fix: Responses API previous_response input items

This adds storing of input items with previous responses and then
restores those input items to prepend to the user's messages list when
using conversation state.

I missed this in the initial implementation, but it makes sense that we
have to store the input items from previous responses so that we can
reconstruct the proper messages stack for multi-turn conversations -
just the output from previous responses isn't enough context for the
models to follow the turns and the original instructions.

Signed-off-by: Ben Browning <bbrownin@redhat.com>

llama_stack/apis/agents/openai_responses.py

@@ -131,3 +131,20 @@ OpenAIResponseInputTool = Annotated[
     Field(discriminator="type"),
 ]
 register_schema(OpenAIResponseInputTool, name="OpenAIResponseInputTool")
+
+
+@json_schema_type
+class OpenAIResponseInputItemMessage(OpenAIResponseInputMessage):
+    id: str
+
+
+@json_schema_type
+class OpenAIResponseInputItemList(BaseModel):
+    data: list[OpenAIResponseInputItemMessage]
+    object: Literal["list"] = "list"
+
+
+@json_schema_type
+class OpenAIResponsePreviousResponseWithInputItems(BaseModel):
+    input_items: OpenAIResponseInputItemList
+    response: OpenAIResponseObject

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

@@ -12,7 +12,10 @@ from typing import cast
 from openai.types.chat import ChatCompletionToolParam
 
 from llama_stack.apis.agents.openai_responses import (
+    OpenAIResponseInputItemList,
+    OpenAIResponseInputItemMessage,
     OpenAIResponseInputMessage,
+    OpenAIResponseInputMessageContent,
     OpenAIResponseInputMessageContentImage,
     OpenAIResponseInputMessageContentText,
     OpenAIResponseInputTool,
@@ -24,6 +27,7 @@ from llama_stack.apis.agents.openai_responses import (
     OpenAIResponseOutputMessage,
     OpenAIResponseOutputMessageContentOutputText,
     OpenAIResponseOutputMessageWebSearchToolCall,
+    OpenAIResponsePreviousResponseWithInputItems,
 )
 from llama_stack.apis.inference.inference import (
     Inference,
@@ -50,9 +54,56 @@ logger = get_logger(name=__name__, category="openai_responses")
 OPENAI_RESPONSES_PREFIX = "openai_responses:"
 
 
-async def _previous_response_to_messages(previous_response: OpenAIResponseObject) -> list[OpenAIMessageParam]:
+async def _convert_response_input_content_to_chat_content_parts(
+    input_content: list[OpenAIResponseInputMessageContent],
+) -> list[OpenAIChatCompletionContentPartParam]:
+    """
+    Convert a list of input content items to a list of chat completion content parts
+    """
+    content_parts = []
+    for input_content_part in input_content:
+        if isinstance(input_content_part, OpenAIResponseInputMessageContentText):
+            content_parts.append(OpenAIChatCompletionContentPartTextParam(text=input_content_part.text))
+        elif isinstance(input_content_part, OpenAIResponseInputMessageContentImage):
+            if input_content_part.image_url:
+                image_url = OpenAIImageURL(url=input_content_part.image_url, detail=input_content_part.detail)
+                content_parts.append(OpenAIChatCompletionContentPartImageParam(image_url=image_url))
+    return content_parts
+
+
+async def _convert_response_input_to_chat_user_content(
+    input: str | list[OpenAIResponseInputMessage],
+) -> str | list[OpenAIChatCompletionContentPartParam]:
+    user_content: str | list[OpenAIChatCompletionContentPartParam] = ""
+    if isinstance(input, list):
+        user_content = []
+        for user_input in input:
+            if isinstance(user_input.content, list):
+                user_content.extend(await _convert_response_input_content_to_chat_content_parts(user_input.content))
+            else:
+                user_content.append(OpenAIChatCompletionContentPartTextParam(text=user_input.content))
+    else:
+        user_content = input
+    return user_content
+
+
+async def _previous_response_to_messages(
+    previous_response: OpenAIResponsePreviousResponseWithInputItems,
+) -> list[OpenAIMessageParam]:
     messages: list[OpenAIMessageParam] = []
-    for output_message in previous_response.output:
+    for previous_message in previous_response.input_items.data:
+        previous_content = await _convert_response_input_content_to_chat_content_parts(previous_message.content)
+        if previous_message.role == "user":
+            converted_message = OpenAIUserMessageParam(content=previous_content)
+        elif previous_message.role == "assistant":
+            converted_message = OpenAIAssistantMessageParam(content=previous_content)
+        else:
+            # TODO: handle other message roles? unclear if system/developer roles are
+            # used in previous responses
+            continue
+        messages.append(converted_message)
+
+    for output_message in previous_response.response.output:
         if isinstance(output_message, OpenAIResponseOutputMessage):
             messages.append(OpenAIAssistantMessageParam(content=output_message.content[0].text))
     return messages
@@ -90,15 +141,19 @@ class OpenAIResponsesImpl:
         self.tool_groups_api = tool_groups_api
         self.tool_runtime_api = tool_runtime_api
 
-    async def get_openai_response(
+    async def _get_previous_response_with_input(self, id: str) -> OpenAIResponsePreviousResponseWithInputItems:
-        self,
-        id: str,
-    ) -> OpenAIResponseObject:
         key = f"{OPENAI_RESPONSES_PREFIX}{id}"
         response_json = await self.persistence_store.get(key=key)
         if response_json is None:
             raise ValueError(f"OpenAI response with id '{id}' not found")
-        return OpenAIResponseObject.model_validate_json(response_json)
+        return OpenAIResponsePreviousResponseWithInputItems.model_validate_json(response_json)
+
+    async def get_openai_response(
+        self,
+        id: str,
+    ) -> OpenAIResponseObject:
+        response_with_input = await self._get_previous_response_with_input(id)
+        return response_with_input.response
 
     async def create_openai_response(
         self,
@@ -114,27 +169,10 @@ class OpenAIResponsesImpl:
 
         messages: list[OpenAIMessageParam] = []
         if previous_response_id:
-            previous_response = await self.get_openai_response(previous_response_id)
+            previous_response_with_input = await self._get_previous_response_with_input(previous_response_id)
-            messages.extend(await _previous_response_to_messages(previous_response))
+            messages.extend(await _previous_response_to_messages(previous_response_with_input))
-        # TODO: refactor this user_content parsing out into a separate method
+
-        user_content: str | list[OpenAIChatCompletionContentPartParam] = ""
+        user_content = await _convert_response_input_to_chat_user_content(input)
-        if isinstance(input, list):
-            user_content = []
-            for user_input in input:
-                if isinstance(user_input.content, list):
-                    for user_input_content in user_input.content:
-                        if isinstance(user_input_content, OpenAIResponseInputMessageContentText):
-                            user_content.append(OpenAIChatCompletionContentPartTextParam(text=user_input_content.text))
-                        elif isinstance(user_input_content, OpenAIResponseInputMessageContentImage):
-                            if user_input_content.image_url:
-                                image_url = OpenAIImageURL(
-                                    url=user_input_content.image_url, detail=user_input_content.detail
-                                )
-                                user_content.append(OpenAIChatCompletionContentPartImageParam(image_url=image_url))
-                else:
-                    user_content.append(OpenAIChatCompletionContentPartTextParam(text=user_input.content))
-        else:
-            user_content = input
         messages.append(OpenAIUserMessageParam(content=user_content))
 
         chat_tools = await self._convert_response_tools_to_chat_tools(tools) if tools else None
@@ -198,10 +236,32 @@ class OpenAIResponsesImpl:
 
         if store:
             # Store in kvstore
+
+            if isinstance(input, str):
+                # synthesize a message from the input string
+                input_content = OpenAIResponseInputMessageContentText(text=input)
+                input_content_item = OpenAIResponseInputItemMessage(
+                    role="user",
+                    content=[input_content],
+                    id=f"msg_{uuid.uuid4()}",
+                )
+                input_items_data = [input_content_item]
+            else:
+                # we already have a list of messages
+                input_items_data = []
+                for input_item in input:
+                    input_items_data.append(
+                        OpenAIResponseInputItemMessage(id=f"msg_{uuid.uuid4()}", **input_item.model_dump())
+                    )
+            input_items = OpenAIResponseInputItemList(data=input_items_data)
+            prev_response = OpenAIResponsePreviousResponseWithInputItems(
+                input_items=input_items,
+                response=response,
+            )
             key = f"{OPENAI_RESPONSES_PREFIX}{response.id}"
             await self.persistence_store.set(
                 key=key,
-                value=response.model_dump_json(),
+                value=prev_response.model_dump_json(),
             )
 
         if stream: