phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-10-23 00:27:26 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

chore(cleanup)!: kill vector_db references as far as possible (#3864)

Browse source 
There should not be "vector db" anywhere.
This commit is contained in:
Ashwin Bharambe 2025-10-20 20:06:16 -07:00 • committed by GitHub
parent 444f6c88f3
commit 122de785c4
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
46 changed files with 701 additions and 822 deletions
Download patch file Download diff file Expand all files Collapse all files

4
llama_stack/core/routers/__init__.py
View file

@ -29,7 +29,7 @@ async def get_routing_table_impl(
from ..routing_tables.scoring_functions import ScoringFunctionsRoutingTable
from ..routing_tables.shields import ShieldsRoutingTable
from ..routing_tables.toolgroups import ToolGroupsRoutingTable
from ..routing_tables.vector_dbs import VectorDBsRoutingTable
from ..routing_tables.vector_stores import VectorStoresRoutingTable
api_to_tables = {
"models": ModelsRoutingTable,
@ -38,7 +38,7 @@ async def get_routing_table_impl(
"scoring_functions": ScoringFunctionsRoutingTable,
"benchmarks": BenchmarksRoutingTable,
"tool_groups": ToolGroupsRoutingTable,
"vector_dbs": VectorDBsRoutingTable,
"vector_stores": VectorStoresRoutingTable,
}
if api.value not in api_to_tables:

12
llama_stack/core/routers/tool_runtime.py
View file

@ -37,24 +37,24 @@ class ToolRuntimeRouter(ToolRuntime):
async def query(
self,
content: InterleavedContent,
vector_db_ids: list[str],
vector_store_ids: list[str],
query_config: RAGQueryConfig | None = None,
) -> RAGQueryResult:
logger.debug(f"ToolRuntimeRouter.RagToolImpl.query: {vector_db_ids}")
logger.debug(f"ToolRuntimeRouter.RagToolImpl.query: {vector_store_ids}")
provider = await self.routing_table.get_provider_impl("knowledge_search")
return await provider.query(content, vector_db_ids, query_config)
return await provider.query(content, vector_store_ids, query_config)
async def insert(
self,
documents: list[RAGDocument],
vector_db_id: str,
vector_store_id: str,
chunk_size_in_tokens: int = 512,
) -> None:
logger.debug(
f"ToolRuntimeRouter.RagToolImpl.insert: {vector_db_id}, {len(documents)} documents, chunk_size={chunk_size_in_tokens}"
f"ToolRuntimeRouter.RagToolImpl.insert: {vector_store_id}, {len(documents)} documents, chunk_size={chunk_size_in_tokens}"
)
provider = await self.routing_table.get_provider_impl("insert_into_memory")
return await provider.insert(documents, vector_db_id, chunk_size_in_tokens)
return await provider.insert(documents, vector_store_id, chunk_size_in_tokens)
def __init__(
self,

47
llama_stack/core/routers/vector_io.py
View file

@ -71,25 +71,6 @@ class VectorIORouter(VectorIO):
raise ValueError(f"Embedding model '{embedding_model_id}' not found or not an embedding model")
async def register_vector_db(
self,
vector_db_id: str,
embedding_model: str,
embedding_dimension: int | None = 384,
provider_id: str | None = None,
vector_db_name: str | None = None,
provider_vector_db_id: str | None = None,
) -> None:
logger.debug(f"VectorIORouter.register_vector_db: {vector_db_id}, {embedding_model}")
await self.routing_table.register_vector_db(
vector_db_id,
embedding_model,
embedding_dimension,
provider_id,
vector_db_name,
provider_vector_db_id,
)
async def insert_chunks(
self,
vector_db_id: str,
@ -165,22 +146,22 @@ class VectorIORouter(VectorIO):
else:
provider_id = list(self.routing_table.impls_by_provider_id.keys())[0]
vector_db_id = f"vs_{uuid.uuid4()}"
registered_vector_db = await self.routing_table.register_vector_db(
vector_db_id=vector_db_id,
vector_store_id = f"vs_{uuid.uuid4()}"
registered_vector_store = await self.routing_table.register_vector_store(
vector_store_id=vector_store_id,
embedding_model=embedding_model,
embedding_dimension=embedding_dimension,
provider_id=provider_id,
provider_vector_db_id=vector_db_id,
vector_db_name=params.name,
provider_vector_store_id=vector_store_id,
vector_store_name=params.name,
)
provider = await self.routing_table.get_provider_impl(registered_vector_db.identifier)
provider = await self.routing_table.get_provider_impl(registered_vector_store.identifier)
# Update model_extra with registered values so provider uses the already-registered vector_db
# Update model_extra with registered values so provider uses the already-registered vector_store
if params.model_extra is None:
params.model_extra = {}
params.model_extra["provider_vector_db_id"] = registered_vector_db.provider_resource_id
params.model_extra["provider_id"] = registered_vector_db.provider_id
params.model_extra["provider_vector_store_id"] = registered_vector_store.provider_resource_id
params.model_extra["provider_id"] = registered_vector_store.provider_id
if embedding_model is not None:
params.model_extra["embedding_model"] = embedding_model
if embedding_dimension is not None:
@ -198,15 +179,15 @@ class VectorIORouter(VectorIO):
logger.debug(f"VectorIORouter.openai_list_vector_stores: limit={limit}")
# Route to default provider for now - could aggregate from all providers in the future
# call retrieve on each vector dbs to get list of vector stores
vector_dbs = await self.routing_table.get_all_with_type("vector_db")
vector_stores = await self.routing_table.get_all_with_type("vector_store")
all_stores = []
for vector_db in vector_dbs:
for vector_store in vector_stores:
try:
provider = await self.routing_table.get_provider_impl(vector_db.identifier)
vector_store = await provider.openai_retrieve_vector_store(vector_db.identifier)
provider = await self.routing_table.get_provider_impl(vector_store.identifier)
vector_store = await provider.openai_retrieve_vector_store(vector_store.identifier)
all_stores.append(vector_store)
except Exception as e:
logger.error(f"Error retrieving vector store {vector_db.identifier}: {e}")
logger.error(f"Error retrieving vector store {vector_store.identifier}: {e}")
continue
# Sort by created_at