISC",
-      "dependencies": {
-        "foreground-child": "^3.1.0",
-        "jackspeak": "^3.1.2",
-        "minimatch": "^9.0.4",
-        "minipass": "^7.1.2",
-        "package-json-from-dist": "^1.0.0",
-        "path-scurry": "^1.11.1"
-      },
-      "bin": {
-        "glob": "dist/esm/bin.mjs"
-      },
-      "funding": {
-        "url": "https://github.com/sponsors/isaacs"
-      }
-    },
-    "node_modules/sucrase/node_modules/minimatch": {
-      "version": "9.0.5",
-      "resolved": "https://registry.npmjs.org/minimatch/-/minimatch-9.0.5.tgz",
-      "integrity": "sha512-G6T0ZX48xgozx7587koeX9Ys2NYy6Gmv//P89sEte9V9whIapMNF4idKxnW2QtCcLiTWlb/wfCabAtAFWhhBow==",
-      "license": "ISC",
-      "dependencies": {
-        "brace-expansion": "^2.0.1"
-      },
-      "engines": {
-        "node": ">=16 || 14 >=14.17"
-      },
-      "funding": {
-        "url": "https://github.com/sponsors/isaacs"
-      }
-    },
     "node_modules/supports-color": {
       "version": "7.2.0",
       "resolved": "https://registry.npmjs.org/supports-color/-/supports-color-7.2.0.tgz",
@@ -21620,12 +21542,6 @@
         "url": "https://github.com/chalk/strip-ansi?sponsor=1"
       }
     },
-    "node_modules/wrappy": {
-      "version": "1.0.2",
-      "resolved": "https://registry.npmjs.org/wrappy/-/wrappy-1.0.2.tgz",
-      "integrity": "sha512-l4Sp/DRseor9wL6EvV2+TuQn63dMkPjZ/sp9XkghTEbV9KlPS1xUsZ3u7/IQO4wxtcFB4bgpQPRcR3QCvezPcQ==",
-      "license": "ISC"
-    },
     "node_modules/write-file-atomic": {
       "version": "3.0.3",
       "resolved": "https://registry.npmjs.org/write-file-atomic/-/write-file-atomic-3.0.3.tgz",

docs/package.json

@@ -31,6 +31,9 @@
     "react-dom": "^19.0.0",
     "remark-code-import": "^1.2.0"
   },
+  "overrides": {
+    "glob": "^10.5.0"
+  },
   "browserslist": {
     "production": [
       ">0.5%",

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

@@ -6705,9 +6705,21 @@ components:
           title: Object
           default: vector_store.file
         attributes:
-          additionalProperties: true
+          additionalProperties:
+            anyOf:
+            - type: string
+              maxLength: 512
+            - type: number
+            - type: boolean
+            title: string | number | boolean
+          propertyNames:
+            type: string
+            maxLength: 64
           type: object
+          maxProperties: 16
           title: Attributes
+          description: Set of 16 key-value pairs that can be attached to an object. This can be useful for storing additional information about the object in a structured format, and querying for objects via API or the dashboard. Keys are strings with a maximum length of 64 characters. Values are strings with a maximum length of 512 characters, booleans, or numbers.
+          x-oaiTypeLabel: map
         chunking_strategy:
           oneOf:
           - $ref: '#/components/schemas/VectorStoreChunkingStrategyAuto'

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

@@ -6061,9 +6061,21 @@ components:
           title: Object
           default: vector_store.file
         attributes:
-          additionalProperties: true
+          additionalProperties:
+            anyOf:
+            - type: string
+              maxLength: 512
+            - type: number
+            - type: boolean
+            title: string | number | boolean
+          propertyNames:
+            type: string
+            maxLength: 64
           type: object
+          maxProperties: 16
           title: Attributes
+          description: Set of 16 key-value pairs that can be attached to an object. This can be useful for storing additional information about the object in a structured format, and querying for objects via API or the dashboard. Keys are strings with a maximum length of 64 characters. Values are strings with a maximum length of 512 characters, booleans, or numbers.
+          x-oaiTypeLabel: map
         chunking_strategy:
           oneOf:
           - $ref: '#/components/schemas/VectorStoreChunkingStrategyAuto'

docs/static/llama-stack-spec.yaml

@@ -8883,9 +8883,21 @@ components:
           title: Object
           default: vector_store.file
         attributes:
-          additionalProperties: true
+          additionalProperties:
+            anyOf:
+            - type: string
+              maxLength: 512
+            - type: number
+            - type: boolean
+            title: string | number | boolean
+          propertyNames:
+            type: string
+            maxLength: 64
           type: object
+          maxProperties: 16
           title: Attributes
+          description: Set of 16 key-value pairs that can be attached to an object. This can be useful for storing additional information about the object in a structured format, and querying for objects via API or the dashboard. Keys are strings with a maximum length of 64 characters. Values are strings with a maximum length of 512 characters, booleans, or numbers.
+          x-oaiTypeLabel: map
         chunking_strategy:
           oneOf:
           - $ref: '#/components/schemas/VectorStoreChunkingStrategyAuto'

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

@@ -9862,9 +9862,21 @@ components:
           title: Object
           default: vector_store.file
         attributes:
-          additionalProperties: true
+          additionalProperties:
+            anyOf:
+            - type: string
+              maxLength: 512
+            - type: number
+            - type: boolean
+            title: string | number | boolean
+          propertyNames:
+            type: string
+            maxLength: 64
           type: object
+          maxProperties: 16
           title: Attributes
+          description: Set of 16 key-value pairs that can be attached to an object. This can be useful for storing additional information about the object in a structured format, and querying for objects via API or the dashboard. Keys are strings with a maximum length of 64 characters. Values are strings with a maximum length of 512 characters, booleans, or numbers.
+          x-oaiTypeLabel: map
         chunking_strategy:
           oneOf:
           - $ref: '#/components/schemas/VectorStoreChunkingStrategyAuto'

scripts/docker.sh

@@ -287,9 +287,9 @@ start_container() {
     # On macOS/Windows, use host.docker.internal to reach host from container
     # On Linux with --network host, use localhost
     if [[ "$(uname)" == "Darwin" ]] || [[ "$(uname)" == *"MINGW"* ]]; then
-        OLLAMA_URL="${OLLAMA_URL:-http://host.docker.internal:11434}"
+        OLLAMA_URL="${OLLAMA_URL:-http://host.docker.internal:11434/v1}"
     else
-        OLLAMA_URL="${OLLAMA_URL:-http://localhost:11434}"
+        OLLAMA_URL="${OLLAMA_URL:-http://localhost:11434/v1}"
     fi
     DOCKER_ENV_VARS="$DOCKER_ENV_VARS -e OLLAMA_URL=$OLLAMA_URL"
 

scripts/get_setup_env.py

@@ -16,16 +16,16 @@ import sys
 from tests.integration.suites import SETUP_DEFINITIONS, SUITE_DEFINITIONS
 
 
-def get_setup_env_vars(setup_name, suite_name=None):
+def get_setup_config(setup_name, suite_name=None):
     """
-    Get environment variables for a setup, with optional suite default fallback.
+    Get full configuration (env vars + defaults) for a setup.
 
     Args:
         setup_name: Name of the setup (e.g., 'ollama', 'gpt')
         suite_name: Optional suite name to get default setup if setup_name is None
 
     Returns:
-        Dictionary of environment variables
+        Dictionary with 'env' and 'defaults' keys
     """
     # If no setup specified, try to get default from suite
     if not setup_name and suite_name:
@@ -34,7 +34,7 @@ def get_setup_env_vars(setup_name, suite_name=None):
             setup_name = suite.default_setup
 
     if not setup_name:
-        return {}
+        return {"env": {}, "defaults": {}}
 
     setup = SETUP_DEFINITIONS.get(setup_name)
     if not setup:
@@ -44,27 +44,31 @@ def get_setup_env_vars(setup_name, suite_name=None):
         )
         sys.exit(1)
 
-    return setup.env
+    return {"env": setup.env, "defaults": setup.defaults}
 
 
 def main():
-    parser = argparse.ArgumentParser(description="Extract environment variables from a test setup")
+    parser = argparse.ArgumentParser(description="Extract environment variables and defaults from a test setup")
     parser.add_argument("--setup", help="Setup name (e.g., ollama, gpt)")
     parser.add_argument("--suite", help="Suite name to get default setup from if --setup not provided")
     parser.add_argument("--format", choices=["bash", "json"], default="bash", help="Output format (default: bash)")
 
     args = parser.parse_args()
 
-    env_vars = get_setup_env_vars(args.setup, args.suite)
+    config = get_setup_config(args.setup, args.suite)
 
     if args.format == "bash":
-        # Output as bash export statements
+        # Output env vars as bash export statements
-        for key, value in env_vars.items():
+        for key, value in config["env"].items():
             print(f"export {key}='{value}'")
+        # Output defaults as bash export statements with LLAMA_STACK_TEST_ prefix
+        for key, value in config["defaults"].items():
+            env_key = f"LLAMA_STACK_TEST_{key.upper()}"
+            print(f"export {env_key}='{value}'")
     elif args.format == "json":
         import json
 
-        print(json.dumps(env_vars))
+        print(json.dumps(config))
 
 
 if __name__ == "__main__":

scripts/install.sh

@@ -640,7 +640,7 @@ cmd=( run -d "${PLATFORM_OPTS[@]}" --name llama-stack \
       --network llama-net \
       -p "${PORT}:${PORT}" \
       "${server_env_opts[@]}" \
-      -e OLLAMA_URL="http://ollama-server:${OLLAMA_PORT}" \
+      -e OLLAMA_URL="http://ollama-server:${OLLAMA_PORT}/v1" \
       "${SERVER_IMAGE}" --port "${PORT}")
 
 log "🦙 Starting Llama Stack..."

scripts/integration-tests.sh

@@ -20,6 +20,7 @@ TEST_PATTERN=""
 INFERENCE_MODE="replay"
 EXTRA_PARAMS=""
 COLLECT_ONLY=false
+TYPESCRIPT_ONLY=false
 
 # Function to display usage
 usage() {
@@ -34,6 +35,7 @@ Options:
     --subdirs STRING         Comma-separated list of test subdirectories to run (overrides suite)
     --pattern STRING         Regex pattern to pass to pytest -k
     --collect-only           Collect tests only without running them (skips server startup)
+    --typescript-only        Skip Python tests and run only TypeScript client tests
     --help                   Show this help message
 
 Suites are defined in tests/integration/suites.py and define which tests to run.
@@ -90,6 +92,10 @@ while [[ $# -gt 0 ]]; do
         COLLECT_ONLY=true
         shift
         ;;
+    --typescript-only)
+        TYPESCRIPT_ONLY=true
+        shift
+        ;;
     --help)
         usage
         exit 0
@@ -181,6 +187,10 @@ echo "$SETUP_ENV"
 eval "$SETUP_ENV"
 echo ""
 
+# Export suite and setup names for TypeScript tests
+export LLAMA_STACK_TEST_SUITE="$TEST_SUITE"
+export LLAMA_STACK_TEST_SETUP="$TEST_SETUP"
+
 ROOT_DIR="$THIS_DIR/.."
 cd $ROOT_DIR
 
@@ -212,6 +222,71 @@ find_available_port() {
     return 1
 }
 
+run_client_ts_tests() {
+    if ! command -v npm &>/dev/null; then
+        echo "npm could not be found; ensure Node.js is installed"
+        return 1
+    fi
+
+    pushd tests/integration/client-typescript >/dev/null
+
+    # Determine if TS_CLIENT_PATH is a directory path or an npm version
+    if [[ -d "$TS_CLIENT_PATH" ]]; then
+        # It's a directory path - use local checkout
+        if [[ ! -f "$TS_CLIENT_PATH/package.json" ]]; then
+            echo "Error: $TS_CLIENT_PATH exists but doesn't look like llama-stack-client-typescript (no package.json)"
+            popd >/dev/null
+            return 1
+        fi
+        echo "Using local llama-stack-client-typescript from: $TS_CLIENT_PATH"
+
+        # Build the TypeScript client first
+        echo "Building TypeScript client..."
+        pushd "$TS_CLIENT_PATH" >/dev/null
+        npm install --silent
+        npm run build --silent
+        popd >/dev/null
+
+        # Install other dependencies first
+        if [[ "${CI:-}" == "true" || "${CI:-}" == "1" ]]; then
+            npm ci --silent
+        else
+            npm install --silent
+        fi
+
+        # Then install the client from local directory
+        echo "Installing llama-stack-client from: $TS_CLIENT_PATH"
+        npm install "$TS_CLIENT_PATH" --silent
+    else
+        # It's an npm version specifier - install from npm
+        echo "Installing llama-stack-client@${TS_CLIENT_PATH} from npm"
+        if [[ "${CI:-}" == "true" || "${CI:-}" == "1" ]]; then
+            npm ci --silent
+            npm install "llama-stack-client@${TS_CLIENT_PATH}" --silent
+        else
+            npm install "llama-stack-client@${TS_CLIENT_PATH}" --silent
+        fi
+    fi
+
+    # Verify installation
+    echo "Verifying llama-stack-client installation..."
+    if npm list llama-stack-client 2>/dev/null | grep -q llama-stack-client; then
+        echo "✅ llama-stack-client successfully installed"
+        npm list llama-stack-client
+    else
+        echo "❌ llama-stack-client not found in node_modules"
+        echo "Installed packages:"
+        npm list --depth=0
+        popd >/dev/null
+        return 1
+    fi
+
+    echo "Running TypeScript tests for suite $TEST_SUITE (setup $TEST_SETUP)"
+    npm test
+
+    popd >/dev/null
+}
+
 # Start Llama Stack Server if needed
 if [[ "$STACK_CONFIG" == *"server:"* && "$COLLECT_ONLY" == false ]]; then
     # Find an available port for the server
@@ -221,6 +296,7 @@ if [[ "$STACK_CONFIG" == *"server:"* && "$COLLECT_ONLY" == false ]]; then
         exit 1
     fi
     export LLAMA_STACK_PORT
+    export TEST_API_BASE_URL="http://localhost:$LLAMA_STACK_PORT"
     echo "Will use port: $LLAMA_STACK_PORT"
 
     stop_server() {
@@ -298,6 +374,7 @@ if [[ "$STACK_CONFIG" == *"docker:"* && "$COLLECT_ONLY" == false ]]; then
         exit 1
     fi
     export LLAMA_STACK_PORT
+    export TEST_API_BASE_URL="http://localhost:$LLAMA_STACK_PORT"
     echo "Will use port: $LLAMA_STACK_PORT"
 
     echo "=== Building Docker Image for distribution: $DISTRO ==="
@@ -473,7 +550,9 @@ if [[ -n "$STACK_CONFIG" ]]; then
     STACK_CONFIG_ARG="--stack-config=$STACK_CONFIG"
 fi
 
-pytest -s -v $PYTEST_TARGET \
+# Run Python tests unless typescript-only mode
+if [[ "$TYPESCRIPT_ONLY" == "false" ]]; then
+    pytest -s -v $PYTEST_TARGET \
         $STACK_CONFIG_ARG \
         --inference-mode="$INFERENCE_MODE" \
         -k "$PYTEST_PATTERN" \
@@ -482,7 +561,12 @@ pytest -s -v $PYTEST_TARGET \
         --embedding-model=sentence-transformers/nomic-ai/nomic-embed-text-v1.5 \
         --color=yes $EXTRA_PARAMS \
         --capture=tee-sys
-exit_code=$?
+    exit_code=$?
+else
+    echo "Skipping Python tests (--typescript-only mode)"
+    exit_code=0
+fi
+
 set +x
 set -e
 
@@ -506,5 +590,10 @@ else
     exit 1
 fi
 
+# Run TypeScript client tests if TS_CLIENT_PATH is set
+if [[ $exit_code -eq 0 && -n "${TS_CLIENT_PATH:-}" && "${LLAMA_STACK_TEST_STACK_CONFIG_TYPE:-}" == "server" ]]; then
+    run_client_ts_tests
+fi
+
 echo ""
 echo "=== Integration Tests Complete ==="

src/llama_stack/distributions/ci-tests/run-with-postgres-store.yaml

@@ -17,32 +17,32 @@ providers:
   - provider_id: ${env.CEREBRAS_API_KEY:+cerebras}
     provider_type: remote::cerebras
     config:
-      base_url: https://api.cerebras.ai
+      base_url: https://api.cerebras.ai/v1
       api_key: ${env.CEREBRAS_API_KEY:=}
   - provider_id: ${env.OLLAMA_URL:+ollama}
     provider_type: remote::ollama
     config:
-      url: ${env.OLLAMA_URL:=http://localhost:11434}
+      base_url: ${env.OLLAMA_URL:=http://localhost:11434/v1}
   - provider_id: ${env.VLLM_URL:+vllm}
     provider_type: remote::vllm
     config:
-      url: ${env.VLLM_URL:=}
+      base_url: ${env.VLLM_URL:=}
       max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
       api_token: ${env.VLLM_API_TOKEN:=fake}
       tls_verify: ${env.VLLM_TLS_VERIFY:=true}
   - provider_id: ${env.TGI_URL:+tgi}
     provider_type: remote::tgi
     config:
-      url: ${env.TGI_URL:=}
+      base_url: ${env.TGI_URL:=}
   - provider_id: fireworks
     provider_type: remote::fireworks
     config:
-      url: https://api.fireworks.ai/inference/v1
+      base_url: https://api.fireworks.ai/inference/v1
       api_key: ${env.FIREWORKS_API_KEY:=}
   - provider_id: together
     provider_type: remote::together
     config:
-      url: https://api.together.xyz/v1
+      base_url: https://api.together.xyz/v1
       api_key: ${env.TOGETHER_API_KEY:=}
   - provider_id: bedrock
     provider_type: remote::bedrock
@@ -52,9 +52,8 @@ providers:
   - provider_id: ${env.NVIDIA_API_KEY:+nvidia}
     provider_type: remote::nvidia
     config:
-      url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com}
+      base_url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com/v1}
       api_key: ${env.NVIDIA_API_KEY:=}
-      append_api_version: ${env.NVIDIA_APPEND_API_VERSION:=True}
   - provider_id: openai
     provider_type: remote::openai
     config:
@@ -76,18 +75,18 @@ providers:
   - provider_id: groq
     provider_type: remote::groq
     config:
-      url: https://api.groq.com
+      base_url: https://api.groq.com/openai/v1
       api_key: ${env.GROQ_API_KEY:=}
   - provider_id: sambanova
     provider_type: remote::sambanova
     config:
-      url: https://api.sambanova.ai/v1
+      base_url: https://api.sambanova.ai/v1
       api_key: ${env.SAMBANOVA_API_KEY:=}
   - provider_id: ${env.AZURE_API_KEY:+azure}
     provider_type: remote::azure
     config:
       api_key: ${env.AZURE_API_KEY:=}
-      api_base: ${env.AZURE_API_BASE:=}
+      base_url: ${env.AZURE_API_BASE:=}
       api_version: ${env.AZURE_API_VERSION:=}
       api_type: ${env.AZURE_API_TYPE:=}
   - provider_id: sentence-transformers

src/llama_stack/distributions/ci-tests/run.yaml

@@ -17,32 +17,32 @@ providers:
   - provider_id: ${env.CEREBRAS_API_KEY:+cerebras}
     provider_type: remote::cerebras
     config:
-      base_url: https://api.cerebras.ai
+      base_url: https://api.cerebras.ai/v1
       api_key: ${env.CEREBRAS_API_KEY:=}
   - provider_id: ${env.OLLAMA_URL:+ollama}
     provider_type: remote::ollama
     config:
-      url: ${env.OLLAMA_URL:=http://localhost:11434}
+      base_url: ${env.OLLAMA_URL:=http://localhost:11434/v1}
   - provider_id: ${env.VLLM_URL:+vllm}
     provider_type: remote::vllm
     config:
-      url: ${env.VLLM_URL:=}
+      base_url: ${env.VLLM_URL:=}
       max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
       api_token: ${env.VLLM_API_TOKEN:=fake}
       tls_verify: ${env.VLLM_TLS_VERIFY:=true}
   - provider_id: ${env.TGI_URL:+tgi}
     provider_type: remote::tgi
     config:
-      url: ${env.TGI_URL:=}
+      base_url: ${env.TGI_URL:=}
   - provider_id: fireworks
     provider_type: remote::fireworks
     config:
-      url: https://api.fireworks.ai/inference/v1
+      base_url: https://api.fireworks.ai/inference/v1
       api_key: ${env.FIREWORKS_API_KEY:=}
   - provider_id: together
     provider_type: remote::together
     config:
-      url: https://api.together.xyz/v1
+      base_url: https://api.together.xyz/v1
       api_key: ${env.TOGETHER_API_KEY:=}
   - provider_id: bedrock
     provider_type: remote::bedrock
@@ -52,9 +52,8 @@ providers:
   - provider_id: ${env.NVIDIA_API_KEY:+nvidia}
     provider_type: remote::nvidia
     config:
-      url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com}
+      base_url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com/v1}
       api_key: ${env.NVIDIA_API_KEY:=}
-      append_api_version: ${env.NVIDIA_APPEND_API_VERSION:=True}
   - provider_id: openai
     provider_type: remote::openai
     config:
@@ -76,18 +75,18 @@ providers:
   - provider_id: groq
     provider_type: remote::groq
     config:
-      url: https://api.groq.com
+      base_url: https://api.groq.com/openai/v1
       api_key: ${env.GROQ_API_KEY:=}
   - provider_id: sambanova
     provider_type: remote::sambanova
     config:
-      url: https://api.sambanova.ai/v1
+      base_url: https://api.sambanova.ai/v1
       api_key: ${env.SAMBANOVA_API_KEY:=}
   - provider_id: ${env.AZURE_API_KEY:+azure}
     provider_type: remote::azure
     config:
       api_key: ${env.AZURE_API_KEY:=}
-      api_base: ${env.AZURE_API_BASE:=}
+      base_url: ${env.AZURE_API_BASE:=}
       api_version: ${env.AZURE_API_VERSION:=}
       api_type: ${env.AZURE_API_TYPE:=}
   - provider_id: sentence-transformers

src/llama_stack/distributions/nvidia/run-with-safety.yaml

@@ -16,9 +16,8 @@ providers:
   - provider_id: nvidia
     provider_type: remote::nvidia
     config:
-      url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com}
+      base_url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com/v1}
       api_key: ${env.NVIDIA_API_KEY:=}
-      append_api_version: ${env.NVIDIA_APPEND_API_VERSION:=True}
   - provider_id: nvidia
     provider_type: remote::nvidia
     config:

src/llama_stack/distributions/nvidia/run.yaml

@@ -16,9 +16,8 @@ providers:
   - provider_id: nvidia
     provider_type: remote::nvidia
     config:
-      url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com}
+      base_url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com/v1}
       api_key: ${env.NVIDIA_API_KEY:=}
-      append_api_version: ${env.NVIDIA_APPEND_API_VERSION:=True}
   vector_io:
   - provider_id: faiss
     provider_type: inline::faiss

src/llama_stack/distributions/open-benchmark/run.yaml

@@ -27,12 +27,12 @@ providers:
   - provider_id: groq
     provider_type: remote::groq
     config:
-      url: https://api.groq.com
+      base_url: https://api.groq.com/openai/v1
       api_key: ${env.GROQ_API_KEY:=}
   - provider_id: together
     provider_type: remote::together
     config:
-      url: https://api.together.xyz/v1
+      base_url: https://api.together.xyz/v1
       api_key: ${env.TOGETHER_API_KEY:=}
   vector_io:
   - provider_id: sqlite-vec

src/llama_stack/distributions/postgres-demo/run.yaml

@@ -11,7 +11,7 @@ providers:
   - provider_id: vllm-inference
     provider_type: remote::vllm
     config:
-      url: ${env.VLLM_URL:=http://localhost:8000/v1}
+      base_url: ${env.VLLM_URL:=}
       max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
       api_token: ${env.VLLM_API_TOKEN:=fake}
       tls_verify: ${env.VLLM_TLS_VERIFY:=true}

src/llama_stack/distributions/starter-gpu/run-with-postgres-store.yaml

@@ -17,32 +17,32 @@ providers:
   - provider_id: ${env.CEREBRAS_API_KEY:+cerebras}
     provider_type: remote::cerebras
     config:
-      base_url: https://api.cerebras.ai
+      base_url: https://api.cerebras.ai/v1
       api_key: ${env.CEREBRAS_API_KEY:=}
   - provider_id: ${env.OLLAMA_URL:+ollama}
     provider_type: remote::ollama
     config:
-      url: ${env.OLLAMA_URL:=http://localhost:11434}
+      base_url: ${env.OLLAMA_URL:=http://localhost:11434/v1}
   - provider_id: ${env.VLLM_URL:+vllm}
     provider_type: remote::vllm
     config:
-      url: ${env.VLLM_URL:=}
+      base_url: ${env.VLLM_URL:=}
       max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
       api_token: ${env.VLLM_API_TOKEN:=fake}
       tls_verify: ${env.VLLM_TLS_VERIFY:=true}
   - provider_id: ${env.TGI_URL:+tgi}
     provider_type: remote::tgi
     config:
-      url: ${env.TGI_URL:=}
+      base_url: ${env.TGI_URL:=}
   - provider_id: fireworks
     provider_type: remote::fireworks
     config:
-      url: https://api.fireworks.ai/inference/v1
+      base_url: https://api.fireworks.ai/inference/v1
       api_key: ${env.FIREWORKS_API_KEY:=}
   - provider_id: together
     provider_type: remote::together
     config:
-      url: https://api.together.xyz/v1
+      base_url: https://api.together.xyz/v1
       api_key: ${env.TOGETHER_API_KEY:=}
   - provider_id: bedrock
     provider_type: remote::bedrock
@@ -52,9 +52,8 @@ providers:
   - provider_id: ${env.NVIDIA_API_KEY:+nvidia}
     provider_type: remote::nvidia
     config:
-      url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com}
+      base_url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com/v1}
       api_key: ${env.NVIDIA_API_KEY:=}
-      append_api_version: ${env.NVIDIA_APPEND_API_VERSION:=True}
   - provider_id: openai
     provider_type: remote::openai
     config:
@@ -76,18 +75,18 @@ providers:
   - provider_id: groq
     provider_type: remote::groq
     config:
-      url: https://api.groq.com
+      base_url: https://api.groq.com/openai/v1
       api_key: ${env.GROQ_API_KEY:=}
   - provider_id: sambanova
     provider_type: remote::sambanova
     config:
-      url: https://api.sambanova.ai/v1
+      base_url: https://api.sambanova.ai/v1
       api_key: ${env.SAMBANOVA_API_KEY:=}
   - provider_id: ${env.AZURE_API_KEY:+azure}
     provider_type: remote::azure
     config:
       api_key: ${env.AZURE_API_KEY:=}
-      api_base: ${env.AZURE_API_BASE:=}
+      base_url: ${env.AZURE_API_BASE:=}
       api_version: ${env.AZURE_API_VERSION:=}
       api_type: ${env.AZURE_API_TYPE:=}
   - provider_id: sentence-transformers

src/llama_stack/distributions/starter-gpu/run.yaml

@@ -17,32 +17,32 @@ providers:
   - provider_id: ${env.CEREBRAS_API_KEY:+cerebras}
     provider_type: remote::cerebras
     config:
-      base_url: https://api.cerebras.ai
+      base_url: https://api.cerebras.ai/v1
       api_key: ${env.CEREBRAS_API_KEY:=}
   - provider_id: ${env.OLLAMA_URL:+ollama}
     provider_type: remote::ollama
     config:
-      url: ${env.OLLAMA_URL:=http://localhost:11434}
+      base_url: ${env.OLLAMA_URL:=http://localhost:11434/v1}
   - provider_id: ${env.VLLM_URL:+vllm}
     provider_type: remote::vllm
     config:
-      url: ${env.VLLM_URL:=}
+      base_url: ${env.VLLM_URL:=}
       max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
       api_token: ${env.VLLM_API_TOKEN:=fake}
       tls_verify: ${env.VLLM_TLS_VERIFY:=true}
   - provider_id: ${env.TGI_URL:+tgi}
     provider_type: remote::tgi
     config:
-      url: ${env.TGI_URL:=}
+      base_url: ${env.TGI_URL:=}
   - provider_id: fireworks
     provider_type: remote::fireworks
     config:
-      url: https://api.fireworks.ai/inference/v1
+      base_url: https://api.fireworks.ai/inference/v1
       api_key: ${env.FIREWORKS_API_KEY:=}
   - provider_id: together
     provider_type: remote::together
     config:
-      url: https://api.together.xyz/v1
+      base_url: https://api.together.xyz/v1
       api_key: ${env.TOGETHER_API_KEY:=}
   - provider_id: bedrock
     provider_type: remote::bedrock
@@ -52,9 +52,8 @@ providers:
   - provider_id: ${env.NVIDIA_API_KEY:+nvidia}
     provider_type: remote::nvidia
     config:
-      url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com}
+      base_url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com/v1}
       api_key: ${env.NVIDIA_API_KEY:=}
-      append_api_version: ${env.NVIDIA_APPEND_API_VERSION:=True}
   - provider_id: openai
     provider_type: remote::openai
     config:
@@ -76,18 +75,18 @@ providers:
   - provider_id: groq
     provider_type: remote::groq
     config:
-      url: https://api.groq.com
+      base_url: https://api.groq.com/openai/v1
       api_key: ${env.GROQ_API_KEY:=}
   - provider_id: sambanova
     provider_type: remote::sambanova
     config:
-      url: https://api.sambanova.ai/v1
+      base_url: https://api.sambanova.ai/v1
       api_key: ${env.SAMBANOVA_API_KEY:=}
   - provider_id: ${env.AZURE_API_KEY:+azure}
     provider_type: remote::azure
     config:
       api_key: ${env.AZURE_API_KEY:=}
-      api_base: ${env.AZURE_API_BASE:=}
+      base_url: ${env.AZURE_API_BASE:=}
       api_version: ${env.AZURE_API_VERSION:=}
       api_type: ${env.AZURE_API_TYPE:=}
   - provider_id: sentence-transformers

src/llama_stack/distributions/starter/run-with-postgres-store.yaml

@@ -17,32 +17,32 @@ providers:
   - provider_id: ${env.CEREBRAS_API_KEY:+cerebras}
     provider_type: remote::cerebras
     config:
-      base_url: https://api.cerebras.ai
+      base_url: https://api.cerebras.ai/v1
       api_key: ${env.CEREBRAS_API_KEY:=}
   - provider_id: ${env.OLLAMA_URL:+ollama}
     provider_type: remote::ollama
     config:
-      url: ${env.OLLAMA_URL:=http://localhost:11434}
+      base_url: ${env.OLLAMA_URL:=http://localhost:11434/v1}
   - provider_id: ${env.VLLM_URL:+vllm}
     provider_type: remote::vllm
     config:
-      url: ${env.VLLM_URL:=}
+      base_url: ${env.VLLM_URL:=}
       max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
       api_token: ${env.VLLM_API_TOKEN:=fake}
       tls_verify: ${env.VLLM_TLS_VERIFY:=true}
   - provider_id: ${env.TGI_URL:+tgi}
     provider_type: remote::tgi
     config:
-      url: ${env.TGI_URL:=}
+      base_url: ${env.TGI_URL:=}
   - provider_id: fireworks
     provider_type: remote::fireworks
     config:
-      url: https://api.fireworks.ai/inference/v1
+      base_url: https://api.fireworks.ai/inference/v1
       api_key: ${env.FIREWORKS_API_KEY:=}
   - provider_id: together
     provider_type: remote::together
     config:
-      url: https://api.together.xyz/v1
+      base_url: https://api.together.xyz/v1
       api_key: ${env.TOGETHER_API_KEY:=}
   - provider_id: bedrock
     provider_type: remote::bedrock
@@ -52,9 +52,8 @@ providers:
   - provider_id: ${env.NVIDIA_API_KEY:+nvidia}
     provider_type: remote::nvidia
     config:
-      url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com}
+      base_url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com/v1}
       api_key: ${env.NVIDIA_API_KEY:=}
-      append_api_version: ${env.NVIDIA_APPEND_API_VERSION:=True}
   - provider_id: openai
     provider_type: remote::openai
     config:
@@ -76,18 +75,18 @@ providers:
   - provider_id: groq
     provider_type: remote::groq
     config:
-      url: https://api.groq.com
+      base_url: https://api.groq.com/openai/v1
       api_key: ${env.GROQ_API_KEY:=}
   - provider_id: sambanova
     provider_type: remote::sambanova
     config:
-      url: https://api.sambanova.ai/v1
+      base_url: https://api.sambanova.ai/v1
       api_key: ${env.SAMBANOVA_API_KEY:=}
   - provider_id: ${env.AZURE_API_KEY:+azure}
     provider_type: remote::azure
     config:
       api_key: ${env.AZURE_API_KEY:=}
-      api_base: ${env.AZURE_API_BASE:=}
+      base_url: ${env.AZURE_API_BASE:=}
       api_version: ${env.AZURE_API_VERSION:=}
       api_type: ${env.AZURE_API_TYPE:=}
   - provider_id: sentence-transformers

src/llama_stack/distributions/starter/run.yaml

@@ -17,32 +17,32 @@ providers:
   - provider_id: ${env.CEREBRAS_API_KEY:+cerebras}
     provider_type: remote::cerebras
     config:
-      base_url: https://api.cerebras.ai
+      base_url: https://api.cerebras.ai/v1
       api_key: ${env.CEREBRAS_API_KEY:=}
   - provider_id: ${env.OLLAMA_URL:+ollama}
     provider_type: remote::ollama
     config:
-      url: ${env.OLLAMA_URL:=http://localhost:11434}
+      base_url: ${env.OLLAMA_URL:=http://localhost:11434/v1}
   - provider_id: ${env.VLLM_URL:+vllm}
     provider_type: remote::vllm
     config:
-      url: ${env.VLLM_URL:=}
+      base_url: ${env.VLLM_URL:=}
       max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
       api_token: ${env.VLLM_API_TOKEN:=fake}
       tls_verify: ${env.VLLM_TLS_VERIFY:=true}
   - provider_id: ${env.TGI_URL:+tgi}
     provider_type: remote::tgi
     config:
-      url: ${env.TGI_URL:=}
+      base_url: ${env.TGI_URL:=}
   - provider_id: fireworks
     provider_type: remote::fireworks
     config:
-      url: https://api.fireworks.ai/inference/v1
+      base_url: https://api.fireworks.ai/inference/v1
       api_key: ${env.FIREWORKS_API_KEY:=}
   - provider_id: together
     provider_type: remote::together
     config:
-      url: https://api.together.xyz/v1
+      base_url: https://api.together.xyz/v1
       api_key: ${env.TOGETHER_API_KEY:=}
   - provider_id: bedrock
     provider_type: remote::bedrock
@@ -52,9 +52,8 @@ providers:
   - provider_id: ${env.NVIDIA_API_KEY:+nvidia}
     provider_type: remote::nvidia
     config:
-      url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com}
+      base_url: ${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com/v1}
       api_key: ${env.NVIDIA_API_KEY:=}
-      append_api_version: ${env.NVIDIA_APPEND_API_VERSION:=True}
   - provider_id: openai
     provider_type: remote::openai
     config:
@@ -76,18 +75,18 @@ providers:
   - provider_id: groq
     provider_type: remote::groq
     config:
-      url: https://api.groq.com
+      base_url: https://api.groq.com/openai/v1
       api_key: ${env.GROQ_API_KEY:=}
   - provider_id: sambanova
     provider_type: remote::sambanova
     config:
-      url: https://api.sambanova.ai/v1
+      base_url: https://api.sambanova.ai/v1
       api_key: ${env.SAMBANOVA_API_KEY:=}
   - provider_id: ${env.AZURE_API_KEY:+azure}
     provider_type: remote::azure
     config:
       api_key: ${env.AZURE_API_KEY:=}
-      api_base: ${env.AZURE_API_BASE:=}
+      base_url: ${env.AZURE_API_BASE:=}
       api_version: ${env.AZURE_API_VERSION:=}
       api_type: ${env.AZURE_API_TYPE:=}
   - provider_id: sentence-transformers

src/llama_stack/distributions/watsonx/run.yaml

@@ -15,7 +15,7 @@ providers:
   - provider_id: watsonx
     provider_type: remote::watsonx
     config:
-      url: ${env.WATSONX_BASE_URL:=https://us-south.ml.cloud.ibm.com}
+      base_url: ${env.WATSONX_BASE_URL:=https://us-south.ml.cloud.ibm.com}
       api_key: ${env.WATSONX_API_KEY:=}
       project_id: ${env.WATSONX_PROJECT_ID:=}
   vector_io:

src/llama_stack/providers/inline/agents/meta_reference/__init__.py

@@ -23,12 +23,14 @@ async def get_provider_impl(
         config,
         deps[Api.inference],
         deps[Api.vector_io],
-        deps[Api.safety],
+        deps.get(Api.safety),
         deps[Api.tool_runtime],
         deps[Api.tool_groups],
         deps[Api.conversations],
-        policy,
+        deps[Api.prompts],
+        deps[Api.files],
         telemetry_enabled,
+        policy,
     )
     await impl.initialize()
     return impl

src/llama_stack/providers/inline/agents/meta_reference/agents.py

@@ -12,6 +12,7 @@ from llama_stack.providers.utils.responses.responses_store import ResponsesStore
 from llama_stack_api import (
     Agents,
     Conversations,
+    Files,
     Inference,
     ListOpenAIResponseInputItem,
     ListOpenAIResponseObject,
@@ -22,6 +23,7 @@ from llama_stack_api import (
     OpenAIResponsePrompt,
     OpenAIResponseText,
     Order,
+    Prompts,
     ResponseGuardrail,
     Safety,
     ToolGroups,
@@ -41,10 +43,12 @@ class MetaReferenceAgentsImpl(Agents):
         config: MetaReferenceAgentsImplConfig,
         inference_api: Inference,
         vector_io_api: VectorIO,
-        safety_api: Safety,
+        safety_api: Safety | None,
         tool_runtime_api: ToolRuntime,
         tool_groups_api: ToolGroups,
         conversations_api: Conversations,
+        prompts_api: Prompts,
+        files_api: Files,
         policy: list[AccessRule],
         telemetry_enabled: bool = False,
     ):
@@ -56,7 +60,8 @@ class MetaReferenceAgentsImpl(Agents):
         self.tool_groups_api = tool_groups_api
         self.conversations_api = conversations_api
         self.telemetry_enabled = telemetry_enabled
-
+        self.prompts_api = prompts_api
+        self.files_api = files_api
         self.in_memory_store = InmemoryKVStoreImpl()
         self.openai_responses_impl: OpenAIResponsesImpl | None = None
         self.policy = policy
@@ -73,6 +78,8 @@ class MetaReferenceAgentsImpl(Agents):
             vector_io_api=self.vector_io_api,
             safety_api=self.safety_api,
             conversations_api=self.conversations_api,
+            prompts_api=self.prompts_api,
+            files_api=self.files_api,
         )
 
     async def shutdown(self) -> None:

src/llama_stack/providers/inline/agents/meta_reference/responses/openai_responses.py

@@ -4,6 +4,7 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
+import re
 import time
 import uuid
 from collections.abc import AsyncIterator
@@ -18,13 +19,17 @@ from llama_stack.providers.utils.responses.responses_store import (
 from llama_stack_api import (
     ConversationItem,
     Conversations,
+    Files,
     Inference,
     InvalidConversationIdError,
     ListOpenAIResponseInputItem,
     ListOpenAIResponseObject,
+    OpenAIChatCompletionContentPartParam,
     OpenAIDeleteResponseObject,
     OpenAIMessageParam,
     OpenAIResponseInput,
+    OpenAIResponseInputMessageContentFile,
+    OpenAIResponseInputMessageContentImage,
     OpenAIResponseInputMessageContentText,
     OpenAIResponseInputTool,
     OpenAIResponseMessage,
@@ -34,7 +39,9 @@ from llama_stack_api import (
     OpenAIResponseText,
     OpenAIResponseTextFormat,
     OpenAISystemMessageParam,
+    OpenAIUserMessageParam,
     Order,
+    Prompts,
     ResponseGuardrailSpec,
     Safety,
     ToolGroups,
@@ -46,6 +53,7 @@ from .streaming import StreamingResponseOrchestrator
 from .tool_executor import ToolExecutor
 from .types import ChatCompletionContext, ToolContext
 from .utils import (
+    convert_response_content_to_chat_content,
     convert_response_input_to_chat_messages,
     convert_response_text_to_chat_response_format,
     extract_guardrail_ids,
@@ -67,8 +75,10 @@ class OpenAIResponsesImpl:
         tool_runtime_api: ToolRuntime,
         responses_store: ResponsesStore,
         vector_io_api: VectorIO,  # VectorIO
-        safety_api: Safety,
+        safety_api: Safety | None,
         conversations_api: Conversations,
+        prompts_api: Prompts,
+        files_api: Files,
     ):
         self.inference_api = inference_api
         self.tool_groups_api = tool_groups_api
@@ -82,6 +92,8 @@ class OpenAIResponsesImpl:
             tool_runtime_api=tool_runtime_api,
             vector_io_api=vector_io_api,
         )
+        self.prompts_api = prompts_api
+        self.files_api = files_api
 
     async def _prepend_previous_response(
         self,
@@ -122,11 +134,13 @@ class OpenAIResponsesImpl:
                 # Use stored messages directly and convert only new input
                 message_adapter = TypeAdapter(list[OpenAIMessageParam])
                 messages = message_adapter.validate_python(previous_response.messages)
-                new_messages = await convert_response_input_to_chat_messages(input, previous_messages=messages)
+                new_messages = await convert_response_input_to_chat_messages(
+                    input, previous_messages=messages, files_api=self.files_api
+                )
                 messages.extend(new_messages)
             else:
                 # Backward compatibility: reconstruct from inputs
-                messages = await convert_response_input_to_chat_messages(all_input)
+                messages = await convert_response_input_to_chat_messages(all_input, files_api=self.files_api)
 
             tool_context.recover_tools_from_previous_response(previous_response)
         elif conversation is not None:
@@ -138,7 +152,7 @@ class OpenAIResponsesImpl:
             all_input = input
             if not conversation_items.data:
                 # First turn - just convert the new input
-                messages = await convert_response_input_to_chat_messages(input)
+                messages = await convert_response_input_to_chat_messages(input, files_api=self.files_api)
             else:
                 if not stored_messages:
                     all_input = conversation_items.data
@@ -154,14 +168,82 @@ class OpenAIResponsesImpl:
                     all_input = input
 
                 messages = stored_messages or []
-                new_messages = await convert_response_input_to_chat_messages(all_input, previous_messages=messages)
+                new_messages = await convert_response_input_to_chat_messages(
+                    all_input, previous_messages=messages, files_api=self.files_api
+                )
                 messages.extend(new_messages)
         else:
             all_input = input
-            messages = await convert_response_input_to_chat_messages(all_input)
+            messages = await convert_response_input_to_chat_messages(all_input, files_api=self.files_api)
 
         return all_input, messages, tool_context
 
+    async def _prepend_prompt(
+        self,
+        messages: list[OpenAIMessageParam],
+        openai_response_prompt: OpenAIResponsePrompt | None,
+    ) -> None:
+        """Prepend prompt template to messages, resolving text/image/file variables.
+
+        :param messages: List of OpenAIMessageParam objects
+        :param openai_response_prompt: (Optional) OpenAIResponsePrompt object with variables
+        :returns: string of utf-8 characters
+        """
+        if not openai_response_prompt or not openai_response_prompt.id:
+            return
+
+        prompt_version = int(openai_response_prompt.version) if openai_response_prompt.version else None
+        cur_prompt = await self.prompts_api.get_prompt(openai_response_prompt.id, prompt_version)
+
+        if not cur_prompt or not cur_prompt.prompt:
+            return
+
+        cur_prompt_text = cur_prompt.prompt
+        cur_prompt_variables = cur_prompt.variables
+
+        if not openai_response_prompt.variables:
+            messages.insert(0, OpenAISystemMessageParam(content=cur_prompt_text))
+            return
+
+        # Validate that all provided variables exist in the prompt
+        for name in openai_response_prompt.variables.keys():
+            if name not in cur_prompt_variables:
+                raise ValueError(f"Variable {name} not found in prompt {openai_response_prompt.id}")
+
+        # Separate text and media variables
+        text_substitutions = {}
+        media_content_parts: list[OpenAIChatCompletionContentPartParam] = []
+
+        for name, value in openai_response_prompt.variables.items():
+            # Text variable found
+            if isinstance(value, OpenAIResponseInputMessageContentText):
+                text_substitutions[name] = value.text
+
+            # Media variable found
+            elif isinstance(value, OpenAIResponseInputMessageContentImage | OpenAIResponseInputMessageContentFile):
+                converted_parts = await convert_response_content_to_chat_content([value], files_api=self.files_api)
+                if isinstance(converted_parts, list):
+                    media_content_parts.extend(converted_parts)
+
+                # Eg: {{product_photo}} becomes "[Image: product_photo]"
+                # This gives the model textual context about what media exists in the prompt
+                var_type = value.type.replace("input_", "").replace("_", " ").title()
+                text_substitutions[name] = f"[{var_type}: {name}]"
+
+        def replace_variable(match: re.Match[str]) -> str:
+            var_name = match.group(1).strip()
+            return str(text_substitutions.get(var_name, match.group(0)))
+
+        pattern = r"\{\{\s*(\w+)\s*\}\}"
+        processed_prompt_text = re.sub(pattern, replace_variable, cur_prompt_text)
+
+        # Insert system message with resolved text
+        messages.insert(0, OpenAISystemMessageParam(content=processed_prompt_text))
+
+        # If we have media, create a new user message because allows to ingest images and files
+        if media_content_parts:
+            messages.append(OpenAIUserMessageParam(content=media_content_parts))
+
     async def get_openai_response(
         self,
         response_id: str,
@@ -273,6 +355,14 @@ class OpenAIResponsesImpl:
 
         guardrail_ids = extract_guardrail_ids(guardrails) if guardrails else []
 
+        # Validate that Safety API is available if guardrails are requested
+        if guardrail_ids and self.safety_api is None:
+            raise ValueError(
+                "Cannot process guardrails: Safety API is not configured.\n\n"
+                "To use guardrails, ensure the Safety API is configured in your stack, or remove "
+                "the 'guardrails' parameter from your request."
+            )
+
         if conversation is not None:
             if previous_response_id is not None:
                 raise ValueError(
@@ -289,6 +379,7 @@ class OpenAIResponsesImpl:
             input=input,
             conversation=conversation,
             model=model,
+            prompt=prompt,
             instructions=instructions,
             previous_response_id=previous_response_id,
             store=store,
@@ -342,6 +433,7 @@ class OpenAIResponsesImpl:
         instructions: str | None = None,
         previous_response_id: str | None = None,
         conversation: str | None = None,
+        prompt: OpenAIResponsePrompt | None = None,
         store: bool | None = True,
         temperature: float | None = None,
         text: OpenAIResponseText | None = None,
@@ -364,6 +456,9 @@ class OpenAIResponsesImpl:
         if instructions:
             messages.insert(0, OpenAISystemMessageParam(content=instructions))
 
+        # Prepend reusable prompt (if provided)
+        await self._prepend_prompt(messages, prompt)
+
         # Structured outputs
         response_format = await convert_response_text_to_chat_response_format(text)
 
@@ -386,6 +481,7 @@ class OpenAIResponsesImpl:
             ctx=ctx,
             response_id=response_id,
             created_at=created_at,
+            prompt=prompt,
             text=text,
             max_infer_iters=max_infer_iters,
             parallel_tool_calls=parallel_tool_calls,

src/llama_stack/providers/inline/agents/meta_reference/responses/streaming.py

@@ -66,6 +66,8 @@ from llama_stack_api import (
     OpenAIResponseUsage,
     OpenAIResponseUsageInputTokensDetails,
     OpenAIResponseUsageOutputTokensDetails,
+    OpenAIToolMessageParam,
+    Safety,
     WebSearchToolTypes,
 )
 
@@ -111,7 +113,7 @@ class StreamingResponseOrchestrator:
         max_infer_iters: int,
         tool_executor,  # Will be the tool execution logic from the main class
         instructions: str | None,
-        safety_api,
+        safety_api: Safety | None,
         guardrail_ids: list[str] | None = None,
         prompt: OpenAIResponsePrompt | None = None,
         parallel_tool_calls: bool | None = None,
@@ -905,10 +907,16 @@ class StreamingResponseOrchestrator:
         """Coordinate execution of both function and non-function tool calls."""
         # Execute non-function tool calls
         for tool_call in non_function_tool_calls:
-            # Check if total calls made to built-in and mcp tools exceed max_tool_calls
+            # if total calls made to built-in and mcp tools exceed max_tool_calls
+            # then create a tool response message indicating the call was skipped
             if self.max_tool_calls is not None and self.accumulated_builtin_tool_calls >= self.max_tool_calls:
                 logger.info(f"Ignoring built-in and mcp tool call since reached the limit of {self.max_tool_calls=}.")
-                break
+                skipped_call_message = OpenAIToolMessageParam(
+                    content=f"Tool call skipped: maximum tool calls limit ({self.max_tool_calls}) reached.",
+                    tool_call_id=tool_call.id,
+                )
+                next_turn_messages.append(skipped_call_message)
+                continue
 
             # Find the item_id for this tool call
             matching_item_id = None

src/llama_stack/providers/inline/agents/meta_reference/responses/utils.py

@@ -5,11 +5,14 @@
 # the root directory of this source tree.
 
 import asyncio
+import base64
+import mimetypes
 import re
 import uuid
 from collections.abc import Sequence
 
 from llama_stack_api import (
+    Files,
     OpenAIAssistantMessageParam,
     OpenAIChatCompletionContentPartImageParam,
     OpenAIChatCompletionContentPartParam,
@@ -18,6 +21,8 @@ from llama_stack_api import (
     OpenAIChatCompletionToolCallFunction,
     OpenAIChoice,
     OpenAIDeveloperMessageParam,
+    OpenAIFile,
+    OpenAIFileFile,
     OpenAIImageURL,
     OpenAIJSONSchema,
     OpenAIMessageParam,
@@ -29,6 +34,7 @@ from llama_stack_api import (
     OpenAIResponseInput,
     OpenAIResponseInputFunctionToolCallOutput,
     OpenAIResponseInputMessageContent,
+    OpenAIResponseInputMessageContentFile,
     OpenAIResponseInputMessageContentImage,
     OpenAIResponseInputMessageContentText,
     OpenAIResponseInputTool,
@@ -37,9 +43,11 @@ from llama_stack_api import (
     OpenAIResponseMessage,
     OpenAIResponseOutputMessageContent,
     OpenAIResponseOutputMessageContentOutputText,
+    OpenAIResponseOutputMessageFileSearchToolCall,
     OpenAIResponseOutputMessageFunctionToolCall,
     OpenAIResponseOutputMessageMCPCall,
     OpenAIResponseOutputMessageMCPListTools,
+    OpenAIResponseOutputMessageWebSearchToolCall,
     OpenAIResponseText,
     OpenAISystemMessageParam,
     OpenAIToolMessageParam,
@@ -49,6 +57,46 @@ from llama_stack_api import (
 )
 
 
+async def extract_bytes_from_file(file_id: str, files_api: Files) -> bytes:
+    """
+    Extract raw bytes from file using the Files API.
+
+    :param file_id: The file identifier (e.g., "file-abc123")
+    :param files_api: Files API instance
+    :returns: Raw file content as bytes
+    :raises: ValueError if file cannot be retrieved
+    """
+    try:
+        response = await files_api.openai_retrieve_file_content(file_id)
+        return bytes(response.body)
+    except Exception as e:
+        raise ValueError(f"Failed to retrieve file content for file_id '{file_id}': {str(e)}") from e
+
+
+def generate_base64_ascii_text_from_bytes(raw_bytes: bytes) -> str:
+    """
+    Converts raw binary bytes into a safe ASCII text representation for URLs
+
+    :param raw_bytes: the actual bytes that represents file content
+    :returns: string of utf-8 characters
+    """
+    return base64.b64encode(raw_bytes).decode("utf-8")
+
+
+def construct_data_url(ascii_text: str, mime_type: str | None) -> str:
+    """
+    Construct data url with decoded data inside
+
+    :param ascii_text: ASCII content
+    :param mime_type: MIME type of file
+    :returns: data url string (eg. data:image/png,base64,%3Ch1%3EHello%2C%20World%21%3C%2Fh1%3E)
+    """
+    if not mime_type:
+        mime_type = "application/octet-stream"
+
+    return f"data:{mime_type};base64,{ascii_text}"
+
+
 async def convert_chat_choice_to_response_message(
     choice: OpenAIChoice,
     citation_files: dict[str, str] | None = None,
@@ -78,11 +126,15 @@ async def convert_chat_choice_to_response_message(
 
 async def convert_response_content_to_chat_content(
     content: str | Sequence[OpenAIResponseInputMessageContent | OpenAIResponseOutputMessageContent],
+    files_api: Files | None,
 ) -> str | list[OpenAIChatCompletionContentPartParam]:
     """
     Convert the content parts from an OpenAI Response API request into OpenAI Chat Completion content parts.
 
     The content schemas of each API look similar, but are not exactly the same.
+
+    :param content: The content to convert
+    :param files_api: Files API for resolving file_id to raw file content (required if content contains files/images)
     """
     if isinstance(content, str):
         return content
@@ -95,9 +147,68 @@ async def convert_response_content_to_chat_content(
         elif isinstance(content_part, OpenAIResponseOutputMessageContentOutputText):
             converted_parts.append(OpenAIChatCompletionContentPartTextParam(text=content_part.text))
         elif isinstance(content_part, OpenAIResponseInputMessageContentImage):
+            detail = content_part.detail
+            image_mime_type = None
             if content_part.image_url:
-                image_url = OpenAIImageURL(url=content_part.image_url, detail=content_part.detail)
+                image_url = OpenAIImageURL(url=content_part.image_url, detail=detail)
                 converted_parts.append(OpenAIChatCompletionContentPartImageParam(image_url=image_url))
+            elif content_part.file_id:
+                if files_api is None:
+                    raise ValueError("file_ids are not supported by this implementation of the Stack")
+                image_file_response = await files_api.openai_retrieve_file(content_part.file_id)
+                if image_file_response.filename:
+                    image_mime_type, _ = mimetypes.guess_type(image_file_response.filename)
+                raw_image_bytes = await extract_bytes_from_file(content_part.file_id, files_api)
+                ascii_text = generate_base64_ascii_text_from_bytes(raw_image_bytes)
+                image_data_url = construct_data_url(ascii_text, image_mime_type)
+                image_url = OpenAIImageURL(url=image_data_url, detail=detail)
+                converted_parts.append(OpenAIChatCompletionContentPartImageParam(image_url=image_url))
+            else:
+                raise ValueError(
+                    f"Image content must have either 'image_url' or 'file_id'. "
+                    f"Got image_url={content_part.image_url}, file_id={content_part.file_id}"
+                )
+        elif isinstance(content_part, OpenAIResponseInputMessageContentFile):
+            resolved_file_data = None
+            file_data = content_part.file_data
+            file_id = content_part.file_id
+            file_url = content_part.file_url
+            filename = content_part.filename
+            file_mime_type = None
+            if not any([file_data, file_id, file_url]):
+                raise ValueError(
+                    f"File content must have at least one of 'file_data', 'file_id', or 'file_url'. "
+                    f"Got file_data={file_data}, file_id={file_id}, file_url={file_url}"
+                )
+            if file_id:
+                if files_api is None:
+                    raise ValueError("file_ids are not supported by this implementation of the Stack")
+
+                file_response = await files_api.openai_retrieve_file(file_id)
+                if not filename:
+                    filename = file_response.filename
+                file_mime_type, _ = mimetypes.guess_type(file_response.filename)
+                raw_file_bytes = await extract_bytes_from_file(file_id, files_api)
+                ascii_text = generate_base64_ascii_text_from_bytes(raw_file_bytes)
+                resolved_file_data = construct_data_url(ascii_text, file_mime_type)
+            elif file_data:
+                if file_data.startswith("data:"):
+                    resolved_file_data = file_data
+                else:
+                    # Raw base64 data, wrap in data URL format
+                    if filename:
+                        file_mime_type, _ = mimetypes.guess_type(filename)
+                    resolved_file_data = construct_data_url(file_data, file_mime_type)
+            elif file_url:
+                resolved_file_data = file_url
+            converted_parts.append(
+                OpenAIFile(
+                    file=OpenAIFileFile(
+                        file_data=resolved_file_data,
+                        filename=filename,
+                    )
+                )
+            )
         elif isinstance(content_part, str):
             converted_parts.append(OpenAIChatCompletionContentPartTextParam(text=content_part))
         else:
@@ -110,12 +221,14 @@ async def convert_response_content_to_chat_content(
 async def convert_response_input_to_chat_messages(
     input: str | list[OpenAIResponseInput],
     previous_messages: list[OpenAIMessageParam] | None = None,
+    files_api: Files | None = None,
 ) -> list[OpenAIMessageParam]:
     """
     Convert the input from an OpenAI Response API request into OpenAI Chat Completion messages.
 
     :param input: The input to convert
     :param previous_messages: Optional previous messages to check for function_call references
+    :param files_api: Files API for resolving file_id to raw file content (optional, required for file/image content)
     """
     messages: list[OpenAIMessageParam] = []
     if isinstance(input, list):
@@ -169,6 +282,12 @@ async def convert_response_input_to_chat_messages(
             elif isinstance(input_item, OpenAIResponseOutputMessageMCPListTools):
                 # the tool list will be handled separately
                 pass
+            elif isinstance(
+                input_item,
+                OpenAIResponseOutputMessageWebSearchToolCall | OpenAIResponseOutputMessageFileSearchToolCall,
+            ):
+                # these tool calls are tracked internally but not converted to chat messages
+                pass
             elif isinstance(input_item, OpenAIResponseMCPApprovalRequest) or isinstance(
                 input_item, OpenAIResponseMCPApprovalResponse
             ):
@@ -176,7 +295,7 @@ async def convert_response_input_to_chat_messages(
                 pass
             elif isinstance(input_item, OpenAIResponseMessage):
                 # Narrow type to OpenAIResponseMessage which has content and role attributes
-                content = await convert_response_content_to_chat_content(input_item.content)
+                content = await convert_response_content_to_chat_content(input_item.content, files_api)
                 message_type = await get_message_type_by_role(input_item.role)
                 if message_type is None:
                     raise ValueError(
@@ -320,11 +439,15 @@ def is_function_tool_call(
     return False
 
 
-async def run_guardrails(safety_api: Safety, messages: str, guardrail_ids: list[str]) -> str | None:
+async def run_guardrails(safety_api: Safety | None, messages: str, guardrail_ids: list[str]) -> str | None:
     """Run guardrails against messages and return violation message if blocked."""
     if not messages:
         return None
 
+    # If safety API is not available, skip guardrails
+    if safety_api is None:
+        return None
+
     # Look up shields to get their provider_resource_id (actual model ID)
     model_ids = []
     # TODO: list_shields not in Safety interface but available at runtime via API routing

src/llama_stack/providers/registry/agents.py

@@ -30,11 +30,15 @@ def available_providers() -> list[ProviderSpec]:
             config_class="llama_stack.providers.inline.agents.meta_reference.MetaReferenceAgentsImplConfig",
             api_dependencies=[
                 Api.inference,
-                Api.safety,
                 Api.vector_io,
                 Api.tool_runtime,
                 Api.tool_groups,
                 Api.conversations,
+                Api.prompts,
+                Api.files,
+            ],
+            optional_api_dependencies=[
+                Api.safety,
             ],
             description="Meta's reference implementation of an agent system that can use tools, access vector databases, and perform complex reasoning tasks.",
         ),

src/llama_stack/providers/remote/inference/azure/azure.py

@@ -4,8 +4,6 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from urllib.parse import urljoin
-
 from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
 
 from .config import AzureConfig
@@ -22,4 +20,4 @@ class AzureInferenceAdapter(OpenAIMixin):
 
         Returns the Azure API base URL from the configuration.
         """
-        return urljoin(str(self.config.api_base), "/openai/v1")
+        return str(self.config.base_url)

src/llama_stack/providers/remote/inference/azure/config.py

@@ -32,8 +32,9 @@ class AzureProviderDataValidator(BaseModel):
 
 @json_schema_type
 class AzureConfig(RemoteInferenceProviderConfig):
-    api_base: HttpUrl = Field(
+    base_url: HttpUrl | None = Field(
-        description="Azure API base for Azure (e.g., https://your-resource-name.openai.azure.com)",
+        default=None,
+        description="Azure API base for Azure (e.g., https://your-resource-name.openai.azure.com/openai/v1)",
     )
     api_version: str | None = Field(
         default_factory=lambda: os.getenv("AZURE_API_VERSION"),
@@ -48,14 +49,14 @@ class AzureConfig(RemoteInferenceProviderConfig):
     def sample_run_config(
         cls,
         api_key: str = "${env.AZURE_API_KEY:=}",
-        api_base: str = "${env.AZURE_API_BASE:=}",
+        base_url: str = "${env.AZURE_API_BASE:=}",
         api_version: str = "${env.AZURE_API_VERSION:=}",
         api_type: str = "${env.AZURE_API_TYPE:=}",
         **kwargs,
     ) -> dict[str, Any]:
         return {
             "api_key": api_key,
-            "api_base": api_base,
+            "base_url": base_url,
             "api_version": api_version,
             "api_type": api_type,
         }

src/llama_stack/providers/remote/inference/cerebras/cerebras.py

@@ -4,8 +4,6 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from urllib.parse import urljoin
-
 from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
 from llama_stack_api import (
     OpenAIEmbeddingsRequestWithExtraBody,
@@ -21,7 +19,7 @@ class CerebrasInferenceAdapter(OpenAIMixin):
     provider_data_api_key_field: str = "cerebras_api_key"
 
     def get_base_url(self) -> str:
-        return urljoin(self.config.base_url, "v1")
+        return str(self.config.base_url)
 
     async def openai_embeddings(
         self,

src/llama_stack/providers/remote/inference/cerebras/config.py

@@ -7,12 +7,12 @@
 import os
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, HttpUrl
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
 
-DEFAULT_BASE_URL = "https://api.cerebras.ai"
+DEFAULT_BASE_URL = "https://api.cerebras.ai/v1"
 
 
 class CerebrasProviderDataValidator(BaseModel):
@@ -24,8 +24,8 @@ class CerebrasProviderDataValidator(BaseModel):
 
 @json_schema_type
 class CerebrasImplConfig(RemoteInferenceProviderConfig):
-    base_url: str = Field(
+    base_url: HttpUrl | None = Field(
-        default=os.environ.get("CEREBRAS_BASE_URL", DEFAULT_BASE_URL),
+        default=HttpUrl(os.environ.get("CEREBRAS_BASE_URL", DEFAULT_BASE_URL)),
         description="Base URL for the Cerebras API",
     )
 

src/llama_stack/providers/remote/inference/databricks/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field, SecretStr
+from pydantic import BaseModel, Field, HttpUrl, SecretStr
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -21,9 +21,9 @@ class DatabricksProviderDataValidator(BaseModel):
 
 @json_schema_type
 class DatabricksImplConfig(RemoteInferenceProviderConfig):
-    url: str | None = Field(
+    base_url: HttpUrl | None = Field(
         default=None,
-        description="The URL for the Databricks model serving endpoint",
+        description="The URL for the Databricks model serving endpoint (should include /serving-endpoints path)",
     )
     auth_credential: SecretStr | None = Field(
         default=None,
@@ -34,11 +34,11 @@ class DatabricksImplConfig(RemoteInferenceProviderConfig):
     @classmethod
     def sample_run_config(
         cls,
-        url: str = "${env.DATABRICKS_HOST:=}",
+        base_url: str = "${env.DATABRICKS_HOST:=}",
         api_token: str = "${env.DATABRICKS_TOKEN:=}",
         **kwargs: Any,
     ) -> dict[str, Any]:
         return {
-            "url": url,
+            "base_url": base_url,
             "api_token": api_token,
         }

src/llama_stack/providers/remote/inference/databricks/databricks.py

@@ -29,15 +29,21 @@ class DatabricksInferenceAdapter(OpenAIMixin):
     }
 
     def get_base_url(self) -> str:
-        return f"{self.config.url}/serving-endpoints"
+        return str(self.config.base_url)
 
     async def list_provider_model_ids(self) -> Iterable[str]:
         # Filter out None values from endpoint names
         api_token = self._get_api_key_from_config_or_provider_data()
+        # WorkspaceClient expects base host without /serving-endpoints suffix
+        base_url_str = str(self.config.base_url)
+        if base_url_str.endswith("/serving-endpoints"):
+            host = base_url_str[:-18]  # Remove '/serving-endpoints'
+        else:
+            host = base_url_str
         return [
             endpoint.name  # type: ignore[misc]
             for endpoint in WorkspaceClient(
-                host=self.config.url, token=api_token
+                host=host, token=api_token
             ).serving_endpoints.list()  # TODO: this is not async
         ]
 

src/llama_stack/providers/remote/inference/fireworks/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import Field
+from pydantic import Field, HttpUrl
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -14,14 +14,14 @@ from llama_stack_api import json_schema_type
 
 @json_schema_type
 class FireworksImplConfig(RemoteInferenceProviderConfig):
-    url: str = Field(
+    base_url: HttpUrl | None = Field(
-        default="https://api.fireworks.ai/inference/v1",
+        default=HttpUrl("https://api.fireworks.ai/inference/v1"),
         description="The URL for the Fireworks server",
     )
 
     @classmethod
     def sample_run_config(cls, api_key: str = "${env.FIREWORKS_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
-            "url": "https://api.fireworks.ai/inference/v1",
+            "base_url": "https://api.fireworks.ai/inference/v1",
             "api_key": api_key,
         }

src/llama_stack/providers/remote/inference/fireworks/fireworks.py

@@ -24,4 +24,4 @@ class FireworksInferenceAdapter(OpenAIMixin):
     provider_data_api_key_field: str = "fireworks_api_key"
 
     def get_base_url(self) -> str:
-        return "https://api.fireworks.ai/inference/v1"
+        return str(self.config.base_url)

src/llama_stack/providers/remote/inference/groq/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, HttpUrl
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -21,14 +21,14 @@ class GroqProviderDataValidator(BaseModel):
 
 @json_schema_type
 class GroqConfig(RemoteInferenceProviderConfig):
-    url: str = Field(
+    base_url: HttpUrl | None = Field(
-        default="https://api.groq.com",
+        default=HttpUrl("https://api.groq.com/openai/v1"),
         description="The URL for the Groq AI server",
     )
 
     @classmethod
     def sample_run_config(cls, api_key: str = "${env.GROQ_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
-            "url": "https://api.groq.com",
+            "base_url": "https://api.groq.com/openai/v1",
             "api_key": api_key,
         }

src/llama_stack/providers/remote/inference/groq/groq.py

@@ -15,4 +15,4 @@ class GroqInferenceAdapter(OpenAIMixin):
     provider_data_api_key_field: str = "groq_api_key"
 
     def get_base_url(self) -> str:
-        return f"{self.config.url}/openai/v1"
+        return str(self.config.base_url)

src/llama_stack/providers/remote/inference/llama_openai_compat/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, HttpUrl
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -21,14 +21,14 @@ class LlamaProviderDataValidator(BaseModel):
 
 @json_schema_type
 class LlamaCompatConfig(RemoteInferenceProviderConfig):
-    openai_compat_api_base: str = Field(
+    base_url: HttpUrl | None = Field(
-        default="https://api.llama.com/compat/v1/",
+        default=HttpUrl("https://api.llama.com/compat/v1/"),
         description="The URL for the Llama API server",
     )
 
     @classmethod
     def sample_run_config(cls, api_key: str = "${env.LLAMA_API_KEY}", **kwargs) -> dict[str, Any]:
         return {
-            "openai_compat_api_base": "https://api.llama.com/compat/v1/",
+            "base_url": "https://api.llama.com/compat/v1/",
             "api_key": api_key,
         }

src/llama_stack/providers/remote/inference/llama_openai_compat/llama.py

@@ -31,7 +31,7 @@ class LlamaCompatInferenceAdapter(OpenAIMixin):
 
         :return: The Llama API base URL
         """
-        return self.config.openai_compat_api_base
+        return str(self.config.base_url)
 
     async def openai_completion(
         self,

src/llama_stack/providers/remote/inference/nvidia/config.py

@@ -7,7 +7,7 @@
 import os
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, HttpUrl
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -44,18 +44,14 @@ class NVIDIAConfig(RemoteInferenceProviderConfig):
     URL of your running NVIDIA NIM and do not need to set the api_key.
     """
 
-    url: str = Field(
+    base_url: HttpUrl | None = Field(
-        default_factory=lambda: os.getenv("NVIDIA_BASE_URL", "https://integrate.api.nvidia.com"),
+        default_factory=lambda: os.getenv("NVIDIA_BASE_URL", "https://integrate.api.nvidia.com/v1"),
         description="A base url for accessing the NVIDIA NIM",
     )
     timeout: int = Field(
         default=60,
         description="Timeout for the HTTP requests",
     )
-    append_api_version: bool = Field(
-        default_factory=lambda: os.getenv("NVIDIA_APPEND_API_VERSION", "True").lower() != "false",
-        description="When set to false, the API version will not be appended to the base_url. By default, it is true.",
-    )
     rerank_model_to_url: dict[str, str] = Field(
         default_factory=lambda: {
             "nv-rerank-qa-mistral-4b:1": "https://ai.api.nvidia.com/v1/retrieval/nvidia/reranking",
@@ -68,13 +64,11 @@ class NVIDIAConfig(RemoteInferenceProviderConfig):
     @classmethod
     def sample_run_config(
         cls,
-        url: str = "${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com}",
+        base_url: HttpUrl | None = "${env.NVIDIA_BASE_URL:=https://integrate.api.nvidia.com/v1}",
         api_key: str = "${env.NVIDIA_API_KEY:=}",
-        append_api_version: bool = "${env.NVIDIA_APPEND_API_VERSION:=True}",
         **kwargs,
     ) -> dict[str, Any]:
         return {
-            "url": url,
+            "base_url": base_url,
             "api_key": api_key,
-            "append_api_version": append_api_version,
         }

src/llama_stack/providers/remote/inference/nvidia/nvidia.py

@@ -44,7 +44,7 @@ class NVIDIAInferenceAdapter(OpenAIMixin):
     }
 
     async def initialize(self) -> None:
-        logger.info(f"Initializing NVIDIAInferenceAdapter({self.config.url})...")
+        logger.info(f"Initializing NVIDIAInferenceAdapter({self.config.base_url})...")
 
         if _is_nvidia_hosted(self.config):
             if not self.config.auth_credential:
@@ -72,7 +72,7 @@ class NVIDIAInferenceAdapter(OpenAIMixin):
 
         :return: The NVIDIA API base URL
         """
-        return f"{self.config.url}/v1" if self.config.append_api_version else self.config.url
+        return str(self.config.base_url)
 
     async def list_provider_model_ids(self) -> Iterable[str]:
         """

src/llama_stack/providers/remote/inference/nvidia/utils.py

@@ -8,4 +8,4 @@ from . import NVIDIAConfig
 
 
 def _is_nvidia_hosted(config: NVIDIAConfig) -> bool:
-    return "integrate.api.nvidia.com" in config.url
+    return "integrate.api.nvidia.com" in str(config.base_url)

src/llama_stack/providers/remote/inference/ollama/config.py

@@ -6,20 +6,22 @@
 
 from typing import Any
 
-from pydantic import Field, SecretStr
+from pydantic import Field, HttpUrl, SecretStr
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 
-DEFAULT_OLLAMA_URL = "http://localhost:11434"
+DEFAULT_OLLAMA_URL = "http://localhost:11434/v1"
 
 
 class OllamaImplConfig(RemoteInferenceProviderConfig):
     auth_credential: SecretStr | None = Field(default=None, exclude=True)
 
-    url: str = DEFAULT_OLLAMA_URL
+    base_url: HttpUrl | None = Field(default=HttpUrl(DEFAULT_OLLAMA_URL))
 
     @classmethod
-    def sample_run_config(cls, url: str = "${env.OLLAMA_URL:=http://localhost:11434}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(
+        cls, base_url: str = "${env.OLLAMA_URL:=http://localhost:11434/v1}", **kwargs
+    ) -> dict[str, Any]:
         return {
-            "url": url,
+            "base_url": base_url,
         }

src/llama_stack/providers/remote/inference/ollama/ollama.py

@@ -55,17 +55,23 @@ class OllamaInferenceAdapter(OpenAIMixin):
         # ollama client attaches itself to the current event loop (sadly?)
         loop = asyncio.get_running_loop()
         if loop not in self._clients:
-            self._clients[loop] = AsyncOllamaClient(host=self.config.url)
+            # Ollama client expects base URL without /v1 suffix
+            base_url_str = str(self.config.base_url)
+            if base_url_str.endswith("/v1"):
+                host = base_url_str[:-3]
+            else:
+                host = base_url_str
+            self._clients[loop] = AsyncOllamaClient(host=host)
         return self._clients[loop]
 
     def get_api_key(self):
         return "NO KEY REQUIRED"
 
     def get_base_url(self):
-        return self.config.url.rstrip("/") + "/v1"
+        return str(self.config.base_url)
 
     async def initialize(self) -> None:
-        logger.info(f"checking connectivity to Ollama at `{self.config.url}`...")
+        logger.info(f"checking connectivity to Ollama at `{self.config.base_url}`...")
         r = await self.health()
         if r["status"] == HealthStatus.ERROR:
             logger.warning(

src/llama_stack/providers/remote/inference/openai/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, HttpUrl
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -21,8 +21,8 @@ class OpenAIProviderDataValidator(BaseModel):
 
 @json_schema_type
 class OpenAIConfig(RemoteInferenceProviderConfig):
-    base_url: str = Field(
+    base_url: HttpUrl | None = Field(
-        default="https://api.openai.com/v1",
+        default=HttpUrl("https://api.openai.com/v1"),
         description="Base URL for OpenAI API",
     )
 

src/llama_stack/providers/remote/inference/openai/openai.py

@@ -35,4 +35,4 @@ class OpenAIInferenceAdapter(OpenAIMixin):
 
         Returns the OpenAI API base URL from the configuration.
         """
-        return self.config.base_url
+        return str(self.config.base_url)

src/llama_stack/providers/remote/inference/passthrough/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import Field
+from pydantic import Field, HttpUrl
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -14,16 +14,16 @@ from llama_stack_api import json_schema_type
 
 @json_schema_type
 class PassthroughImplConfig(RemoteInferenceProviderConfig):
-    url: str = Field(
+    base_url: HttpUrl | None = Field(
         default=None,
         description="The URL for the passthrough endpoint",
     )
 
     @classmethod
     def sample_run_config(
-        cls, url: str = "${env.PASSTHROUGH_URL}", api_key: str = "${env.PASSTHROUGH_API_KEY}", **kwargs
+        cls, base_url: HttpUrl | None = "${env.PASSTHROUGH_URL}", api_key: str = "${env.PASSTHROUGH_API_KEY}", **kwargs
     ) -> dict[str, Any]:
         return {
-            "url": url,
+            "base_url": base_url,
             "api_key": api_key,
         }

src/llama_stack/providers/remote/inference/passthrough/passthrough.py

@@ -82,8 +82,8 @@ class PassthroughInferenceAdapter(NeedsRequestProviderData, Inference):
 
     def _get_passthrough_url(self) -> str:
         """Get the passthrough URL from config or provider data."""
-        if self.config.url is not None:
+        if self.config.base_url is not None:
-            return self.config.url
+            return str(self.config.base_url)
 
         provider_data = self.get_request_provider_data()
         if provider_data is None:

src/llama_stack/providers/remote/inference/runpod/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field, SecretStr
+from pydantic import BaseModel, Field, HttpUrl, SecretStr
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -21,7 +21,7 @@ class RunpodProviderDataValidator(BaseModel):
 
 @json_schema_type
 class RunpodImplConfig(RemoteInferenceProviderConfig):
-    url: str | None = Field(
+    base_url: HttpUrl | None = Field(
         default=None,
         description="The URL for the Runpod model serving endpoint",
     )
@@ -34,6 +34,6 @@ class RunpodImplConfig(RemoteInferenceProviderConfig):
     @classmethod
     def sample_run_config(cls, **kwargs: Any) -> dict[str, Any]:
         return {
-            "url": "${env.RUNPOD_URL:=}",
+            "base_url": "${env.RUNPOD_URL:=}",
             "api_token": "${env.RUNPOD_API_TOKEN}",
         }

src/llama_stack/providers/remote/inference/runpod/runpod.py

@@ -28,7 +28,7 @@ class RunpodInferenceAdapter(OpenAIMixin):
 
     def get_base_url(self) -> str:
         """Get base URL for OpenAI client."""
-        return self.config.url
+        return str(self.config.base_url)
 
     async def openai_chat_completion(
         self,

src/llama_stack/providers/remote/inference/sambanova/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, HttpUrl
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -21,14 +21,14 @@ class SambaNovaProviderDataValidator(BaseModel):
 
 @json_schema_type
 class SambaNovaImplConfig(RemoteInferenceProviderConfig):
-    url: str = Field(
+    base_url: HttpUrl | None = Field(
-        default="https://api.sambanova.ai/v1",
+        default=HttpUrl("https://api.sambanova.ai/v1"),
         description="The URL for the SambaNova AI server",
     )
 
     @classmethod
     def sample_run_config(cls, api_key: str = "${env.SAMBANOVA_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
-            "url": "https://api.sambanova.ai/v1",
+            "base_url": "https://api.sambanova.ai/v1",
             "api_key": api_key,
         }

src/llama_stack/providers/remote/inference/sambanova/sambanova.py

@@ -25,4 +25,4 @@ class SambaNovaInferenceAdapter(OpenAIMixin):
 
         :return: The SambaNova base URL
         """
-        return self.config.url
+        return str(self.config.base_url)

src/llama_stack/providers/remote/inference/tgi/config.py

@@ -5,7 +5,7 @@
 # the root directory of this source tree.
 
 
-from pydantic import BaseModel, Field, SecretStr
+from pydantic import BaseModel, Field, HttpUrl, SecretStr
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -15,18 +15,19 @@ from llama_stack_api import json_schema_type
 class TGIImplConfig(RemoteInferenceProviderConfig):
     auth_credential: SecretStr | None = Field(default=None, exclude=True)
 
-    url: str = Field(
+    base_url: HttpUrl | None = Field(
-        description="The URL for the TGI serving endpoint",
+        default=None,
+        description="The URL for the TGI serving endpoint (should include /v1 path)",
     )
 
     @classmethod
     def sample_run_config(
         cls,
-        url: str = "${env.TGI_URL:=}",
+        base_url: str = "${env.TGI_URL:=}",
         **kwargs,
     ):
         return {
-            "url": url,
+            "base_url": base_url,
         }
 
 

src/llama_stack/providers/remote/inference/tgi/tgi.py

@@ -8,7 +8,7 @@
 from collections.abc import Iterable
 
 from huggingface_hub import AsyncInferenceClient, HfApi
-from pydantic import SecretStr
+from pydantic import HttpUrl, SecretStr
 
 from llama_stack.log import get_logger
 from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
@@ -23,7 +23,7 @@ log = get_logger(name=__name__, category="inference::tgi")
 
 
 class _HfAdapter(OpenAIMixin):
-    url: str
+    base_url: HttpUrl
     api_key: SecretStr
 
     hf_client: AsyncInferenceClient
@@ -36,7 +36,7 @@ class _HfAdapter(OpenAIMixin):
         return "NO KEY REQUIRED"
 
     def get_base_url(self):
-        return self.url
+        return self.base_url
 
     async def list_provider_model_ids(self) -> Iterable[str]:
         return [self.model_id]
@@ -50,14 +50,20 @@ class _HfAdapter(OpenAIMixin):
 
 class TGIAdapter(_HfAdapter):
     async def initialize(self, config: TGIImplConfig) -> None:
-        if not config.url:
+        if not config.base_url:
             raise ValueError("You must provide a URL in run.yaml (or via the TGI_URL environment variable) to use TGI.")
-        log.info(f"Initializing TGI client with url={config.url}")
+        log.info(f"Initializing TGI client with url={config.base_url}")
-        self.hf_client = AsyncInferenceClient(model=config.url, provider="hf-inference")
+        # Extract base URL without /v1 for HF client initialization
+        base_url_str = str(config.base_url).rstrip("/")
+        if base_url_str.endswith("/v1"):
+            base_url_for_client = base_url_str[:-3]
+        else:
+            base_url_for_client = base_url_str
+        self.hf_client = AsyncInferenceClient(model=base_url_for_client, provider="hf-inference")
         endpoint_info = await self.hf_client.get_endpoint_info()
         self.max_tokens = endpoint_info["max_total_tokens"]
         self.model_id = endpoint_info["model_id"]
-        self.url = f"{config.url.rstrip('/')}/v1"
+        self.base_url = config.base_url
         self.api_key = SecretStr("NO_KEY")
 
 

src/llama_stack/providers/remote/inference/together/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import Field
+from pydantic import Field, HttpUrl
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -14,14 +14,14 @@ from llama_stack_api import json_schema_type
 
 @json_schema_type
 class TogetherImplConfig(RemoteInferenceProviderConfig):
-    url: str = Field(
+    base_url: HttpUrl | None = Field(
-        default="https://api.together.xyz/v1",
+        default=HttpUrl("https://api.together.xyz/v1"),
         description="The URL for the Together AI server",
     )
 
     @classmethod
     def sample_run_config(cls, **kwargs) -> dict[str, Any]:
         return {
-            "url": "https://api.together.xyz/v1",
+            "base_url": "https://api.together.xyz/v1",
             "api_key": "${env.TOGETHER_API_KEY:=}",
         }

src/llama_stack/providers/remote/inference/together/together.py

@@ -9,7 +9,6 @@ from collections.abc import Iterable
 from typing import Any, cast
 
 from together import AsyncTogether  # type: ignore[import-untyped]
-from together.constants import BASE_URL  # type: ignore[import-untyped]
 
 from llama_stack.core.request_headers import NeedsRequestProviderData
 from llama_stack.log import get_logger
@@ -42,7 +41,7 @@ class TogetherInferenceAdapter(OpenAIMixin, NeedsRequestProviderData):
     provider_data_api_key_field: str = "together_api_key"
 
     def get_base_url(self):
-        return BASE_URL
+        return str(self.config.base_url)
 
     def _get_client(self) -> AsyncTogether:
         together_api_key = None

src/llama_stack/providers/remote/inference/vllm/config.py

@@ -6,7 +6,7 @@
 
 from pathlib import Path
 
-from pydantic import Field, SecretStr, field_validator
+from pydantic import Field, HttpUrl, SecretStr, field_validator
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -14,7 +14,7 @@ from llama_stack_api import json_schema_type
 
 @json_schema_type
 class VLLMInferenceAdapterConfig(RemoteInferenceProviderConfig):
-    url: str | None = Field(
+    base_url: HttpUrl | None = Field(
         default=None,
         description="The URL for the vLLM model serving endpoint",
     )
@@ -48,11 +48,11 @@ class VLLMInferenceAdapterConfig(RemoteInferenceProviderConfig):
     @classmethod
     def sample_run_config(
         cls,
-        url: str = "${env.VLLM_URL:=}",
+        base_url: str = "${env.VLLM_URL:=}",
         **kwargs,
     ):
         return {
-            "url": url,
+            "base_url": base_url,
             "max_tokens": "${env.VLLM_MAX_TOKENS:=4096}",
             "api_token": "${env.VLLM_API_TOKEN:=fake}",
             "tls_verify": "${env.VLLM_TLS_VERIFY:=true}",

src/llama_stack/providers/remote/inference/vllm/vllm.py

@@ -39,12 +39,12 @@ class VLLMInferenceAdapter(OpenAIMixin):
 
     def get_base_url(self) -> str:
         """Get the base URL from config."""
-        if not self.config.url:
+        if not self.config.base_url:
             raise ValueError("No base URL configured")
-        return self.config.url
+        return str(self.config.base_url)
 
     async def initialize(self) -> None:
-        if not self.config.url:
+        if not self.config.base_url:
             raise ValueError(
                 "You must provide a URL in run.yaml (or via the VLLM_URL environment variable) to use vLLM."
             )

src/llama_stack/providers/remote/inference/watsonx/config.py

@@ -7,7 +7,7 @@
 import os
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, HttpUrl
 
 from llama_stack.providers.utils.inference.model_registry import RemoteInferenceProviderConfig
 from llama_stack_api import json_schema_type
@@ -23,7 +23,7 @@ class WatsonXProviderDataValidator(BaseModel):
 
 @json_schema_type
 class WatsonXConfig(RemoteInferenceProviderConfig):
-    url: str = Field(
+    base_url: HttpUrl | None = Field(
         default_factory=lambda: os.getenv("WATSONX_BASE_URL", "https://us-south.ml.cloud.ibm.com"),
         description="A base url for accessing the watsonx.ai",
     )
@@ -39,7 +39,7 @@ class WatsonXConfig(RemoteInferenceProviderConfig):
     @classmethod
     def sample_run_config(cls, **kwargs) -> dict[str, Any]:
         return {
-            "url": "${env.WATSONX_BASE_URL:=https://us-south.ml.cloud.ibm.com}",
+            "base_url": "${env.WATSONX_BASE_URL:=https://us-south.ml.cloud.ibm.com}",
             "api_key": "${env.WATSONX_API_KEY:=}",
             "project_id": "${env.WATSONX_PROJECT_ID:=}",
         }

src/llama_stack/providers/remote/inference/watsonx/watsonx.py

@@ -255,7 +255,7 @@ class WatsonXInferenceAdapter(LiteLLMOpenAIMixin):
         )
 
     def get_base_url(self) -> str:
-        return self.config.url
+        return str(self.config.base_url)
 
     # Copied from OpenAIMixin
     async def check_model_availability(self, model: str) -> bool:
@@ -316,7 +316,7 @@ class WatsonXInferenceAdapter(LiteLLMOpenAIMixin):
         """
         Retrieves foundation model specifications from the watsonx.ai API.
         """
-        url = f"{self.config.url}/ml/v1/foundation_model_specs?version=2023-10-25"
+        url = f"{str(self.config.base_url)}/ml/v1/foundation_model_specs?version=2023-10-25"
         headers = {
             # Note that there is no authorization header.  Listing models does not require authentication.
             "Content-Type": "application/json",

src/llama_stack/providers/utils/inference/openai_compat.py

@@ -3,23 +3,10 @@
 #
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
-from collections.abc import Iterable
 from typing import (
     Any,
 )
 
-from openai.types.chat import (
-    ChatCompletionContentPartParam as OpenAIChatCompletionContentPartParam,
-)
-
-try:
-    from openai.types.chat import (
-        ChatCompletionMessageFunctionToolCall as OpenAIChatCompletionMessageFunctionToolCall,
-    )
-except ImportError:
-    from openai.types.chat.chat_completion_message_tool_call import (
-        ChatCompletionMessageToolCall as OpenAIChatCompletionMessageFunctionToolCall,
-    )
 from openai.types.chat import (
     ChatCompletionMessageToolCall,
 )
@@ -32,18 +19,6 @@ from llama_stack.models.llama.datatypes import (
     ToolCall,
     ToolDefinition,
 )
-from llama_stack_api import (
-    URL,
-    GreedySamplingStrategy,
-    ImageContentItem,
-    JsonSchemaResponseFormat,
-    OpenAIResponseFormatParam,
-    SamplingParams,
-    TextContentItem,
-    TopKSamplingStrategy,
-    TopPSamplingStrategy,
-    _URLOrData,
-)
 
 logger = get_logger(name=__name__, category="providers::utils")
 
@@ -73,42 +48,6 @@ class OpenAICompatCompletionResponse(BaseModel):
     choices: list[OpenAICompatCompletionChoice]
 
 
-def get_sampling_strategy_options(params: SamplingParams) -> dict:
-    options = {}
-    if isinstance(params.strategy, GreedySamplingStrategy):
-        options["temperature"] = 0.0
-    elif isinstance(params.strategy, TopPSamplingStrategy):
-        if params.strategy.temperature is not None:
-            options["temperature"] = params.strategy.temperature
-        if params.strategy.top_p is not None:
-            options["top_p"] = params.strategy.top_p
-    elif isinstance(params.strategy, TopKSamplingStrategy):
-        options["top_k"] = params.strategy.top_k
-    else:
-        raise ValueError(f"Unsupported sampling strategy: {params.strategy}")
-
-    return options
-
-
-def get_sampling_options(params: SamplingParams | None) -> dict:
-    if not params:
-        return {}
-
-    options = {}
-    if params:
-        options.update(get_sampling_strategy_options(params))
-        if params.max_tokens:
-            options["max_tokens"] = params.max_tokens
-
-        if params.repetition_penalty is not None and params.repetition_penalty != 1.0:
-            options["repeat_penalty"] = params.repetition_penalty
-
-        if params.stop is not None:
-            options["stop"] = params.stop
-
-    return options
-
-
 def text_from_choice(choice) -> str:
     if hasattr(choice, "delta") and choice.delta:
         return choice.delta.content  # type: ignore[no-any-return]  # external OpenAI types lack precise annotations
@@ -253,154 +192,6 @@ def convert_tooldef_to_openai_tool(tool: ToolDefinition) -> dict:
     return out
 
 
-def _convert_stop_reason_to_openai_finish_reason(stop_reason: StopReason) -> str:
-    """
-    Convert a StopReason to an OpenAI chat completion finish_reason.
-    """
-    return {
-        StopReason.end_of_turn: "stop",
-        StopReason.end_of_message: "tool_calls",
-        StopReason.out_of_tokens: "length",
-    }.get(stop_reason, "stop")
-
-
-def _convert_openai_finish_reason(finish_reason: str) -> StopReason:
-    """
-    Convert an OpenAI chat completion finish_reason to a StopReason.
-
-    finish_reason: Literal["stop", "length", "tool_calls", ...]
-        - stop: model hit a natural stop point or a provided stop sequence
-        - length: maximum number of tokens specified in the request was reached
-        - tool_calls: model called a tool
-
-    ->
-
-    class StopReason(Enum):
-        end_of_turn = "end_of_turn"
-        end_of_message = "end_of_message"
-        out_of_tokens = "out_of_tokens"
-    """
-
-    # TODO(mf): are end_of_turn and end_of_message semantics correct?
-    return {
-        "stop": StopReason.end_of_turn,
-        "length": StopReason.out_of_tokens,
-        "tool_calls": StopReason.end_of_message,
-    }.get(finish_reason, StopReason.end_of_turn)
-
-
-def _convert_openai_request_tools(tools: list[dict[str, Any]] | None = None) -> list[ToolDefinition]:
-    lls_tools: list[ToolDefinition] = []
-    if not tools:
-        return lls_tools
-
-    for tool in tools:
-        tool_fn = tool.get("function", {})
-        tool_name = tool_fn.get("name", None)
-        tool_desc = tool_fn.get("description", None)
-        tool_params = tool_fn.get("parameters", None)
-
-        lls_tool = ToolDefinition(
-            tool_name=tool_name,
-            description=tool_desc,
-            input_schema=tool_params,  # Pass through entire JSON Schema
-        )
-        lls_tools.append(lls_tool)
-    return lls_tools
-
-
-def _convert_openai_request_response_format(
-    response_format: OpenAIResponseFormatParam | None = None,
-):
-    if not response_format:
-        return None
-    # response_format can be a dict or a pydantic model
-    response_format_dict = dict(response_format)  # type: ignore[arg-type]  # OpenAIResponseFormatParam union needs dict conversion
-    if response_format_dict.get("type", "") == "json_schema":
-        return JsonSchemaResponseFormat(
-            type="json_schema",  # type: ignore[arg-type]  # Literal["json_schema"] incompatible with expected type
-            json_schema=response_format_dict.get("json_schema", {}).get("schema", ""),
-        )
-    return None
-
-
-def _convert_openai_tool_calls(
-    tool_calls: list[OpenAIChatCompletionMessageFunctionToolCall],
-) -> list[ToolCall]:
-    """
-    Convert an OpenAI ChatCompletionMessageToolCall list into a list of ToolCall.
-
-    OpenAI ChatCompletionMessageToolCall:
-        id: str
-        function: Function
-        type: Literal["function"]
-
-    OpenAI Function:
-        arguments: str
-        name: str
-
-    ->
-
-    ToolCall:
-        call_id: str
-        tool_name: str
-        arguments: Dict[str, ...]
-    """
-    if not tool_calls:
-        return []  # CompletionMessage tool_calls is not optional
-
-    return [
-        ToolCall(
-            call_id=call.id,
-            tool_name=call.function.name,
-            arguments=call.function.arguments,
-        )
-        for call in tool_calls
-    ]
-
-
-def _convert_openai_sampling_params(
-    max_tokens: int | None = None,
-    temperature: float | None = None,
-    top_p: float | None = None,
-) -> SamplingParams:
-    sampling_params = SamplingParams()
-
-    if max_tokens:
-        sampling_params.max_tokens = max_tokens
-
-    # Map an explicit temperature of 0 to greedy sampling
-    if temperature == 0:
-        sampling_params.strategy = GreedySamplingStrategy()
-    else:
-        # OpenAI defaults to 1.0 for temperature and top_p if unset
-        if temperature is None:
-            temperature = 1.0
-        if top_p is None:
-            top_p = 1.0
-        sampling_params.strategy = TopPSamplingStrategy(temperature=temperature, top_p=top_p)  # type: ignore[assignment]  # SamplingParams.strategy union accepts this type
-
-    return sampling_params
-
-
-def openai_content_to_content(content: str | Iterable[OpenAIChatCompletionContentPartParam] | None):
-    if content is None:
-        return ""
-    if isinstance(content, str):
-        return content
-    elif isinstance(content, list):
-        return [openai_content_to_content(c) for c in content]
-    elif hasattr(content, "type"):
-        if content.type == "text":
-            return TextContentItem(type="text", text=content.text)  # type: ignore[attr-defined]  # Iterable narrowed by hasattr check but mypy doesn't track
-        elif content.type == "image_url":
-            return ImageContentItem(type="image", image=_URLOrData(url=URL(uri=content.image_url.url)))  # type: ignore[attr-defined]  # Iterable narrowed by hasattr check but mypy doesn't track
-        else:
-            raise ValueError(f"Unknown content type: {content.type}")
-    else:
-        raise ValueError(f"Unknown content type: {content}")
-
-
 async def prepare_openai_completion_params(**params):
     async def _prepare_value(value: Any) -> Any:
         new_value = value

src/llama_stack/providers/utils/inference/openai_mixin.py

@@ -213,6 +213,19 @@ class OpenAIMixin(NeedsRequestProviderData, ABC, BaseModel):
 
         return api_key
 
+    def _validate_model_allowed(self, provider_model_id: str) -> None:
+        """
+        Validate that the model is in the allowed_models list if configured.
+
+        :param provider_model_id: The provider-specific model ID to validate
+        :raises ValueError: If the model is not in the allowed_models list
+        """
+        if self.config.allowed_models is not None and provider_model_id not in self.config.allowed_models:
+            raise ValueError(
+                f"Model '{provider_model_id}' is not in the allowed models list. "
+                f"Allowed models: {self.config.allowed_models}"
+            )
+
     async def _get_provider_model_id(self, model: str) -> str:
         """
         Get the provider-specific model ID from the model store.
@@ -259,8 +272,11 @@ class OpenAIMixin(NeedsRequestProviderData, ABC, BaseModel):
         Direct OpenAI completion API call.
         """
         # TODO: fix openai_completion to return type compatible with OpenAI's API response
+        provider_model_id = await self._get_provider_model_id(params.model)
+        self._validate_model_allowed(provider_model_id)
+
         completion_kwargs = await prepare_openai_completion_params(
-            model=await self._get_provider_model_id(params.model),
+            model=provider_model_id,
             prompt=params.prompt,
             best_of=params.best_of,
             echo=params.echo,
@@ -292,6 +308,9 @@ class OpenAIMixin(NeedsRequestProviderData, ABC, BaseModel):
         """
         Direct OpenAI chat completion API call.
         """
+        provider_model_id = await self._get_provider_model_id(params.model)
+        self._validate_model_allowed(provider_model_id)
+
         messages = params.messages
 
         if self.download_images:
@@ -313,7 +332,7 @@ class OpenAIMixin(NeedsRequestProviderData, ABC, BaseModel):
             messages = [await _localize_image_url(m) for m in messages]
 
         request_params = await prepare_openai_completion_params(
-            model=await self._get_provider_model_id(params.model),
+            model=provider_model_id,
             messages=messages,
             frequency_penalty=params.frequency_penalty,
             function_call=params.function_call,
@@ -351,10 +370,13 @@ class OpenAIMixin(NeedsRequestProviderData, ABC, BaseModel):
         """
         Direct OpenAI embeddings API call.
         """
+        provider_model_id = await self._get_provider_model_id(params.model)
+        self._validate_model_allowed(provider_model_id)
+
         # Build request params conditionally to avoid NotGiven/Omit type mismatch
         # The OpenAI SDK uses Omit in signatures but NOT_GIVEN has type NotGiven
         request_params: dict[str, Any] = {
-            "model": await self._get_provider_model_id(params.model),
+            "model": provider_model_id,
             "input": params.input,
         }
         if params.encoding_format is not None:

src/llama_stack_api/vector_io.py

@@ -11,7 +11,7 @@
 from typing import Annotated, Any, Literal, Protocol, runtime_checkable
 
 from fastapi import Body, Query
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, field_validator
 
 from llama_stack_api.common.tracing import telemetry_traceable
 from llama_stack_api.inference import InterleavedContent
@@ -372,6 +372,65 @@ VectorStoreFileStatus = Literal["completed"] | Literal["in_progress"] | Literal[
 register_schema(VectorStoreFileStatus, name="VectorStoreFileStatus")
 
 
+# VectorStoreFileAttributes type with OpenAPI constraints
+VectorStoreFileAttributes = Annotated[
+    dict[str, Annotated[str, Field(max_length=512)] | float | bool],
+    Field(
+        max_length=16,
+        json_schema_extra={
+            "propertyNames": {"type": "string", "maxLength": 64},
+            "x-oaiTypeLabel": "map",
+        },
+        description=(
+            "Set of 16 key-value pairs that can be attached to an object. This can be "
+            "useful for storing additional information about the object in a structured "
+            "format, and querying for objects via API or the dashboard. Keys are strings "
+            "with a maximum length of 64 characters. Values are strings with a maximum "
+            "length of 512 characters, booleans, or numbers."
+        ),
+    ),
+]
+
+
+def _sanitize_vector_store_attributes(metadata: dict[str, Any] | None) -> dict[str, str | float | bool]:
+    """
+    Sanitize metadata to VectorStoreFileAttributes spec (max 16 properties, primitives only).
+
+    Converts dict[str, Any] to dict[str, str | float | bool]:
+    - Preserves: str (truncated to 512 chars), bool, int/float (as float)
+    - Converts: list -> comma-separated string
+    - Filters: dict, None, other types
+    - Enforces: max 16 properties, max 64 char keys, max 512 char string values
+    """
+    if not metadata:
+        return {}
+
+    sanitized: dict[str, str | float | bool] = {}
+    for key, value in metadata.items():
+        # Enforce max 16 properties
+        if len(sanitized) >= 16:
+            break
+
+        # Enforce max 64 char keys
+        if len(key) > 64:
+            continue
+
+        # Convert to supported primitive types
+        if isinstance(value, bool):
+            sanitized[key] = value
+        elif isinstance(value, int | float):
+            sanitized[key] = float(value)
+        elif isinstance(value, str):
+            # Enforce max 512 char string values
+            sanitized[key] = value[:512] if len(value) > 512 else value
+        elif isinstance(value, list):
+            # Convert lists to comma-separated strings (max 512 chars)
+            list_str = ", ".join(str(item) for item in value)
+            sanitized[key] = list_str[:512] if len(list_str) > 512 else list_str
+
+    return sanitized
+
+
 @json_schema_type
 class VectorStoreFileObject(BaseModel):
     """OpenAI Vector Store File object.
@@ -389,7 +448,7 @@ class VectorStoreFileObject(BaseModel):
 
     id: str
     object: str = "vector_store.file"
-    attributes: dict[str, Any] = Field(default_factory=dict)
+    attributes: VectorStoreFileAttributes = Field(default_factory=dict)
     chunking_strategy: VectorStoreChunkingStrategy
     created_at: int
     last_error: VectorStoreFileLastError | None = None
@@ -397,6 +456,12 @@ class VectorStoreFileObject(BaseModel):
     usage_bytes: int = 0
     vector_store_id: str
 
+    @field_validator("attributes", mode="before")
+    @classmethod
+    def _validate_attributes(cls, v: dict[str, Any] | None) -> dict[str, str | float | bool]:
+        """Sanitize attributes to match VectorStoreFileAttributes OpenAPI spec."""
+        return _sanitize_vector_store_attributes(v)
+
 
 @json_schema_type
 class VectorStoreListFilesResponse(BaseModel):

tests/integration/README.md

@@ -211,3 +211,23 @@ def test_asymmetric_embeddings(llama_stack_client, embedding_model_id):
 
     assert query_response.embeddings is not None
 ```
+
+## TypeScript Client Replays
+
+TypeScript SDK tests can run alongside Python tests when testing against `server:<config>` stacks. Set `TS_CLIENT_PATH` to the path or version of `llama-stack-client-typescript` to enable:
+
+```bash
+# Use published npm package (responses suite)
+TS_CLIENT_PATH=^0.3.2 scripts/integration-tests.sh --stack-config server:ci-tests --suite responses --setup gpt
+
+# Use local checkout from ~/.cache (recommended for development)
+git clone https://github.com/llamastack/llama-stack-client-typescript.git ~/.cache/llama-stack-client-typescript
+TS_CLIENT_PATH=~/.cache/llama-stack-client-typescript scripts/integration-tests.sh --stack-config server:ci-tests --suite responses --setup gpt
+
+# Run base suite with TypeScript tests
+TS_CLIENT_PATH=~/.cache/llama-stack-client-typescript scripts/integration-tests.sh --stack-config server:ci-tests --suite base --setup ollama
+```
+
+TypeScript tests run immediately after Python tests pass, using the same replay fixtures. The mapping between Python suites/setups and TypeScript test files is defined in `tests/integration/client-typescript/suites.json`.
+
+If `TS_CLIENT_PATH` is unset, TypeScript tests are skipped entirely.

tests/integration/agents/test_openai_responses.py

@@ -516,169 +516,3 @@ def test_response_with_instructions(openai_client, client_with_models, text_mode
 
     # Verify instructions from previous response was not carried over to the next response
     assert response_with_instructions2.instructions == instructions2
-
-
-@pytest.mark.skip(reason="Tool calling is not reliable.")
-def test_max_tool_calls_with_function_tools(openai_client, client_with_models, text_model_id):
-    """Test handling of max_tool_calls with function tools in responses."""
-    if isinstance(client_with_models, LlamaStackAsLibraryClient):
-        pytest.skip("OpenAI responses are not supported when testing with library client yet.")
-
-    client = openai_client
-    max_tool_calls = 1
-
-    tools = [
-        {
-            "type": "function",
-            "name": "get_weather",
-            "description": "Get weather information for a specified location",
-            "parameters": {
-                "type": "object",
-                "properties": {
-                    "location": {
-                        "type": "string",
-                        "description": "The city name (e.g., 'New York', 'London')",
-                    },
-                },
-            },
-        },
-        {
-            "type": "function",
-            "name": "get_time",
-            "description": "Get current time for a specified location",
-            "parameters": {
-                "type": "object",
-                "properties": {
-                    "location": {
-                        "type": "string",
-                        "description": "The city name (e.g., 'New York', 'London')",
-                    },
-                },
-            },
-        },
-    ]
-
-    # First create a response that triggers function tools
-    response = client.responses.create(
-        model=text_model_id,
-        input="Can you tell me the weather in Paris and the current time?",
-        tools=tools,
-        stream=False,
-        max_tool_calls=max_tool_calls,
-    )
-
-    # Verify we got two function calls and that the max_tool_calls do not affect function tools
-    assert len(response.output) == 2
-    assert response.output[0].type == "function_call"
-    assert response.output[0].name == "get_weather"
-    assert response.output[0].status == "completed"
-    assert response.output[1].type == "function_call"
-    assert response.output[1].name == "get_time"
-    assert response.output[0].status == "completed"
-
-    # Verify we have a valid max_tool_calls field
-    assert response.max_tool_calls == max_tool_calls
-
-
-def test_max_tool_calls_invalid(openai_client, client_with_models, text_model_id):
-    """Test handling of invalid max_tool_calls in responses."""
-    if isinstance(client_with_models, LlamaStackAsLibraryClient):
-        pytest.skip("OpenAI responses are not supported when testing with library client yet.")
-
-    client = openai_client
-
-    input = "Search for today's top technology news."
-    invalid_max_tool_calls = 0
-    tools = [
-        {"type": "web_search"},
-    ]
-
-    # Create a response with an invalid max_tool_calls value i.e. 0
-    # Handle ValueError from LLS and BadRequestError from OpenAI client
-    with pytest.raises((ValueError, BadRequestError)) as excinfo:
-        client.responses.create(
-            model=text_model_id,
-            input=input,
-            tools=tools,
-            stream=False,
-            max_tool_calls=invalid_max_tool_calls,
-        )
-
-    error_message = str(excinfo.value)
-    assert f"Invalid max_tool_calls={invalid_max_tool_calls}; should be >= 1" in error_message, (
-        f"Expected error message about invalid max_tool_calls, got: {error_message}"
-    )
-
-
-def test_max_tool_calls_with_builtin_tools(openai_client, client_with_models, text_model_id):
-    """Test handling of max_tool_calls with built-in tools in responses."""
-    if isinstance(client_with_models, LlamaStackAsLibraryClient):
-        pytest.skip("OpenAI responses are not supported when testing with library client yet.")
-
-    client = openai_client
-
-    input = "Search for today's top technology and a positive news story. You MUST make exactly two separate web search calls."
-    max_tool_calls = [1, 5]
-    tools = [
-        {"type": "web_search"},
-    ]
-
-    # First create a response that triggers web_search tools without max_tool_calls
-    response = client.responses.create(
-        model=text_model_id,
-        input=input,
-        tools=tools,
-        stream=False,
-    )
-
-    # Verify we got two web search calls followed by a message
-    assert len(response.output) == 3
-    assert response.output[0].type == "web_search_call"
-    assert response.output[0].status == "completed"
-    assert response.output[1].type == "web_search_call"
-    assert response.output[1].status == "completed"
-    assert response.output[2].type == "message"
-    assert response.output[2].status == "completed"
-    assert response.output[2].role == "assistant"
-
-    # Next create a response that triggers web_search tools with max_tool_calls set to 1
-    response_2 = client.responses.create(
-        model=text_model_id,
-        input=input,
-        tools=tools,
-        stream=False,
-        max_tool_calls=max_tool_calls[0],
-    )
-
-    # Verify we got one web search tool call followed by a message
-    assert len(response_2.output) == 2
-    assert response_2.output[0].type == "web_search_call"
-    assert response_2.output[0].status == "completed"
-    assert response_2.output[1].type == "message"
-    assert response_2.output[1].status == "completed"
-    assert response_2.output[1].role == "assistant"
-
-    # Verify we have a valid max_tool_calls field
-    assert response_2.max_tool_calls == max_tool_calls[0]
-
-    # Finally create a response that triggers web_search tools with max_tool_calls set to 5
-    response_3 = client.responses.create(
-        model=text_model_id,
-        input=input,
-        tools=tools,
-        stream=False,
-        max_tool_calls=max_tool_calls[1],
-    )
-
-    # Verify we got two web search calls followed by a message
-    assert len(response_3.output) == 3
-    assert response_3.output[0].type == "web_search_call"
-    assert response_3.output[0].status == "completed"
-    assert response_3.output[1].type == "web_search_call"
-    assert response_3.output[1].status == "completed"
-    assert response_3.output[2].type == "message"
-    assert response_3.output[2].status == "completed"
-    assert response_3.output[2].role == "assistant"
-
-    # Verify we have a valid max_tool_calls field
-    assert response_3.max_tool_calls == max_tool_calls[1]

tests/integration/client-typescript/__tests__/inference.test.ts

@@ -0,0 +1,104 @@
+// 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.
+
+/**
+ * Integration tests for Inference API (Chat Completions).
+ * Ported from: llama-stack/tests/integration/inference/test_openai_completion.py
+ *
+ * IMPORTANT: Test cases must match EXACTLY with Python tests to use recorded API responses.
+ */
+
+import { createTestClient, requireTextModel } from '../setup';
+
+describe('Inference API - Chat Completions', () => {
+  // Test cases matching llama-stack/tests/integration/test_cases/inference/chat_completion.json
+  const chatCompletionTestCases = [
+    {
+      id: 'non_streaming_01',
+      question: 'Which planet do humans live on?',
+      expected: 'earth',
+      testId:
+        'tests/integration/inference/test_openai_completion.py::test_openai_chat_completion_non_streaming[client_with_models-txt=ollama/llama3.2:3b-instruct-fp16-inference:chat_completion:non_streaming_01]',
+    },
+    {
+      id: 'non_streaming_02',
+      question: 'Which planet has rings around it with a name starting with letter S?',
+      expected: 'saturn',
+      testId:
+        'tests/integration/inference/test_openai_completion.py::test_openai_chat_completion_non_streaming[client_with_models-txt=ollama/llama3.2:3b-instruct-fp16-inference:chat_completion:non_streaming_02]',
+    },
+  ];
+
+  const streamingTestCases = [
+    {
+      id: 'streaming_01',
+      question: "What's the name of the Sun in latin?",
+      expected: 'sol',
+      testId:
+        'tests/integration/inference/test_openai_completion.py::test_openai_chat_completion_streaming[client_with_models-txt=ollama/llama3.2:3b-instruct-fp16-inference:chat_completion:streaming_01]',
+    },
+    {
+      id: 'streaming_02',
+      question: 'What is the name of the US captial?',
+      expected: 'washington',
+      testId:
+        'tests/integration/inference/test_openai_completion.py::test_openai_chat_completion_streaming[client_with_models-txt=ollama/llama3.2:3b-instruct-fp16-inference:chat_completion:streaming_02]',
+    },
+  ];
+
+  test.each(chatCompletionTestCases)(
+    'chat completion non-streaming: $id',
+    async ({ question, expected, testId }) => {
+      const client = createTestClient(testId);
+      const textModel = requireTextModel();
+
+      const response = await client.chat.completions.create({
+        model: textModel,
+        messages: [
+          {
+            role: 'user',
+            content: question,
+          },
+        ],
+        stream: false,
+      });
+
+      // Non-streaming responses have choices with message property
+      const choice = response.choices[0];
+      expect(choice).toBeDefined();
+      if (!choice || !('message' in choice)) {
+        throw new Error('Expected non-streaming response with message');
+      }
+      const content = choice.message.content;
+      expect(content).toBeDefined();
+      const messageContent = typeof content === 'string' ? content.toLowerCase().trim() : '';
+      expect(messageContent.length).toBeGreaterThan(0);
+      expect(messageContent).toContain(expected.toLowerCase());
+    },
+  );
+
+  test.each(streamingTestCases)('chat completion streaming: $id', async ({ question, expected, testId }) => {
+    const client = createTestClient(testId);
+    const textModel = requireTextModel();
+
+    const stream = await client.chat.completions.create({
+      model: textModel,
+      messages: [{ role: 'user', content: question }],
+      stream: true,
+    });
+
+    const streamedContent: string[] = [];
+    for await (const chunk of stream) {
+      if (chunk.choices && chunk.choices.length > 0 && chunk.choices[0]?.delta?.content) {
+        streamedContent.push(chunk.choices[0].delta.content);
+      }
+    }
+
+    expect(streamedContent.length).toBeGreaterThan(0);
+    const fullContent = streamedContent.join('').toLowerCase().trim();
+    expect(fullContent).toContain(expected.toLowerCase());
+  });
+});

tests/integration/client-typescript/__tests__/responses.test.ts

@@ -0,0 +1,132 @@
+// 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.
+
+/**
+ * Integration tests for Responses API.
+ * Ported from: llama-stack/tests/integration/responses/test_basic_responses.py
+ *
+ * IMPORTANT: Test cases and IDs must match EXACTLY with Python tests to use recorded API responses.
+ */
+
+import { createTestClient, requireTextModel, getResponseOutputText } from '../setup';
+
+describe('Responses API - Basic', () => {
+  // Test cases matching llama-stack/tests/integration/responses/fixtures/test_cases.py
+  const basicTestCases = [
+    {
+      id: 'earth',
+      input: 'Which planet do humans live on?',
+      expected: 'earth',
+      // Use client_with_models fixture to match non-streaming recordings
+      testId:
+        'tests/integration/responses/test_basic_responses.py::test_response_non_streaming_basic[client_with_models-txt=openai/gpt-4o-earth]',
+    },
+    {
+      id: 'saturn',
+      input: 'Which planet has rings around it with a name starting with letter S?',
+      expected: 'saturn',
+      testId:
+        'tests/integration/responses/test_basic_responses.py::test_response_non_streaming_basic[client_with_models-txt=openai/gpt-4o-saturn]',
+    },
+  ];
+
+  test.each(basicTestCases)('non-streaming basic response: $id', async ({ input, expected, testId }) => {
+    // Create client with test_id for all requests
+    const client = createTestClient(testId);
+    const textModel = requireTextModel();
+
+    // Create a response
+    const response = await client.responses.create({
+      model: textModel,
+      input,
+      stream: false,
+    });
+
+    // Verify response has content
+    const outputText = getResponseOutputText(response).toLowerCase().trim();
+    expect(outputText.length).toBeGreaterThan(0);
+    expect(outputText).toContain(expected.toLowerCase());
+
+    // Verify usage is reported
+    expect(response.usage).toBeDefined();
+    expect(response.usage!.input_tokens).toBeGreaterThan(0);
+    expect(response.usage!.output_tokens).toBeGreaterThan(0);
+    expect(response.usage!.total_tokens).toBe(response.usage!.input_tokens + response.usage!.output_tokens);
+
+    // Verify stored response matches
+    const retrievedResponse = await client.responses.retrieve(response.id);
+    expect(getResponseOutputText(retrievedResponse)).toBe(getResponseOutputText(response));
+
+    // Test follow-up with previous_response_id
+    const nextResponse = await client.responses.create({
+      model: textModel,
+      input: 'Repeat your previous response in all caps.',
+      previous_response_id: response.id,
+    });
+    const nextOutputText = getResponseOutputText(nextResponse).trim();
+    expect(nextOutputText).toContain(expected.toUpperCase());
+  });
+
+  test.each(basicTestCases)('streaming basic response: $id', async ({ input, expected, testId }) => {
+    // Modify test_id for streaming variant
+    const streamingTestId = testId.replace(
+      'test_response_non_streaming_basic',
+      'test_response_streaming_basic',
+    );
+    const client = createTestClient(streamingTestId);
+    const textModel = requireTextModel();
+
+    // Create a streaming response
+    const stream = await client.responses.create({
+      model: textModel,
+      input,
+      stream: true,
+    });
+
+    const events: any[] = [];
+    let responseId = '';
+
+    for await (const chunk of stream) {
+      events.push(chunk);
+
+      if (chunk.type === 'response.created') {
+        // Verify response.created is the first event
+        expect(events.length).toBe(1);
+        expect(chunk.response.status).toBe('in_progress');
+        responseId = chunk.response.id;
+      } else if (chunk.type === 'response.completed') {
+        // Verify response.completed comes after response.created
+        expect(events.length).toBeGreaterThanOrEqual(2);
+        expect(chunk.response.status).toBe('completed');
+        expect(chunk.response.id).toBe(responseId);
+
+        // Verify content quality
+        const outputText = getResponseOutputText(chunk.response).toLowerCase().trim();
+        expect(outputText.length).toBeGreaterThan(0);
+        expect(outputText).toContain(expected.toLowerCase());
+
+        // Verify usage is reported
+        expect(chunk.response.usage).toBeDefined();
+        expect(chunk.response.usage!.input_tokens).toBeGreaterThan(0);
+        expect(chunk.response.usage!.output_tokens).toBeGreaterThan(0);
+        expect(chunk.response.usage!.total_tokens).toBe(
+          chunk.response.usage!.input_tokens + chunk.response.usage!.output_tokens,
+        );
+      }
+    }
+
+    // Verify we got both events
+    expect(events.length).toBeGreaterThanOrEqual(2);
+    const firstEvent = events[0];
+    const lastEvent = events[events.length - 1];
+    expect(firstEvent.type).toBe('response.created');
+    expect(lastEvent.type).toBe('response.completed');
+
+    // Verify stored response matches streamed response
+    const retrievedResponse = await client.responses.retrieve(responseId);
+    expect(getResponseOutputText(retrievedResponse)).toBe(getResponseOutputText(lastEvent.response));
+  });
+});

tests/integration/client-typescript/jest.integration.config.js

@@ -0,0 +1,31 @@
+// 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.
+
+/** @type {import('ts-jest').JestConfigWithTsJest} */
+module.exports = {
+  preset: 'ts-jest/presets/default-esm',
+  testEnvironment: 'node',
+  extensionsToTreatAsEsm: ['.ts'],
+  moduleNameMapper: {
+    '^(\\.{1,2}/.*)\\.js$': '$1',
+  },
+  transform: {
+    '^.+\\.tsx?$': [
+      'ts-jest',
+      {
+        useESM: true,
+        tsconfig: {
+          module: 'ES2022',
+          moduleResolution: 'bundler',
+        },
+      },
+    ],
+  },
+  testMatch: ['<rootDir>/__tests__/**/*.test.ts'],
+  setupFilesAfterEnv: ['<rootDir>/setup.ts'],
+  testTimeout: 60000, // 60 seconds (integration tests can be slow)
+  watchman: false, // Disable watchman to avoid permission issues
+};

tests/integration/client-typescript/package.json

@@ -0,0 +1,18 @@
+{
+  "name": "llama-stack-typescript-integration-tests",
+  "version": "0.0.1",
+  "private": true,
+  "description": "TypeScript client integration tests for Llama Stack",
+  "scripts": {
+    "test": "node run-tests.js"
+  },
+  "devDependencies": {
+    "@swc/core": "^1.3.102",
+    "@swc/jest": "^0.2.29",
+    "@types/jest": "^29.4.0",
+    "@types/node": "^20.0.0",
+    "jest": "^29.4.0",
+    "ts-jest": "^29.1.0",
+    "typescript": "^5.0.0"
+  }
+}

tests/integration/client-typescript/run-tests.js

@@ -0,0 +1,63 @@
+#!/usr/bin/env node
+// 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.
+
+/**
+ * Test runner that finds and executes TypeScript tests based on suite/setup mapping.
+ * Called by integration-tests.sh via npm test.
+ */
+
+const fs = require('fs');
+const path = require('path');
+const { execSync } = require('child_process');
+
+const suite = process.env.LLAMA_STACK_TEST_SUITE;
+const setup = process.env.LLAMA_STACK_TEST_SETUP || '';
+
+if (!suite) {
+  console.error('Error: LLAMA_STACK_TEST_SUITE environment variable is required');
+  process.exit(1);
+}
+
+// Read suites.json to find matching test files
+const suitesPath = path.join(__dirname, 'suites.json');
+if (!fs.existsSync(suitesPath)) {
+  console.log(`No TypeScript tests configured (${suitesPath} not found)`);
+  process.exit(0);
+}
+
+const suites = JSON.parse(fs.readFileSync(suitesPath, 'utf-8'));
+
+// Find matching entry
+let testFiles = [];
+for (const entry of suites) {
+  if (entry.suite !== suite) {
+    continue;
+  }
+  const entrySetup = entry.setup || '';
+  if (entrySetup && entrySetup !== setup) {
+    continue;
+  }
+  testFiles = entry.files || [];
+  break;
+}
+
+if (testFiles.length === 0) {
+  console.log(`No TypeScript integration tests mapped for suite ${suite} (setup ${setup})`);
+  process.exit(0);
+}
+
+console.log(`Running TypeScript tests for suite ${suite} (setup ${setup}): ${testFiles.join(', ')}`);
+
+// Run Jest with the mapped test files
+try {
+  execSync(`npx jest --config jest.integration.config.js ${testFiles.join(' ')}`, {
+    stdio: 'inherit',
+    cwd: __dirname,
+  });
+} catch (error) {
+  process.exit(error.status || 1);
+}

tests/integration/client-typescript/setup.ts

@@ -0,0 +1,162 @@
+// 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.
+
+/**
+ * Global setup for integration tests.
+ * This file mimics pytest's fixture system by providing shared test configuration.
+ */
+
+import LlamaStackClient from 'llama-stack-client';
+
+/**
+ * Load test configuration from the Python setup system.
+ * This reads setup definitions from tests/integration/suites.py via get_setup_env.py.
+ */
+function loadTestConfig() {
+  const baseURL = process.env['TEST_API_BASE_URL'];
+  const setupName = process.env['LLAMA_STACK_TEST_SETUP'];
+  const textModel = process.env['LLAMA_STACK_TEST_TEXT_MODEL'];
+  const embeddingModel = process.env['LLAMA_STACK_TEST_EMBEDDING_MODEL'];
+
+  if (!baseURL) {
+    throw new Error(
+      'TEST_API_BASE_URL is required for integration tests. ' +
+        'Run tests using: ./scripts/integration-test.sh',
+    );
+  }
+
+  return {
+    baseURL,
+    textModel,
+    embeddingModel,
+    setupName,
+  };
+}
+
+// Read configuration from environment variables (set by scripts/integration-test.sh)
+export const TEST_CONFIG = loadTestConfig();
+
+// Validate required configuration
+beforeAll(() => {
+  console.log('\n=== Integration Test Configuration ===');
+  console.log(`Base URL: ${TEST_CONFIG.baseURL}`);
+  console.log(`Setup: ${TEST_CONFIG.setupName || 'NOT SET'}`);
+  console.log(
+    `Text Model: ${TEST_CONFIG.textModel || 'NOT SET - tests requiring text model will be skipped'}`,
+  );
+  console.log(
+    `Embedding Model: ${
+      TEST_CONFIG.embeddingModel || 'NOT SET - tests requiring embedding model will be skipped'
+    }`,
+  );
+  console.log('=====================================\n');
+});
+
+/**
+ * Create a client instance for integration tests.
+ * Mimics pytest's `llama_stack_client` fixture.
+ *
+ * @param testId - Test ID to send in X-LlamaStack-Provider-Data header for replay mode.
+ *                 Format: "tests/integration/responses/test_basic_responses.py::test_name[params]"
+ */
+export function createTestClient(testId?: string): LlamaStackClient {
+  const headers: Record<string, string> = {};
+
+  // In server mode with replay, send test ID for recording isolation
+  if (process.env['LLAMA_STACK_TEST_STACK_CONFIG_TYPE'] === 'server' && testId) {
+    headers['X-LlamaStack-Provider-Data'] = JSON.stringify({
+      __test_id: testId,
+    });
+  }
+
+  return new LlamaStackClient({
+    baseURL: TEST_CONFIG.baseURL,
+    timeout: 60000, // 60 seconds
+    defaultHeaders: headers,
+  });
+}
+
+/**
+ * Skip test if required model is not configured.
+ * Mimics pytest's `skip_if_no_model` autouse fixture.
+ */
+export function skipIfNoModel(modelType: 'text' | 'embedding'): typeof test {
+  const model = modelType === 'text' ? TEST_CONFIG.textModel : TEST_CONFIG.embeddingModel;
+
+  if (!model) {
+    const envVar = modelType === 'text' ? 'LLAMA_STACK_TEST_TEXT_MODEL' : 'LLAMA_STACK_TEST_EMBEDDING_MODEL';
+    const message = `Skipping: ${modelType} model not configured (set ${envVar})`;
+    return test.skip.bind(test) as typeof test;
+  }
+
+  return test;
+}
+
+/**
+ * Get the configured text model, throwing if not set.
+ * Use this in tests that absolutely require a text model.
+ */
+export function requireTextModel(): string {
+  if (!TEST_CONFIG.textModel) {
+    throw new Error(
+      'LLAMA_STACK_TEST_TEXT_MODEL environment variable is required. ' +
+        'Run tests using: ./scripts/integration-test.sh',
+    );
+  }
+  return TEST_CONFIG.textModel;
+}
+
+/**
+ * Get the configured embedding model, throwing if not set.
+ * Use this in tests that absolutely require an embedding model.
+ */
+export function requireEmbeddingModel(): string {
+  if (!TEST_CONFIG.embeddingModel) {
+    throw new Error(
+      'LLAMA_STACK_TEST_EMBEDDING_MODEL environment variable is required. ' +
+        'Run tests using: ./scripts/integration-test.sh',
+    );
+  }
+  return TEST_CONFIG.embeddingModel;
+}
+
+/**
+ * Extracts aggregated text output from a ResponseObject.
+ * This concatenates all text content from the response's output array.
+ *
+ * Copied from llama-stack-client's response-helpers until it's available in published version.
+ */
+export function getResponseOutputText(response: any): string {
+  const pieces: string[] = [];
+
+  for (const output of response.output ?? []) {
+    if (!output || output.type !== 'message') {
+      continue;
+    }
+
+    const content = output.content;
+    if (typeof content === 'string') {
+      pieces.push(content);
+      continue;
+    }
+
+    if (!Array.isArray(content)) {
+      continue;
+    }
+
+    for (const item of content) {
+      if (typeof item === 'string') {
+        pieces.push(item);
+        continue;
+      }
+      if (item && item.type === 'output_text' && 'text' in item && typeof item.text === 'string') {
+        pieces.push(item.text);
+      }
+    }
+  }
+
+  return pieces.join('');
+}

tests/integration/client-typescript/suites.json

@@ -0,0 +1,12 @@
+[
+  {
+    "suite": "responses",
+    "setup": "gpt",
+    "files": ["__tests__/responses.test.ts"]
+  },
+  {
+    "suite": "base",
+    "setup": "ollama",
+    "files": ["__tests__/inference.test.ts"]
+  }
+]

tests/integration/client-typescript/tsconfig.json

@@ -0,0 +1,16 @@
+{
+  "compilerOptions": {
+    "target": "ES2022",
+    "module": "ES2022",
+    "lib": ["ES2022"],
+    "moduleResolution": "bundler",
+    "esModuleInterop": true,
+    "allowSyntheticDefaultImports": true,
+    "strict": true,
+    "skipLibCheck": true,
+    "resolveJsonModule": true,
+    "types": ["jest", "node"]
+  },
+  "include": ["**/*.ts"],
+  "exclude": ["node_modules"]
+}

tests/integration/responses/recordings/1997dc007d202497ce456683d24ddde3553f0db5d5a673146d8bb99c072e77cd.json

@@ -0,0 +1,773 @@
+{
+  "test_id": "tests/integration/responses/test_tool_responses.py::test_max_tool_calls_with_mcp_tools[client_with_models-txt=openai/gpt-4o]",
+  "request": {
+    "method": "POST",
+    "url": "https://api.openai.com/v1/v1/chat/completions",
+    "headers": {},
+    "body": {
+      "model": "gpt-4o",
+      "messages": [
+        {
+          "role": "user",
+          "content": "Get the experiment ID for 'boiling_point' and get the user ID for 'charlie'"
+        }
+      ],
+      "stream": true,
+      "stream_options": {
+        "include_usage": true
+      },
+      "tools": [
+        {
+          "type": "function",
+          "function": {
+            "name": "get_user_id",
+            "description": "\n        Get the user ID for a given username. This ID is needed for other operations.\n\n        :param username: The username to look up\n        :return: The user ID for the username\n        ",
+            "parameters": {
+              "properties": {
+                "username": {
+                  "title": "Username",
+                  "type": "string"
+                }
+              },
+              "required": [
+                "username"
+              ],
+              "title": "get_user_idArguments",
+              "type": "object"
+            }
+          }
+        },
+        {
+          "type": "function",
+          "function": {
+            "name": "get_user_permissions",
+            "description": "\n        Get the permissions for a user ID. Requires a valid user ID from get_user_id.\n\n        :param user_id: The user ID to check permissions for\n        :return: The permissions for the user\n        ",
+            "parameters": {
+              "properties": {
+                "user_id": {
+                  "title": "User Id",
+                  "type": "string"
+                }
+              },
+              "required": [
+                "user_id"
+              ],
+              "title": "get_user_permissionsArguments",
+              "type": "object"
+            }
+          }
+        },
+        {
+          "type": "function",
+          "function": {
+            "name": "check_file_access",
+            "description": "\n        Check if a user can access a specific file. Requires a valid user ID.\n\n        :param user_id: The user ID to check access for\n        :param filename: The filename to check access to\n        :return: Whether the user can access the file (yes/no)\n        ",
+            "parameters": {
+              "properties": {
+                "user_id": {
+                  "title": "User Id",
+                  "type": "string"
+                },
+                "filename": {
+                  "title": "Filename",
+                  "type": "string"
+                }
+              },
+              "required": [
+                "user_id",
+                "filename"
+              ],
+              "title": "check_file_accessArguments",
+              "type": "object"
+            }
+          }
+        },
+        {
+          "type": "function",
+          "function": {
+            "name": "get_experiment_id",
+            "description": "\n        Get the experiment ID for a given experiment name. This ID is needed to get results.\n\n        :param experiment_name: The name of the experiment\n        :return: The experiment ID\n        ",
+            "parameters": {
+              "properties": {
+                "experiment_name": {
+                  "title": "Experiment Name",
+                  "type": "string"
+                }
+              },
+              "required": [
+                "experiment_name"
+              ],
+              "title": "get_experiment_idArguments",
+              "type": "object"
+            }
+          }
+        },
+        {
+          "type": "function",
+          "function": {
+            "name": "get_experiment_results",
+            "description": "\n        Get the results for an experiment ID. Requires a valid experiment ID from get_experiment_id.\n\n        :param experiment_id: The experiment ID to get results for\n        :return: The experiment results\n        ",
+            "parameters": {
+              "properties": {
+                "experiment_id": {
+                  "title": "Experiment Id",
+                  "type": "string"
+                }
+              },
+              "required": [
+                "experiment_id"
+              ],
+              "title": "get_experiment_resultsArguments",
+              "type": "object"
+            }
+          }
+        }
+      ]
+    },
+    "endpoint": "/v1/chat/completions",
+    "model": "gpt-4o"
+  },
+  "response": {
+    "body": [
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": "assistant",
+                "tool_calls": null
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "1V9w3bXnppL"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": "call_y8S7JKR2Qhu4Bh1uxdHRcNDg",
+                    "function": {
+                      "arguments": "",
+                      "name": "get_experiment_id"
+                    },
+                    "type": "function"
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "YEsj"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "{\"ex",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "n"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "perim",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "Q"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "ent_na",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": ""
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "me\":",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "U"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": " \"boi",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": ""
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "ling_p",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": ""
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "oint",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "ha"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "\"}",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "d5D"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": "call_HELkyZOm2fzLx2CeTH3bEcS2",
+                    "function": {
+                      "arguments": "",
+                      "name": "get_user_id"
+                    },
+                    "type": "function"
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "0LbsjDcKz6"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "{\"us",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "c"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "ernam",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "9"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "e\": \"c",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "7C0WFn181I3y3l"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "harl",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "wf"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "ie\"}",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "r"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": null
+              },
+              "finish_reason": "tool_calls",
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "FAci"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-1997dc007d20",
+          "choices": [],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": {
+            "completion_tokens": 51,
+            "prompt_tokens": 393,
+            "total_tokens": 444,
+            "completion_tokens_details": {
+              "accepted_prediction_tokens": 0,
+              "audio_tokens": 0,
+              "reasoning_tokens": 0,
+              "rejected_prediction_tokens": 0
+            },
+            "prompt_tokens_details": {
+              "audio_tokens": 0,
+              "cached_tokens": 0
+            }
+          },
+          "obfuscation": "6xgpRRdKjviPT"
+        }
+      }
+    ],
+    "is_streaming": true
+  },
+  "id_normalization_mapping": {}
+}

tests/integration/responses/recordings/463ab0e2f2914026cfa3c742259c43af318468eb4ef84fd4008ebb40824b7e86.json

@@ -0,0 +1,593 @@
+{
+  "test_id": "tests/integration/responses/test_tool_responses.py::test_max_tool_calls_with_function_tools[openai_client-txt=openai/gpt-4o]",
+  "request": {
+    "method": "POST",
+    "url": "https://api.openai.com/v1/v1/chat/completions",
+    "headers": {},
+    "body": {
+      "model": "gpt-4o",
+      "messages": [
+        {
+          "role": "user",
+          "content": "Can you tell me the weather in Paris and the current time?"
+        }
+      ],
+      "stream": true,
+      "stream_options": {
+        "include_usage": true
+      },
+      "tools": [
+        {
+          "type": "function",
+          "function": {
+            "type": "function",
+            "name": "get_weather",
+            "description": "Get weather information for a specified location",
+            "parameters": {
+              "type": "object",
+              "properties": {
+                "location": {
+                  "type": "string",
+                  "description": "The city name (e.g., 'New York', 'London')"
+                }
+              }
+            },
+            "strict": null
+          }
+        },
+        {
+          "type": "function",
+          "function": {
+            "type": "function",
+            "name": "get_time",
+            "description": "Get current time for a specified location",
+            "parameters": {
+              "type": "object",
+              "properties": {
+                "location": {
+                  "type": "string",
+                  "description": "The city name (e.g., 'New York', 'London')"
+                }
+              }
+            },
+            "strict": null
+          }
+        }
+      ]
+    },
+    "endpoint": "/v1/chat/completions",
+    "model": "gpt-4o"
+  },
+  "response": {
+    "body": [
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": "assistant",
+                "tool_calls": null
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "QmTXstGvpa8"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": "call_HJMoLtHXfCzhlMQOfqIKt0n3",
+                    "function": {
+                      "arguments": "",
+                      "name": "get_weather"
+                    },
+                    "type": "function"
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "iFjmkK23KL"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "{\"lo",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "7"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "catio",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "L"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "n\": \"P",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "THa6gWbrWhVmZ6"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "aris",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "eL"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "\"}",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "jng"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": "call_vGKvTKZM7aALMaUw3Jas7lRg",
+                    "function": {
+                      "arguments": "",
+                      "name": "get_time"
+                    },
+                    "type": "function"
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "LSailgMcgSl54"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "{\"lo",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "z"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "catio",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "4"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "n\": \"P",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "0engr6vRvqXTEP"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "aris",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "Pe"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "\"}",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "LU9"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": null
+              },
+              "finish_reason": "tool_calls",
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": null,
+          "obfuscation": "kD7d"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-463ab0e2f291",
+          "choices": [],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_b1442291a8",
+          "usage": {
+            "completion_tokens": 44,
+            "prompt_tokens": 110,
+            "total_tokens": 154,
+            "completion_tokens_details": {
+              "accepted_prediction_tokens": 0,
+              "audio_tokens": 0,
+              "reasoning_tokens": 0,
+              "rejected_prediction_tokens": 0
+            },
+            "prompt_tokens_details": {
+              "audio_tokens": 0,
+              "cached_tokens": 0
+            }
+          },
+          "obfuscation": "R4ICoxqTqj7ZY"
+        }
+      }
+    ],
+    "is_streaming": true
+  },
+  "id_normalization_mapping": {}
+}

tests/integration/responses/recordings/b218af7fa0663e60b12633f54cfddbcf60a1fedd85c501850b9f7e759443809f.json

@@ -0,0 +1,773 @@
+{
+  "test_id": "tests/integration/responses/test_tool_responses.py::test_max_tool_calls_with_mcp_tools[openai_client-txt=openai/gpt-4o]",
+  "request": {
+    "method": "POST",
+    "url": "https://api.openai.com/v1/v1/chat/completions",
+    "headers": {},
+    "body": {
+      "model": "gpt-4o",
+      "messages": [
+        {
+          "role": "user",
+          "content": "Get the experiment ID for 'boiling_point' and get the user ID for 'charlie'"
+        }
+      ],
+      "stream": true,
+      "stream_options": {
+        "include_usage": true
+      },
+      "tools": [
+        {
+          "type": "function",
+          "function": {
+            "name": "get_user_id",
+            "description": "\n        Get the user ID for a given username. This ID is needed for other operations.\n\n        :param username: The username to look up\n        :return: The user ID for the username\n        ",
+            "parameters": {
+              "properties": {
+                "username": {
+                  "title": "Username",
+                  "type": "string"
+                }
+              },
+              "required": [
+                "username"
+              ],
+              "title": "get_user_idArguments",
+              "type": "object"
+            }
+          }
+        },
+        {
+          "type": "function",
+          "function": {
+            "name": "get_user_permissions",
+            "description": "\n        Get the permissions for a user ID. Requires a valid user ID from get_user_id.\n\n        :param user_id: The user ID to check permissions for\n        :return: The permissions for the user\n        ",
+            "parameters": {
+              "properties": {
+                "user_id": {
+                  "title": "User Id",
+                  "type": "string"
+                }
+              },
+              "required": [
+                "user_id"
+              ],
+              "title": "get_user_permissionsArguments",
+              "type": "object"
+            }
+          }
+        },
+        {
+          "type": "function",
+          "function": {
+            "name": "check_file_access",
+            "description": "\n        Check if a user can access a specific file. Requires a valid user ID.\n\n        :param user_id: The user ID to check access for\n        :param filename: The filename to check access to\n        :return: Whether the user can access the file (yes/no)\n        ",
+            "parameters": {
+              "properties": {
+                "user_id": {
+                  "title": "User Id",
+                  "type": "string"
+                },
+                "filename": {
+                  "title": "Filename",
+                  "type": "string"
+                }
+              },
+              "required": [
+                "user_id",
+                "filename"
+              ],
+              "title": "check_file_accessArguments",
+              "type": "object"
+            }
+          }
+        },
+        {
+          "type": "function",
+          "function": {
+            "name": "get_experiment_id",
+            "description": "\n        Get the experiment ID for a given experiment name. This ID is needed to get results.\n\n        :param experiment_name: The name of the experiment\n        :return: The experiment ID\n        ",
+            "parameters": {
+              "properties": {
+                "experiment_name": {
+                  "title": "Experiment Name",
+                  "type": "string"
+                }
+              },
+              "required": [
+                "experiment_name"
+              ],
+              "title": "get_experiment_idArguments",
+              "type": "object"
+            }
+          }
+        },
+        {
+          "type": "function",
+          "function": {
+            "name": "get_experiment_results",
+            "description": "\n        Get the results for an experiment ID. Requires a valid experiment ID from get_experiment_id.\n\n        :param experiment_id: The experiment ID to get results for\n        :return: The experiment results\n        ",
+            "parameters": {
+              "properties": {
+                "experiment_id": {
+                  "title": "Experiment Id",
+                  "type": "string"
+                }
+              },
+              "required": [
+                "experiment_id"
+              ],
+              "title": "get_experiment_resultsArguments",
+              "type": "object"
+            }
+          }
+        }
+      ]
+    },
+    "endpoint": "/v1/chat/completions",
+    "model": "gpt-4o"
+  },
+  "response": {
+    "body": [
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": "assistant",
+                "tool_calls": null
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "N5OTLR9CfmU"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": "call_z8P1RQv54BLxyMlRdMFkcCGd",
+                    "function": {
+                      "arguments": "",
+                      "name": "get_experiment_id"
+                    },
+                    "type": "function"
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "3EKK"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "{\"ex",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "R"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "perim",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "Q"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "ent_na",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": ""
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "me\":",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "6"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": " \"boi",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": ""
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "ling_p",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": ""
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "oint",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "pw"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 0,
+                    "id": null,
+                    "function": {
+                      "arguments": "\"}",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "Gfk"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": "call_I5tcLgyMADoVwLKDj9HkTCs5",
+                    "function": {
+                      "arguments": "",
+                      "name": "get_user_id"
+                    },
+                    "type": "function"
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "Yp7IueDs5V"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "{\"us",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "8"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "ernam",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "X"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "e\": \"c",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "2oif8BwVnTCnAF"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "harl",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "hv"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": [
+                  {
+                    "index": 1,
+                    "id": null,
+                    "function": {
+                      "arguments": "ie\"}",
+                      "name": null
+                    },
+                    "type": null
+                  }
+                ]
+              },
+              "finish_reason": null,
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "C"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [
+            {
+              "delta": {
+                "content": null,
+                "function_call": null,
+                "refusal": null,
+                "role": null,
+                "tool_calls": null
+              },
+              "finish_reason": "tool_calls",
+              "index": 0,
+              "logprobs": null
+            }
+          ],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": null,
+          "obfuscation": "ctjO"
+        }
+      },
+      {
+        "__type__": "openai.types.chat.chat_completion_chunk.ChatCompletionChunk",
+        "__data__": {
+          "id": "rec-b218af7fa066",
+          "choices": [],
+          "created": 0,
+          "model": "gpt-4o-2024-08-06",
+          "object": "chat.completion.chunk",
+          "service_tier": "default",
+          "system_fingerprint": "fp_c98e05ca17",
+          "usage": {
+            "completion_tokens": 51,
+            "prompt_tokens": 393,
+            "total_tokens": 444,
+            "completion_tokens_details": {
+              "accepted_prediction_tokens": 0,
+              "audio_tokens": 0,
+              "reasoning_tokens": 0,
+              "rejected_prediction_tokens": 0
+            },
+            "prompt_tokens_details": {
+              "audio_tokens": 0,
+              "cached_tokens": 0
+            }
+          },
+          "obfuscation": "fclbZeBSSKN4C"
+        }
+      }
+    ],
+    "is_streaming": true
+  },
+  "id_normalization_mapping": {}
+}