🔎 feat: Add a Single-Owner Presence Check for Embedded Files

[berry-13]

Summary

LibreChat needs to know whether a file is already embedded for an agent before embedding it again (LibreChat-AI/LibreChat#17206). No route can answer that today. /query and /documents read the caller and the entity_id together, so a copy the user embedded under their own id answers for the agent, and with enough of those chunks ranking first the agent's own never reach the top k.

POST /exists takes file_ids and an optional entity_id and returns the ids holding chunks owned by exactly that entity, or by the caller when none is named. It reuses the scoped get_filtered_ids on both stores, runs no embedding and returns no content. A named entity replaces the caller instead of widening the set, so the route only narrows what the same request could already read through /query. The README's retrieval scope section documents it.

Change Type

• New feature (non-breaking change which adds functionality)

Testing

• pytest tests/test_authorization.py: 17 passed, including new cases for an entity whose file id the caller also holds, the caller-only scope (a foreign file reads as absent), and an unauthenticated deployment.
• Negative control: with the route on resolve_scope, the entity case fails because the caller's copy answers for the agent.
• Full unit suite: 312 passed. Black 24.4.0 is clean on the touched files.
• The pgvector integration suite was not run locally (needs Docker); get_filtered_ids itself is unchanged and covered there.

Checklist

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have added tests that prove my feature works
• Local unit tests pass with my changes
• I have updated the documentation

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-18T09:28:08.318516Z	1fb7404	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 89668bc33a

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Offload Atlas existence lookups from the event loop

When VECTOR_DB_TYPE=atlas-mongo, this branch calls AtlasMongoVector.get_filtered_ids synchronously; that method executes the blocking PyMongo collection.distinct operation (app/services/vector_store/atlas_mongo_vector.py:79-86), so Mongo/network latency here stalls the FastAPI event-loop thread and unrelated concurrent requests. Run this lookup through run_in_executor, as the other non-PG query branches do.

Useful? React with 👍 / 👎.

[berry-13]

Fixed in 1fb7404: the synchronous branch now dispatches get_filtered_ids through run_in_executor on the app thread pool, as /query does. test_sync_store_presence_check_leaves_event_loop_responsive asserts the lookup runs on a worker thread while the loop stays free; it fails with the direct call restored.