Commit graph

6 commits

Author SHA1 Message Date
Yuan Tang
a79a083e39
Fix broken pgvector provider and memory leaks (#947)
This PR fixes the broken pgvector provider as well as wraps all cursor
object creations with context manager to ensure that they get properly
closed to avoid potential memory leaks.

```
> pytest llama_stack/providers/tests/vector_io/test_vector_io.py   -m "pgvector" --env EMBEDDING_DIMENSION=384 --env PGVECTOR_PORT=7432 --env PGVECTOR_DB=db --env PGVECTOR_USER=user --env PGVECTOR_PASSWORD=pass   -v -s --tb=short --disable-warnings

llama_stack/providers/tests/vector_io/test_vector_io.py::TestVectorIO::test_banks_list[-pgvector] PASSED
llama_stack/providers/tests/vector_io/test_vector_io.py::TestVectorIO::test_banks_register[-pgvector] PASSED
llama_stack/providers/tests/vector_io/test_vector_io.py::TestVectorIO::test_query_documents[-pgvector] The scores are: [0.8168284974053789, 0.8080469278964486, 0.8050996198466661]
PASSED
```

---------

Signed-off-by: Yuan Tang <terrytangyuan@gmail.com>
2025-02-05 09:32:05 -08:00
Yuan Tang
83a51c7bfb
Properly close PGVector DB connection during shutdown() (#931)
The connection to the DB was not closed during shutdown.

Signed-off-by: Yuan Tang <terrytangyuan@gmail.com>
2025-02-02 21:23:13 -08:00
Yuan Tang
34ab7a3b6c
Fix precommit check after moving to ruff (#927)
Lint check in main branch is failing. This fixes the lint check after we
moved to ruff in https://github.com/meta-llama/llama-stack/pull/921. We
need to move to a `ruff.toml` file as well as fixing and ignoring some
additional checks.

Signed-off-by: Yuan Tang <terrytangyuan@gmail.com>
2025-02-02 06:46:45 -08:00
Ashwin Bharambe
f2feb7d15c
Fix Chroma adapter (#893)
Chroma method had the wrong signature.

## Test Plan

Start Chroma: `chroma run --path /tmp/foo/chroma2 --host localhost
--port 6001`

Modify run.yaml to include Chroma server pointing to localhost:6001 and
run `llama stack run`

Then:

```bash
LLAMA_STACK_BASE_URL=http://localhost:8321 pytest -s -v agents/test_agents.py -k rag
```

passes
2025-01-28 13:19:47 -08:00
Ashwin Bharambe
c9e5578151
[memory refactor][5/n] Migrate all vector_io providers (#835)
See https://github.com/meta-llama/llama-stack/issues/827 for the broader
design.

This PR finishes off all the stragglers and migrates everything to the
new naming.
2025-01-22 10:17:59 -08:00
Ashwin Bharambe
3ae8585b65
[memory refactor][1/n] Rename Memory -> VectorIO, MemoryBanks -> VectorDBs (#828)
See https://github.com/meta-llama/llama-stack/issues/827 for the broader
design.

This is the first part:

- delete other kinds of memory banks (keyvalue, keyword, graph) for now;
we will introduce a keyvalue store API as part of this design but not
use it in the RAG tool yet.
- renaming of the APIs
2025-01-22 09:59:30 -08:00