feat: Add metadata field to request and response

This changes adds Optional metadata field to OpenAI compatible request and
response object.

fixes: #3564
Signed-off-by: Abhishek Bongale <abhishekbongale@outlook.com>

client-sdks/stainless/openapi.yml

@@ -6796,6 +6796,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
         input:
           items:
             anyOf:
@@ -7199,6 +7205,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
       type: object
       required:
       - input
@@ -7330,6 +7342,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
       type: object
       required:
       - created_at

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

@@ -3639,6 +3639,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
         input:
           items:
             anyOf:
@@ -4042,6 +4048,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
       type: object
       required:
       - input
@@ -4173,6 +4185,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
       type: object
       required:
       - created_at

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

@@ -3336,6 +3336,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
         input:
           items:
             anyOf:
@@ -3736,6 +3742,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
       type: object
       required:
       - created_at

docs/static/llama-stack-spec.yaml

@@ -5817,6 +5817,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
         input:
           items:
             anyOf:
@@ -6220,6 +6226,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
       type: object
       required:
       - input
@@ -6351,6 +6363,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
       type: object
       required:
       - created_at

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

@@ -6796,6 +6796,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
         input:
           items:
             anyOf:
@@ -7199,6 +7205,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
       type: object
       required:
       - input
@@ -7330,6 +7342,12 @@ components:
           anyOf:
           - type: integer
           - type: 'null'
+        metadata:
+          anyOf:
+          - additionalProperties:
+              type: string
+            type: object
+          - type: 'null'
       type: object
       required:
       - created_at

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

@@ -111,6 +111,7 @@ class MetaReferenceAgentsImpl(Agents):
         max_infer_iters: int | None = 10,
         guardrails: list[ResponseGuardrail] | None = None,
         max_tool_calls: int | None = None,
+        metadata: dict[str, str] | None = None,
     ) -> OpenAIResponseObject:
         assert self.openai_responses_impl is not None, "OpenAI responses not initialized"
         result = await self.openai_responses_impl.create_openai_response(
@@ -130,6 +131,7 @@ class MetaReferenceAgentsImpl(Agents):
             guardrails,
             parallel_tool_calls,
             max_tool_calls,
+            metadata,
         )
         return result  # type: ignore[no-any-return]
 

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

@@ -336,6 +336,7 @@ class OpenAIResponsesImpl:
         guardrails: list[str | ResponseGuardrailSpec] | None = None,
         parallel_tool_calls: bool | None = None,
         max_tool_calls: int | None = None,
+        metadata: dict[str, str] | None = None,
     ):
         stream = bool(stream)
         text = OpenAIResponseText(format=OpenAIResponseTextFormat(type="text")) if text is None else text
@@ -390,6 +391,7 @@ class OpenAIResponsesImpl:
             guardrail_ids=guardrail_ids,
             parallel_tool_calls=parallel_tool_calls,
             max_tool_calls=max_tool_calls,
+            metadata=metadata,
         )
 
         if stream:
@@ -442,6 +444,7 @@ class OpenAIResponsesImpl:
         guardrail_ids: list[str] | None = None,
         parallel_tool_calls: bool | None = True,
         max_tool_calls: int | None = None,
+        metadata: dict[str, str] | None = None,
     ) -> AsyncIterator[OpenAIResponseObjectStream]:
         # These should never be None when called from create_openai_response (which sets defaults)
         # but we assert here to help mypy understand the types
@@ -490,6 +493,7 @@ class OpenAIResponsesImpl:
             guardrail_ids=guardrail_ids,
             instructions=instructions,
             max_tool_calls=max_tool_calls,
+            metadata=metadata,
         )
 
         # Stream the response

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

@@ -118,6 +118,7 @@ class StreamingResponseOrchestrator:
         prompt: OpenAIResponsePrompt | None = None,
         parallel_tool_calls: bool | None = None,
         max_tool_calls: int | None = None,
+        metadata: dict[str, str] | None = None,
     ):
         self.inference_api = inference_api
         self.ctx = ctx
@@ -135,6 +136,7 @@ class StreamingResponseOrchestrator:
         self.parallel_tool_calls = parallel_tool_calls
         # Max number of total calls to built-in tools that can be processed in a response
         self.max_tool_calls = max_tool_calls
+        self.metadata = metadata
         self.sequence_number = 0
         # Store MCP tool mapping that gets built during tool processing
         self.mcp_tool_to_server: dict[str, OpenAIResponseInputToolMCP] = (
@@ -162,6 +164,7 @@ class StreamingResponseOrchestrator:
             model=self.ctx.model,
             status="completed",
             output=[OpenAIResponseMessage(role="assistant", content=[refusal_content], type="message")],
+            metadata=self.metadata,
         )
 
         return OpenAIResponseObjectStreamResponseCompleted(response=refusal_response)
@@ -197,6 +200,7 @@ class StreamingResponseOrchestrator:
             prompt=self.prompt,
             parallel_tool_calls=self.parallel_tool_calls,
             max_tool_calls=self.max_tool_calls,
+            metadata=self.metadata,
         )
 
     async def create_response(self) -> AsyncIterator[OpenAIResponseObjectStream]:

src/llama_stack_api/agents.py

@@ -89,6 +89,7 @@ class Agents(Protocol):
             ),
         ] = None,
         max_tool_calls: int | None = None,
+        metadata: dict[str, str] | None = None,
     ) -> OpenAIResponseObject | AsyncIterator[OpenAIResponseObjectStream]:
         """Create a model response.
 
@@ -100,6 +101,7 @@ class Agents(Protocol):
         :param include: (Optional) Additional fields to include in the response.
         :param guardrails: (Optional) List of guardrails to apply during response generation. Can be guardrail IDs (strings) or guardrail specifications.
         :param max_tool_calls: (Optional) Max number of total calls to built-in tools that can be processed in a response.
+        :param metadata: (Optional) Dictionary of metadata key-value pairs to attach to the response.
         :returns: An OpenAIResponseObject.
         """
         ...

src/llama_stack_api/openai_responses.py

@@ -597,6 +597,7 @@ class OpenAIResponseObject(BaseModel):
     :param usage: (Optional) Token usage information for the response
     :param instructions: (Optional) System message inserted into the model's context
     :param max_tool_calls: (Optional) Max number of total calls to built-in tools that can be processed in a response
+    :param metadata: (Optional) Dictionary of metadata key-value pairs
     """
 
     created_at: int
@@ -619,6 +620,7 @@ class OpenAIResponseObject(BaseModel):
     usage: OpenAIResponseUsage | None = None
     instructions: str | None = None
     max_tool_calls: int | None = None
+    metadata: dict[str, str] | None = None
 
 
 @json_schema_type