chore(yaml)!: move registered resources to a sub-key (#3861)

**NOTE: this is a backwards incompatible change to the run-configs.**

A small QOL update, but this will prove useful when I do a rename for
"vector_dbs" to "vector_stores" next.

Moves all the `models, shields, ...` keys in run-config under a
`registered_resources` sub-key.

llama_stack/core/datatypes.py

@@ -422,6 +422,18 @@ def process_cors_config(cors_config: bool | CORSConfig | None) -> CORSConfig | N
     raise ValueError(f"Expected bool or CORSConfig, got {type(cors_config).__name__}")
 
 
+class RegisteredResources(BaseModel):
+    """Registry of resources available in the distribution."""
+
+    models: list[ModelInput] = Field(default_factory=list)
+    shields: list[ShieldInput] = Field(default_factory=list)
+    vector_dbs: list[VectorDBInput] = Field(default_factory=list)
+    datasets: list[DatasetInput] = Field(default_factory=list)
+    scoring_fns: list[ScoringFnInput] = Field(default_factory=list)
+    benchmarks: list[BenchmarkInput] = Field(default_factory=list)
+    tool_groups: list[ToolGroupInput] = Field(default_factory=list)
+
+
 class ServerConfig(BaseModel):
     port: int = Field(
         default=8321,
@@ -491,14 +503,10 @@ can be instantiated multiple times (with different configs) if necessary.
         description="Catalog of named storage backends and references available to the stack",
     )
 
-    # registry of "resources" in the distribution
+    registered_resources: RegisteredResources = Field(
-    models: list[ModelInput] = Field(default_factory=list)
+        default_factory=RegisteredResources,
-    shields: list[ShieldInput] = Field(default_factory=list)
+        description="Registry of resources available in the distribution",
-    vector_dbs: list[VectorDBInput] = Field(default_factory=list)
+    )
-    datasets: list[DatasetInput] = Field(default_factory=list)
-    scoring_fns: list[ScoringFnInput] = Field(default_factory=list)
-    benchmarks: list[BenchmarkInput] = Field(default_factory=list)
-    tool_groups: list[ToolGroupInput] = Field(default_factory=list)
 
     logging: LoggingConfig | None = Field(default=None, description="Configuration for Llama Stack Logging")
 

llama_stack/core/stack.py

@@ -110,7 +110,7 @@ TEST_RECORDING_CONTEXT = None
 
 async def register_resources(run_config: StackRunConfig, impls: dict[Api, Any]):
     for rsrc, api, register_method, list_method in RESOURCES:
-        objects = getattr(run_config, rsrc)
+        objects = getattr(run_config.registered_resources, rsrc)
         if api not in impls:
             continue
 

llama_stack/distributions/ci-tests/run.yaml

@@ -247,22 +247,23 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
-models: []
+registered_resources:
-shields:
+  models: []
-- shield_id: llama-guard
+  shields:
+  - shield_id: llama-guard
     provider_id: ${env.SAFETY_MODEL:+llama-guard}
     provider_shield_id: ${env.SAFETY_MODEL:=}
-- shield_id: code-scanner
+  - shield_id: code-scanner
     provider_id: ${env.CODE_SCANNER_MODEL:+code-scanner}
     provider_shield_id: ${env.CODE_SCANNER_MODEL:=}
-vector_dbs: []
+  vector_dbs: []
-datasets: []
+  datasets: []
-scoring_fns: []
+  scoring_fns: []
-benchmarks: []
+  benchmarks: []
-tool_groups:
+  tool_groups:
-- toolgroup_id: builtin::websearch
+  - toolgroup_id: builtin::websearch
     provider_id: tavily-search
-- toolgroup_id: builtin::rag
+  - toolgroup_id: builtin::rag
     provider_id: rag-runtime
 server:
   port: 8321

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

@@ -109,30 +109,31 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
-models:
+registered_resources:
-- metadata: {}
+  models:
+  - metadata: {}
     model_id: ${env.INFERENCE_MODEL}
     provider_id: tgi0
     model_type: llm
-- metadata: {}
+  - metadata: {}
     model_id: ${env.SAFETY_MODEL}
     provider_id: tgi1
     model_type: llm
-- metadata:
+  - metadata:
       embedding_dimension: 768
     model_id: nomic-embed-text-v1.5
     provider_id: sentence-transformers
     model_type: embedding
-shields:
+  shields:
-- shield_id: ${env.SAFETY_MODEL}
+  - shield_id: ${env.SAFETY_MODEL}
-vector_dbs: []
+  vector_dbs: []
-datasets: []
+  datasets: []
-scoring_fns: []
+  scoring_fns: []
-benchmarks: []
+  benchmarks: []
-tool_groups:
+  tool_groups:
-- toolgroup_id: builtin::websearch
+  - toolgroup_id: builtin::websearch
     provider_id: brave-search
-- toolgroup_id: builtin::rag
+  - toolgroup_id: builtin::rag
     provider_id: rag-runtime
 server:
   port: 8321

llama_stack/distributions/dell/run.yaml

@@ -105,25 +105,26 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
-models:
+registered_resources:
-- metadata: {}
+  models:
+  - metadata: {}
     model_id: ${env.INFERENCE_MODEL}
     provider_id: tgi0
     model_type: llm
-- metadata:
+  - metadata:
       embedding_dimension: 768
     model_id: nomic-embed-text-v1.5
     provider_id: sentence-transformers
     model_type: embedding
-shields: []
+  shields: []
-vector_dbs: []
+  vector_dbs: []
-datasets: []
+  datasets: []
-scoring_fns: []
+  scoring_fns: []
-benchmarks: []
+  benchmarks: []
-tool_groups:
+  tool_groups:
-- toolgroup_id: builtin::websearch
+  - toolgroup_id: builtin::websearch
     provider_id: brave-search
-- toolgroup_id: builtin::rag
+  - toolgroup_id: builtin::rag
     provider_id: rag-runtime
 server:
   port: 8321

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

@@ -122,30 +122,31 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
-models:
+registered_resources:
-- metadata: {}
+  models:
+  - metadata: {}
     model_id: ${env.INFERENCE_MODEL}
     provider_id: meta-reference-inference
     model_type: llm
-- metadata: {}
+  - metadata: {}
     model_id: ${env.SAFETY_MODEL}
     provider_id: meta-reference-safety
     model_type: llm
-- metadata:
+  - metadata:
       embedding_dimension: 768
     model_id: nomic-embed-text-v1.5
     provider_id: sentence-transformers
     model_type: embedding
-shields:
+  shields:
-- shield_id: ${env.SAFETY_MODEL}
+  - shield_id: ${env.SAFETY_MODEL}
-vector_dbs: []
+  vector_dbs: []
-datasets: []
+  datasets: []
-scoring_fns: []
+  scoring_fns: []
-benchmarks: []
+  benchmarks: []
-tool_groups:
+  tool_groups:
-- toolgroup_id: builtin::websearch
+  - toolgroup_id: builtin::websearch
     provider_id: tavily-search
-- toolgroup_id: builtin::rag
+  - toolgroup_id: builtin::rag
     provider_id: rag-runtime
 server:
   port: 8321

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

@@ -112,25 +112,26 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
-models:
+registered_resources:
-- metadata: {}
+  models:
+  - metadata: {}
     model_id: ${env.INFERENCE_MODEL}
     provider_id: meta-reference-inference
     model_type: llm
-- metadata:
+  - metadata:
       embedding_dimension: 768
     model_id: nomic-embed-text-v1.5
     provider_id: sentence-transformers
     model_type: embedding
-shields: []
+  shields: []
-vector_dbs: []
+  vector_dbs: []
-datasets: []
+  datasets: []
-scoring_fns: []
+  scoring_fns: []
-benchmarks: []
+  benchmarks: []
-tool_groups:
+  tool_groups:
-- toolgroup_id: builtin::websearch
+  - toolgroup_id: builtin::websearch
     provider_id: tavily-search
-- toolgroup_id: builtin::rag
+  - toolgroup_id: builtin::rag
     provider_id: rag-runtime
 server:
   port: 8321

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

@@ -111,24 +111,25 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
-models:
+registered_resources:
-- metadata: {}
+  models:
+  - metadata: {}
     model_id: ${env.INFERENCE_MODEL}
     provider_id: nvidia
     model_type: llm
-- metadata: {}
+  - metadata: {}
     model_id: ${env.SAFETY_MODEL}
     provider_id: nvidia
     model_type: llm
-shields:
+  shields:
-- shield_id: ${env.SAFETY_MODEL}
+  - shield_id: ${env.SAFETY_MODEL}
     provider_id: nvidia
-vector_dbs: []
+  vector_dbs: []
-datasets: []
+  datasets: []
-scoring_fns: []
+  scoring_fns: []
-benchmarks: []
+  benchmarks: []
-tool_groups:
+  tool_groups:
-- toolgroup_id: builtin::rag
+  - toolgroup_id: builtin::rag
     provider_id: rag-runtime
 server:
   port: 8321

llama_stack/distributions/nvidia/run.yaml

@@ -100,14 +100,15 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
-models: []
+registered_resources:
-shields: []
+  models: []
-vector_dbs: []
+  shields: []
-datasets: []
+  vector_dbs: []
-scoring_fns: []
+  datasets: []
-benchmarks: []
+  scoring_fns: []
-tool_groups:
+  benchmarks: []
-- toolgroup_id: builtin::rag
+  tool_groups:
+  - toolgroup_id: builtin::rag
     provider_id: rag-runtime
 server:
   port: 8321

llama_stack/distributions/open-benchmark/run.yaml

@@ -142,108 +142,109 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
-models:
+registered_resources:
-- metadata: {}
+  models:
+  - metadata: {}
     model_id: gpt-4o
     provider_id: openai
     provider_model_id: gpt-4o
     model_type: llm
-- metadata: {}
+  - metadata: {}
     model_id: claude-3-5-sonnet-latest
     provider_id: anthropic
     provider_model_id: claude-3-5-sonnet-latest
     model_type: llm
-- metadata: {}
+  - metadata: {}
     model_id: gemini/gemini-1.5-flash
     provider_id: gemini
     provider_model_id: gemini/gemini-1.5-flash
     model_type: llm
-- metadata: {}
+  - metadata: {}
     model_id: meta-llama/Llama-3.3-70B-Instruct
     provider_id: groq
     provider_model_id: groq/llama-3.3-70b-versatile
     model_type: llm
-- metadata: {}
+  - metadata: {}
     model_id: meta-llama/Llama-3.1-405B-Instruct
     provider_id: together
     provider_model_id: meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo
     model_type: llm
-shields:
+  shields:
-- shield_id: meta-llama/Llama-Guard-3-8B
+  - shield_id: meta-llama/Llama-Guard-3-8B
-vector_dbs: []
+  vector_dbs: []
-datasets:
+  datasets:
-- purpose: eval/messages-answer
+  - purpose: eval/messages-answer
     source:
       type: uri
       uri: huggingface://datasets/llamastack/simpleqa?split=train
     metadata: {}
     dataset_id: simpleqa
-- purpose: eval/messages-answer
+  - purpose: eval/messages-answer
     source:
       type: uri
       uri: huggingface://datasets/llamastack/mmlu_cot?split=test&name=all
     metadata: {}
     dataset_id: mmlu_cot
-- purpose: eval/messages-answer
+  - purpose: eval/messages-answer
     source:
       type: uri
       uri: huggingface://datasets/llamastack/gpqa_0shot_cot?split=test&name=gpqa_main
     metadata: {}
     dataset_id: gpqa_cot
-- purpose: eval/messages-answer
+  - purpose: eval/messages-answer
     source:
       type: uri
       uri: huggingface://datasets/llamastack/math_500?split=test
     metadata: {}
     dataset_id: math_500
-- purpose: eval/messages-answer
+  - purpose: eval/messages-answer
     source:
       type: uri
       uri: huggingface://datasets/llamastack/IfEval?split=train
     metadata: {}
     dataset_id: ifeval
-- purpose: eval/messages-answer
+  - purpose: eval/messages-answer
     source:
       type: uri
       uri: huggingface://datasets/llamastack/docvqa?split=val
     metadata: {}
     dataset_id: docvqa
-scoring_fns: []
+  scoring_fns: []
-benchmarks:
+  benchmarks:
-- dataset_id: simpleqa
+  - dataset_id: simpleqa
     scoring_functions:
     - llm-as-judge::405b-simpleqa
     metadata: {}
     benchmark_id: meta-reference-simpleqa
-- dataset_id: mmlu_cot
+  - dataset_id: mmlu_cot
     scoring_functions:
     - basic::regex_parser_multiple_choice_answer
     metadata: {}
     benchmark_id: meta-reference-mmlu-cot
-- dataset_id: gpqa_cot
+  - dataset_id: gpqa_cot
     scoring_functions:
     - basic::regex_parser_multiple_choice_answer
     metadata: {}
     benchmark_id: meta-reference-gpqa-cot
-- dataset_id: math_500
+  - dataset_id: math_500
     scoring_functions:
     - basic::regex_parser_math_response
     metadata: {}
     benchmark_id: meta-reference-math-500
-- dataset_id: ifeval
+  - dataset_id: ifeval
     scoring_functions:
     - basic::ifeval
     metadata: {}
     benchmark_id: meta-reference-ifeval
-- dataset_id: docvqa
+  - dataset_id: docvqa
     scoring_functions:
     - basic::docvqa
     metadata: {}
     benchmark_id: meta-reference-docvqa
-tool_groups:
+  tool_groups:
-- toolgroup_id: builtin::websearch
+  - toolgroup_id: builtin::websearch
     provider_id: tavily-search
-- toolgroup_id: builtin::rag
+  - toolgroup_id: builtin::rag
     provider_id: rag-runtime
 server:
   port: 8321

llama_stack/distributions/postgres-demo/run.yaml

@@ -87,26 +87,27 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
-models:
+registered_resources:
-- metadata: {}
+  models:
+  - metadata: {}
     model_id: ${env.INFERENCE_MODEL}
     provider_id: vllm-inference
     model_type: llm
-- metadata:
+  - metadata:
       embedding_dimension: 768
     model_id: nomic-embed-text-v1.5
     provider_id: sentence-transformers
     model_type: embedding
-shields:
+  shields:
-- shield_id: meta-llama/Llama-Guard-3-8B
+  - shield_id: meta-llama/Llama-Guard-3-8B
-vector_dbs: []
+  vector_dbs: []
-datasets: []
+  datasets: []
-scoring_fns: []
+  scoring_fns: []
-benchmarks: []
+  benchmarks: []
-tool_groups:
+  tool_groups:
-- toolgroup_id: builtin::websearch
+  - toolgroup_id: builtin::websearch
     provider_id: tavily-search
-- toolgroup_id: builtin::rag
+  - toolgroup_id: builtin::rag
     provider_id: rag-runtime
 server:
   port: 8321

llama_stack/distributions/starter-gpu/run.yaml

@@ -250,22 +250,23 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
-models: []
+registered_resources:
-shields:
+  models: []
-- shield_id: llama-guard
+  shields:
+  - shield_id: llama-guard
     provider_id: ${env.SAFETY_MODEL:+llama-guard}
     provider_shield_id: ${env.SAFETY_MODEL:=}
-- shield_id: code-scanner
+  - shield_id: code-scanner
     provider_id: ${env.CODE_SCANNER_MODEL:+code-scanner}
     provider_shield_id: ${env.CODE_SCANNER_MODEL:=}
-vector_dbs: []
+  vector_dbs: []
-datasets: []
+  datasets: []
-scoring_fns: []
+  scoring_fns: []
-benchmarks: []
+  benchmarks: []
-tool_groups:
+  tool_groups:
-- toolgroup_id: builtin::websearch
+  - toolgroup_id: builtin::websearch
     provider_id: tavily-search
-- toolgroup_id: builtin::rag
+  - toolgroup_id: builtin::rag
     provider_id: rag-runtime
 server:
   port: 8321

llama_stack/distributions/starter/run.yaml

@@ -247,22 +247,23 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
-models: []
+registered_resources:
-shields:
+  models: []
-- shield_id: llama-guard
+  shields:
+  - shield_id: llama-guard
     provider_id: ${env.SAFETY_MODEL:+llama-guard}
     provider_shield_id: ${env.SAFETY_MODEL:=}
-- shield_id: code-scanner
+  - shield_id: code-scanner
     provider_id: ${env.CODE_SCANNER_MODEL:+code-scanner}
     provider_shield_id: ${env.CODE_SCANNER_MODEL:=}
-vector_dbs: []
+  vector_dbs: []
-datasets: []
+  datasets: []
-scoring_fns: []
+  scoring_fns: []
-benchmarks: []
+  benchmarks: []
-tool_groups:
+  tool_groups:
-- toolgroup_id: builtin::websearch
+  - toolgroup_id: builtin::websearch
     provider_id: tavily-search
-- toolgroup_id: builtin::rag
+  - toolgroup_id: builtin::rag
     provider_id: rag-runtime
 server:
   port: 8321

llama_stack/distributions/template.py

@@ -272,6 +272,7 @@ class RunConfigSettings(BaseModel):
             "apis": apis,
             "providers": provider_configs,
             "storage": storage_config,
+            "registered_resources": {
                 "models": [m.model_dump(exclude_none=True) for m in (self.default_models or [])],
                 "shields": [s.model_dump(exclude_none=True) for s in (self.default_shields or [])],
                 "vector_dbs": [],
@@ -279,6 +280,7 @@ class RunConfigSettings(BaseModel):
                 "scoring_fns": [],
                 "benchmarks": [b.model_dump(exclude_none=True) for b in (self.default_benchmarks or [])],
                 "tool_groups": [t.model_dump(exclude_none=True) for t in (self.default_tool_groups or [])],
+            },
             "server": {
                 "port": 8321,
             },

llama_stack/distributions/watsonx/run.yaml

@@ -115,16 +115,17 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
-models: []
+registered_resources:
-shields: []
+  models: []
-vector_dbs: []
+  shields: []
-datasets: []
+  vector_dbs: []
-scoring_fns: []
+  datasets: []
-benchmarks: []
+  scoring_fns: []
-tool_groups:
+  benchmarks: []
-- toolgroup_id: builtin::websearch
+  tool_groups:
+  - toolgroup_id: builtin::websearch
     provider_id: tavily-search
-- toolgroup_id: builtin::rag
+  - toolgroup_id: builtin::rag
     provider_id: rag-runtime
 server:
   port: 8321