feat: Enable Filters in OpenAI Search API by franciscojavierarceo · Pull Request #4471 · ogx-ai/ogx

franciscojavierarceo · 2026-01-08T20:30:06Z

What does this PR do?

Finishes #2435; i.e., adds support for OpenAI Search API typed filters for vector IO.

This PR adds support for OpenAI's typed metadata filter API across vector IO providers, enabling callers to filter search results by metadata fields using comparison operators (eq, ne, gt, gte, lt, lte, in, nin) and compound logic (and, or).

Core changes:

API & routing (core/routers/vector_io.py): Propagates filter parameters from the query request through the provider dispatch layer.
Filter type definitions (providers/utils/vector_io/filters.py): Introduces ComparisonFilter, CompoundFilter, and Filter union type used across all providers.
Inline providers (faiss, sqlite-vec): Implement filter evaluation in-memory (faiss) and via SQL WHERE clauses using JSON_EXTRACT (sqlite-vec). Operator dispatch uses a module-level COMPARISON_OPS / SQL_OPS dict instead of if/elif chains.
Remote providers (Milvus, Chroma, Elasticsearch, pgvector, Qdrant, Weaviate, OCI): Translate filters to each backend's native expression or query language.
OpenAI vector store mixin (openai_vector_store_mixin.py): Post-retrieval Python-side filter evaluation using a shared COMPARISON_OPERATORS dispatch dict, covering providers that don't support server-side filtering natively.
Milvus bug fix: filter parameter was not being passed to keyword search — corrected.

Test Plan

New integration test suite: tests/integration/vector_io/test_vector_io_filters.py — covers all 8 comparison operators and compound (and/or) logic
against faiss and sqlite-vec
New integration test: test_openai_vector_store_search_with_typed_filters in test_openai_vector_stores.py — validates eq/lte/compound-or filters over
the OpenAI vector store path
Run integration tests locally via:

faiss

.venv/bin/python -m pytest tests/integration/vector_io/test_vector_io_filters.py
--stack-config="inference=inline::sentence-transformers,vector_io=inline::faiss,files=inline::localfs" -v

sqlite-vec

.venv/bin/python -m pytest tests/integration/vector_io/test_vector_io_filters.py
--stack-config="inference=inline::sentence-transformers,vector_io=inline::sqlite-vec,files=inline::localfs" -v

OpenAI vector store filter (mixin path)

.venv/bin/python -m pytest tests/integration/vector_io/test_openai_vector_stores.py::test_openai_vector_store_search_with_typed_filters
--stack-config="inference=inline::sentence-transformers,vector_io=inline::sqlite-vec,files=inline::localfs" -v

github-actions · 2026-02-03T18:38:56Z

✱ Stainless preview builds

This PR will update the llama-stack-client SDKs with the following commit message.

feat: Enable Filters in OpenAI Search API

⚠️

llama-stack-client-openapi studio · code

There was a regression in your SDK.
generate ⚠️

⚠️

llama-stack-client-python studio · conflict

There was a regression in your SDK.

⚠️

llama-stack-client-go studio · conflict

There was a regression in your SDK.

⚠️

llama-stack-client-node studio · conflict

There was a regression in your SDK.

This comment is auto-generated by GitHub Actions and is automatically kept up to date as you push.
If you push custom code to the preview branch, re-run this workflow to update the comment.
Last updated: 2026-03-10 17:34:16 UTC

mergify · 2026-02-05T05:12:27Z

This pull request has merge conflicts that must be resolved before it can be merged. @franciscojavierarceo please rebase it. https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

Signed-off-by: Francisco Javier Arceo <farceo@redhat.com>

…ltering Signed-off-by: Francisco Javier Arceo <farceo@redhat.com>

Signed-off-by: Francisco Javier Arceo <farceo@redhat.com>

cdoern

lgtm, this is still in draft but looks pretty clean to me!

Adds support for dynamic metadata filtering to the Solr vector IO provider, enabling users to filter RAG query results by metadata fields at runtime. NOTE: Dynamic filters from requests are **combined with** the existing static `chunk_filter_query` configuration (e.g., `"is_chunk:true"`) rather than replacing it. This preserves backward compatibility and maintains the internal schema filtering needed for chunk window expansion to work correctly. Both filters are joined with AND logic: fq=(is_chunk:true AND ) This approach avoids breaking changes while enabling flexible metadata filtering for use cases like filtering by platform, version, document type, etc. - Upstream PR: ogx-ai/ogx#4471

…sults (backport #6093) (#6117) ## Summary - The catch-all `except Exception` in `openai_search_vector_store` was silently swallowing backend errors and returning empty results with HTTP 200 - Clients had no way to distinguish "no matching documents" from "the search failed" - This masked bugs like the milvus-lite 3.0 `chunk_content` KeyError (#6089), where file_search appeared to work but retrieval silently returned nothing - Now re-raises the exception so it surfaces as a 500 to the client - Also fixes `test_openai_vector_store_with_chunks` which was passing `filters={"topic": "ai"}` (the old shorthand format) instead of the typed format `{"type": "eq", "key": "topic", "value": "ai"}` required since #4471. The error was silently swallowed, and the assertion loop over empty results never executed, so the test appeared to pass. Closes #6092 ## Test plan - Added unit test `test_search_vector_store_propagates_backend_errors` that verifies a `KeyError` from the backend propagates to the caller - Fixed `test_openai_vector_store_with_chunks` filter format so it actually exercises the filter path - Reproduced locally with milvus-lite 3.0: before the fix, search returns 200 with empty results; after, returns 500<hr>This is an automatic backport of pull request #6093 done by [Mergify](https://mergify.com). Signed-off-by: Derek Higgins <derekh@redhat.com> Co-authored-by: Derek Higgins <derekh@redhat.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: Charlie Doern <cdoern@redhat.com>

…sults (ogx-ai#6093) ## Summary - The catch-all `except Exception` in `openai_search_vector_store` was silently swallowing backend errors and returning empty results with HTTP 200 - Clients had no way to distinguish "no matching documents" from "the search failed" - This masked bugs like the milvus-lite 3.0 `chunk_content` KeyError (ogx-ai#6089), where file_search appeared to work but retrieval silently returned nothing - Now re-raises the exception so it surfaces as a 500 to the client - Also fixes `test_openai_vector_store_with_chunks` which was passing `filters={"topic": "ai"}` (the old shorthand format) instead of the typed format `{"type": "eq", "key": "topic", "value": "ai"}` required since ogx-ai#4471. The error was silently swallowed, and the assertion loop over empty results never executed, so the test appeared to pass. Closes ogx-ai#6092 ## Test plan - Added unit test `test_search_vector_store_propagates_backend_errors` that verifies a `KeyError` from the backend propagates to the caller - Fixed `test_openai_vector_store_with_chunks` filter format so it actually exercises the filter path - Reproduced locally with milvus-lite 3.0: before the fix, search returns 200 with empty results; after, returns 500 --------- Signed-off-by: Derek Higgins <derekh@redhat.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: Charlie Doern <cdoern@redhat.com>

meta-cla Bot added the CLA Signed This label is managed by the Meta Open Source bot. label Jan 8, 2026

mergify Bot added the needs-rebase label Jan 16, 2026

franciscojavierarceo force-pushed the search-filters branch from 375e38f to 76520d2 Compare January 16, 2026 05:15

mergify Bot removed the needs-rebase label Jan 16, 2026

franciscojavierarceo force-pushed the search-filters branch from 552e0fb to 415cb9e Compare January 26, 2026 04:13

mergify Bot added the needs-rebase label Jan 26, 2026

franciscojavierarceo force-pushed the search-filters branch 3 times, most recently from 9e9b839 to c93216a Compare January 30, 2026 01:49

mergify Bot added needs-rebase and removed needs-rebase labels Jan 30, 2026

franciscojavierarceo closed this Jan 31, 2026

franciscojavierarceo force-pushed the search-filters branch from 7241907 to 8701d91 Compare January 31, 2026 19:29

franciscojavierarceo reopened this Jan 31, 2026

mergify Bot added needs-rebase and removed needs-rebase labels Jan 31, 2026

mergify Bot added the needs-rebase label Feb 5, 2026

franciscojavierarceo force-pushed the search-filters branch from 0a46cc9 to 8125175 Compare February 9, 2026 20:17

mergify Bot added needs-rebase and removed needs-rebase labels Feb 9, 2026

franciscojavierarceo force-pushed the search-filters branch from 9eee28c to a36ef7e Compare February 10, 2026 21:55

mergify Bot added needs-rebase and removed needs-rebase labels Feb 10, 2026

franciscojavierarceo force-pushed the search-filters branch 2 times, most recently from e9e8e75 to 5d99e8b Compare February 11, 2026 15:43

franciscojavierarceo added 2 commits March 10, 2026 09:17

rmeove unused tests

517375e

Signed-off-by: Francisco Javier Arceo <farceo@redhat.com>

updated faiss approach to be useful

79b20b0

Signed-off-by: Francisco Javier Arceo <farceo@redhat.com>

franciscojavierarceo force-pushed the search-filters branch from ffc59e9 to 79b20b0 Compare March 10, 2026 13:18

mergify Bot removed the needs-rebase label Mar 10, 2026

franciscojavierarceo and others added 5 commits March 10, 2026 10:18

incorporate feedback

9a48049

Signed-off-by: Francisco Javier Arceo <farceo@redhat.com>

cleanup implementation to move logic pre-filtering instead of post fi…

90f5b4f

…ltering Signed-off-by: Francisco Javier Arceo <farceo@redhat.com>

filter out unsupported vector store providers

1e46c10

Signed-off-by: Francisco Javier Arceo <farceo@redhat.com>

Merge branch 'main' into search-filters

8081276

updated filtering to include milvus

9363d55

Signed-off-by: Francisco Javier Arceo <farceo@redhat.com>

cdoern approved these changes Mar 10, 2026

View reviewed changes

franciscojavierarceo marked this pull request as ready for review March 10, 2026 17:32

franciscojavierarceo requested review from ashwinb, bbrowning, ehhuang, leseb, mattf and raghotham as code owners March 10, 2026 17:32

franciscojavierarceo merged commit d1ccbfe into ogx-ai:main Mar 10, 2026
63 checks passed

anik120 mentioned this pull request Apr 23, 2026

LCORE-1446: Add dynamic filter support to Solr vector IO provider lightspeed-core/lightspeed-providers#119

Merged

17 tasks

derekhiggins mentioned this pull request Jun 12, 2026

fix(vector_io): propagate search errors instead of returning empty results #6093

Merged

mergify Bot mentioned this pull request Jun 15, 2026

fix(vector_io): propagate search errors instead of returning empty results (backport #6093) #6117

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: Enable Filters in OpenAI Search API#4471

feat: Enable Filters in OpenAI Search API#4471
franciscojavierarceo merged 13 commits into
ogx-ai:mainfrom
franciscojavierarceo:search-filters

franciscojavierarceo commented Jan 8, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Feb 3, 2026 •

edited

Loading

Uh oh!

mergify Bot commented Feb 5, 2026

Uh oh!

cdoern left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

franciscojavierarceo commented Jan 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this PR do?

Test Plan

faiss

sqlite-vec

OpenAI vector store filter (mixin path)

Uh oh!

github-actions Bot commented Feb 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✱ Stainless preview builds

Uh oh!

mergify Bot commented Feb 5, 2026

Uh oh!

cdoern left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

franciscojavierarceo commented Jan 8, 2026 •

edited

Loading

github-actions Bot commented Feb 3, 2026 •

edited

Loading