phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-07-26 22:19:49 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat(registry): more flexible model lookup (#2859)

Browse source 
This PR updates model registration and lookup behavior to be slightly
more general / flexible. See
https://github.com/meta-llama/llama-stack/issues/2843 for more details.

Note that this change is backwards compatible given the design of the
`lookup_model()` method.

## Test Plan

Added unit tests
This commit is contained in:
Ashwin Bharambe 2025-07-22 15:22:48 -07:00 • committed by GitHub
parent 9736f096f6
commit 3b83032555
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
15 changed files with 265 additions and 75 deletions
Download patch file Download diff file Expand all files Collapse all files

37
llama_stack/distribution/routing_tables/vector_dbs.py
View file

@ -27,7 +27,7 @@ from llama_stack.distribution.datatypes import (
)
from llama_stack.log import get_logger
from .common import CommonRoutingTableImpl
from .common import CommonRoutingTableImpl, lookup_model
logger = get_logger(name=__name__, category="core")
@ -51,8 +51,7 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
provider_vector_db_id: str | None = None,
vector_db_name: str | None = None,
) -> VectorDB:
if provider_vector_db_id is None:
provider_vector_db_id = vector_db_id
provider_vector_db_id = provider_vector_db_id or vector_db_id
if provider_id is None:
if len(self.impls_by_provider_id) > 0:
provider_id = list(self.impls_by_provider_id.keys())[0]
@ -62,7 +61,7 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
)
else:
raise ValueError("No provider available. Please configure a vector_io provider.")
model = await self.get_object_by_identifier("model", embedding_model)
model = await lookup_model(self, embedding_model)
if model is None:
raise ValueError(f"Model {embedding_model} not found")
if model.model_type != ModelType.embedding:
@ -93,7 +92,8 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
vector_store_id: str,
) -> VectorStoreObject:
await self.assert_action_allowed("read", "vector_db", vector_store_id)
return await self.get_provider_impl(vector_store_id).openai_retrieve_vector_store(vector_store_id)
provider = await self.get_provider_impl(vector_store_id)
return await provider.openai_retrieve_vector_store(vector_store_id)
async def openai_update_vector_store(
self,
@ -103,7 +103,8 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
metadata: dict[str, Any] | None = None,
) -> VectorStoreObject:
await self.assert_action_allowed("update", "vector_db", vector_store_id)
return await self.get_provider_impl(vector_store_id).openai_update_vector_store(
provider = await self.get_provider_impl(vector_store_id)
return await provider.openai_update_vector_store(
vector_store_id=vector_store_id,
name=name,
expires_after=expires_after,
@ -115,7 +116,8 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
vector_store_id: str,
) -> VectorStoreDeleteResponse:
await self.assert_action_allowed("delete", "vector_db", vector_store_id)
result = await self.get_provider_impl(vector_store_id).openai_delete_vector_store(vector_store_id)
provider = await self.get_provider_impl(vector_store_id)
result = await provider.openai_delete_vector_store(vector_store_id)
await self.unregister_vector_db(vector_store_id)
return result
@ -130,7 +132,8 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
search_mode: str | None = "vector",
) -> VectorStoreSearchResponsePage:
await self.assert_action_allowed("read", "vector_db", vector_store_id)
return await self.get_provider_impl(vector_store_id).openai_search_vector_store(
provider = await self.get_provider_impl(vector_store_id)
return await provider.openai_search_vector_store(
vector_store_id=vector_store_id,
query=query,
filters=filters,
@ -148,7 +151,8 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
chunking_strategy: VectorStoreChunkingStrategy | None = None,
) -> VectorStoreFileObject:
await self.assert_action_allowed("update", "vector_db", vector_store_id)
return await self.get_provider_impl(vector_store_id).openai_attach_file_to_vector_store(
provider = await self.get_provider_impl(vector_store_id)
return await provider.openai_attach_file_to_vector_store(
vector_store_id=vector_store_id,
file_id=file_id,
attributes=attributes,
@ -165,7 +169,8 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
filter: VectorStoreFileStatus | None = None,
) -> list[VectorStoreFileObject]:
await self.assert_action_allowed("read", "vector_db", vector_store_id)
return await self.get_provider_impl(vector_store_id).openai_list_files_in_vector_store(
provider = await self.get_provider_impl(vector_store_id)
return await provider.openai_list_files_in_vector_store(
vector_store_id=vector_store_id,
limit=limit,
order=order,
@ -180,7 +185,8 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
file_id: str,
) -> VectorStoreFileObject:
await self.assert_action_allowed("read", "vector_db", vector_store_id)
return await self.get_provider_impl(vector_store_id).openai_retrieve_vector_store_file(
provider = await self.get_provider_impl(vector_store_id)
return await provider.openai_retrieve_vector_store_file(
vector_store_id=vector_store_id,
file_id=file_id,
)
@ -191,7 +197,8 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
file_id: str,
) -> VectorStoreFileContentsResponse:
await self.assert_action_allowed("read", "vector_db", vector_store_id)
return await self.get_provider_impl(vector_store_id).openai_retrieve_vector_store_file_contents(
provider = await self.get_provider_impl(vector_store_id)
return await provider.openai_retrieve_vector_store_file_contents(
vector_store_id=vector_store_id,
file_id=file_id,
)
@ -203,7 +210,8 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
attributes: dict[str, Any],
) -> VectorStoreFileObject:
await self.assert_action_allowed("update", "vector_db", vector_store_id)
return await self.get_provider_impl(vector_store_id).openai_update_vector_store_file(
provider = await self.get_provider_impl(vector_store_id)
return await provider.openai_update_vector_store_file(
vector_store_id=vector_store_id,
file_id=file_id,
attributes=attributes,
@ -215,7 +223,8 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
file_id: str,
) -> VectorStoreFileDeleteResponse:
await self.assert_action_allowed("delete", "vector_db", vector_store_id)
return await self.get_provider_impl(vector_store_id).openai_delete_vector_store_file(
provider = await self.get_provider_impl(vector_store_id)
return await provider.openai_delete_vector_store_file(
vector_store_id=vector_store_id,
file_id=file_id,
)