phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-07-17 10:28:11 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

sqlite-vec support for Responses file_search

Browse source 
This wires up the Files API optional dependency into sqlite_vec and
adds the localfs Files provider to our starter template, so that
Responses API file_search tool works out of the box for sqlite_vec in
that template.

Some additional testing with this provider plus some other inference
models led me to loosen the verification test results checking a bit -
not for the tool call, but just around the assistant response with the
file_search tool call. Some providers, such as OpenAI SaaS, make
multiple tool calls to resolve the query sometimes, especially when it
cannot find an answer so tries a few permutations before returning
empty results to the user in that test.

Signed-off-by: Ben Browning <bbrownin@redhat.com>
This commit is contained in:
Ben Browning 2025-06-13 14:01:34 -04:00
parent ec09524a91
commit 7a71d9ebd8
8 changed files with 30 additions and 24 deletions
Download patch file Download diff file Expand all files Collapse all files

2
llama_stack/providers/inline/vector_io/sqlite_vec/__init__.py
View file

@ -15,6 +15,6 @@ async def get_provider_impl(config: SQLiteVectorIOConfig, deps: dict[Api, Any]):
from .sqlite_vec import SQLiteVecVectorIOAdapter
assert isinstance(config, SQLiteVectorIOConfig), f"Unexpected config type: {type(config)}"
impl = SQLiteVecVectorIOAdapter(config, deps[Api.inference])
impl = SQLiteVecVectorIOAdapter(config, deps[Api.inference], deps.get(Api.files, None))
await impl.initialize()
return impl

14
llama_stack/providers/inline/vector_io/sqlite_vec/sqlite_vec.py
View file

@ -17,6 +17,7 @@ import numpy as np
import sqlite_vec
from numpy.typing import NDArray
from llama_stack.apis.files.files import Files
from llama_stack.apis.inference.inference import Inference
from llama_stack.apis.vector_dbs import VectorDB
from llama_stack.apis.vector_io import (
@ -24,7 +25,6 @@ from llama_stack.apis.vector_io import (
QueryChunksResponse,
VectorIO,
)
from llama_stack.apis.vector_io.vector_io import VectorStoreChunkingStrategy, VectorStoreFileObject
from llama_stack.providers.datatypes import VectorDBsProtocolPrivate
from llama_stack.providers.utils.memory.openai_vector_store_mixin import OpenAIVectorStoreMixin
from llama_stack.providers.utils.memory.vector_store import EmbeddingIndex, VectorDBWithIndex
@ -302,9 +302,10 @@ class SQLiteVecVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtoc
and creates a cache of VectorDBWithIndex instances (each wrapping a SQLiteVecIndex).
"""
def __init__(self, config, inference_api: Inference) -> None:
def __init__(self, config, inference_api: Inference, files_api: Files | None) -> None:
self.config = config
self.inference_api = inference_api
self.files_api = files_api
self.cache: dict[str, VectorDBWithIndex] = {}
self.openai_vector_stores: dict[str, dict[str, Any]] = {}
@ -490,15 +491,6 @@ class SQLiteVecVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtoc
raise ValueError(f"Vector DB {vector_db_id} not found")
return await self.cache[vector_db_id].query_chunks(query, params)
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:
raise NotImplementedError("OpenAI Vector Stores Files API is not supported in sqlite_vec")
def generate_chunk_id(document_id: str, chunk_text: str) -> str:
"""Generate a unique chunk ID using a hash of document ID and chunk text."""