feat: File search tool for Responses API (#2426)

# What does this PR do?

This is an initial working prototype of wiring up the `file_search`
builtin tool for the Responses API to our existing rag knowledge search
tool.

This is me seeing what I could pull together on top of the bits we
already have merged. This may not be the ideal way to implement this,
and things like how I shuffle the vector store ids from the original
response API tool request to the actual tool execution feel a bit hacky
(grep for `tool_kwargs["vector_db_ids"]` in `_execute_tool_call` to see
what I mean).

## Test Plan

I stubbed in some new tests to exercise this using text and pdf
documents.

Note that this is currently under tests/verification only because it
sometimes flakes with tool calling of the small Llama-3.2-3B model we
run in CI (and that I use as an example below). We'd want to make the
test a bit more robust in some way if we moved this over to
tests/integration and ran it in CI.

### OpenAI SaaS (to verify test correctness)

```
pytest -sv tests/verifications/openai_api/test_responses.py \
  -k 'file_search' \
  --base-url=https://api.openai.com/v1 \
  --model=gpt-4o
```

### Fireworks with faiss vector store

```
llama stack run llama_stack/templates/fireworks/run.yaml

pytest -sv tests/verifications/openai_api/test_responses.py \
  -k 'file_search' \
  --base-url=http://localhost:8321/v1/openai/v1 \
  --model=meta-llama/Llama-3.3-70B-Instruct
```

### Ollama with faiss vector store

This sometimes flakes on Ollama because the quantized small model
doesn't always choose to call the tool to answer the user's question.
But, it often works.

```
ollama run llama3.2:3b

INFERENCE_MODEL="meta-llama/Llama-3.2-3B-Instruct" \
llama stack run ./llama_stack/templates/ollama/run.yaml \
  --image-type venv \
  --env OLLAMA_URL="http://0.0.0.0:11434"

pytest -sv tests/verifications/openai_api/test_responses.py \
  -k'file_search' \
  --base-url=http://localhost:8321/v1/openai/v1 \
  --model=meta-llama/Llama-3.2-3B-Instruct
```

### OpenAI provider with sqlite-vec vector store

```
llama stack run ./llama_stack/templates/starter/run.yaml --image-type venv

 pytest -sv tests/verifications/openai_api/test_responses.py \
  -k 'file_search' \
  --base-url=http://localhost:8321/v1/openai/v1 \
  --model=openai/gpt-4o-mini
```

### Ensure existing vector store integration tests still pass

```
ollama run llama3.2:3b

INFERENCE_MODEL="meta-llama/Llama-3.2-3B-Instruct" \
llama stack run ./llama_stack/templates/ollama/run.yaml \
  --image-type venv \
  --env OLLAMA_URL="http://0.0.0.0:11434"

LLAMA_STACK_CONFIG=http://localhost:8321 \
pytest -sv tests/integration/vector_io \
  --text-model "meta-llama/Llama-3.2-3B-Instruct" \
  --embedding-model=all-MiniLM-L6-v2
```

---------

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

llama_stack/apis/agents/openai_responses.py

@@ -81,6 +81,15 @@ class OpenAIResponseOutputMessageWebSearchToolCall(BaseModel):
     type: Literal["web_search_call"] = "web_search_call"
 
 
+@json_schema_type
+class OpenAIResponseOutputMessageFileSearchToolCall(BaseModel):
+    id: str
+    queries: list[str]
+    status: str
+    type: Literal["file_search_call"] = "file_search_call"
+    results: list[dict[str, Any]] | None = None
+
+
 @json_schema_type
 class OpenAIResponseOutputMessageFunctionToolCall(BaseModel):
     call_id: str
@@ -119,6 +128,7 @@ class OpenAIResponseOutputMessageMCPListTools(BaseModel):
 OpenAIResponseOutput = Annotated[
     OpenAIResponseMessage
     | OpenAIResponseOutputMessageWebSearchToolCall
+    | OpenAIResponseOutputMessageFileSearchToolCall
     | OpenAIResponseOutputMessageFunctionToolCall
     | OpenAIResponseOutputMessageMCPCall
     | OpenAIResponseOutputMessageMCPListTools,
@@ -362,6 +372,7 @@ class OpenAIResponseInputFunctionToolCallOutput(BaseModel):
 OpenAIResponseInput = Annotated[
     # Responses API allows output messages to be passed in as input
     OpenAIResponseOutputMessageWebSearchToolCall
+    | OpenAIResponseOutputMessageFileSearchToolCall
     | OpenAIResponseOutputMessageFunctionToolCall
     | OpenAIResponseInputFunctionToolCallOutput
     |
@@ -397,9 +408,10 @@ class FileSearchRankingOptions(BaseModel):
 @json_schema_type
 class OpenAIResponseInputToolFileSearch(BaseModel):
     type: Literal["file_search"] = "file_search"
-    vector_store_id: list[str]
+    vector_store_ids: list[str]
+    filters: dict[str, Any] | None = None
+    max_num_results: int | None = Field(default=10, ge=1, le=50)
     ranking_options: FileSearchRankingOptions | None = None
-    # TODO: add filters
 
 
 class ApprovalFilter(BaseModel):

llama_stack/apis/vector_io/vector_io.py

@@ -8,7 +8,7 @@
 #
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
-from typing import Any, Literal, Protocol, runtime_checkable
+from typing import Annotated, Any, Literal, Protocol, runtime_checkable
 
 from pydantic import BaseModel, Field
 
@@ -16,6 +16,7 @@ from llama_stack.apis.inference import InterleavedContent
 from llama_stack.apis.vector_dbs import VectorDB
 from llama_stack.providers.utils.telemetry.trace_protocol import trace_protocol
 from llama_stack.schema_utils import json_schema_type, webmethod
+from llama_stack.strong_typing.schema import register_schema
 
 
 class Chunk(BaseModel):
@@ -133,6 +134,50 @@ class VectorStoreDeleteResponse(BaseModel):
     deleted: bool = True
 
 
+@json_schema_type
+class VectorStoreChunkingStrategyAuto(BaseModel):
+    type: Literal["auto"] = "auto"
+
+
+@json_schema_type
+class VectorStoreChunkingStrategyStaticConfig(BaseModel):
+    chunk_overlap_tokens: int = 400
+    max_chunk_size_tokens: int = Field(800, ge=100, le=4096)
+
+
+@json_schema_type
+class VectorStoreChunkingStrategyStatic(BaseModel):
+    type: Literal["static"] = "static"
+    static: VectorStoreChunkingStrategyStaticConfig
+
+
+VectorStoreChunkingStrategy = Annotated[
+    VectorStoreChunkingStrategyAuto | VectorStoreChunkingStrategyStatic, Field(discriminator="type")
+]
+register_schema(VectorStoreChunkingStrategy, name="VectorStoreChunkingStrategy")
+
+
+@json_schema_type
+class VectorStoreFileLastError(BaseModel):
+    code: Literal["server_error"] | Literal["rate_limit_exceeded"]
+    message: str
+
+
+@json_schema_type
+class VectorStoreFileObject(BaseModel):
+    """OpenAI Vector Store File object."""
+
+    id: str
+    object: str = "vector_store.file"
+    attributes: dict[str, Any] = Field(default_factory=dict)
+    chunking_strategy: VectorStoreChunkingStrategy
+    created_at: int
+    last_error: VectorStoreFileLastError | None = None
+    status: Literal["completed"] | Literal["in_progress"] | Literal["cancelled"] | Literal["failed"]
+    usage_bytes: int = 0
+    vector_store_id: str
+
+
 class VectorDBStore(Protocol):
     def get_vector_db(self, vector_db_id: str) -> VectorDB | None: ...
 
@@ -290,3 +335,21 @@ class VectorIO(Protocol):
         :returns: A VectorStoreSearchResponse containing the search results.
         """
         ...
+
+    @webmethod(route="/openai/v1/vector_stores/{vector_store_id}/files", method="POST")
+    async def openai_attach_file_to_vector_store(
+        self,
+        vector_store_id: str,
+        file_id: str,
+        attributes: dict[str, Any] | None = None,
+        chunking_strategy: VectorStoreChunkingStrategy | None = None,
+    ) -> VectorStoreFileObject:
+        """Attach a file to a vector store.
+
+        :param vector_store_id: The ID of the vector store to attach the file to.
+        :param file_id: The ID of the file to attach to the vector store.
+        :param attributes: The key-value attributes stored with the file, which can be used for filtering.
+        :param chunking_strategy: The chunking strategy to use for the file.
+        :returns: A VectorStoreFileObject representing the attached file.
+        """
+        ...