Skip to content

Security: fairbearlab/bear-app-rag

SECURITY.md

Security Policy

Supported versions

Only the latest release of bear-app-rag receives security fixes.

Reporting a vulnerability

Please do not open a public issue for security problems. Use GitHub's private vulnerability reporting: Security → Report a vulnerability on this repository. You'll get an acknowledgement within 7 days.

Local-first threat model

bear-app-rag runs entirely on your machine. Indexing, sync, and search make no network requests after the one-time embedding-model download, and tests/test_privacy.py enforces that boundary in CI. Your notes are read from Bear's SQLite database read-only and are never uploaded anywhere. The only credential the project uses is ANTHROPIC_API_KEY for the opt-in eval judge, which is injected at runtime and never written to the tree.

This narrows the exposure: the realistic risks are local ones — a malicious note influencing agent output through MCP, or the vector store leaking notes the user expected to be excluded (archived and tag-filtered notes are covered by tests).

Known upstream advisories

ChromaDB 1.5.9 (the latest release, no fixed version available at the time of writing) carries four open advisories, all scoped to the ChromaDB HTTP server and its auth layer:

This project uses chromadb.PersistentClient in-process (bear_rag/store.py), never starts or exposes the HTTP server, and never configures chromadb's authn/authz providers, so none of the vulnerable endpoints are reachable. The only remaining in-process path, rebuilding an embedding function from persisted collection config, is written solely by this process's own default ONNX embedding function; an attacker would already need write access to the local Chroma directory. CI's pip-audit step (and make audit) ignores these four IDs; each ignore is removed once a fixed release ships.

What's already in place

  • Dependencies are monitored by Dependabot; CI runs a vulnerability scan on every push.
  • GitHub Actions are pinned to commit SHAs and run with minimal token permissions.
  • This repository is scored by OpenSSF Scorecard.

There aren't any published security advisories