longbrain Files

Highlights

• Document search upgrade: BGE-M3 doc embedder, reranker, and enrichment (SEARCH_SPEC Sprints 0-3)
• Memory reliability (Phase 1 complete): document versioning, recency decay, context budgeting, trace, admission gate
• Connectors (Phase 2 start): Google Drive connector — first connector
• Security: optional shared-secret auth, tiered rate limits, write/destructive action audit
• API: stateless embedding and completion endpoints
• Eval: scale/live recall modes, id-based ranking, memory audit script
• Setup: curl-able bootstrap installer, auto-detect and fix doc-embed mismatch
• UI: cleanup action for superseded document chunks
• Codex adapter: fix to persist LONGBRAIN_API_KEY across logins/reboots (launchd env sync)

Full commit list

• feat(install): add curl-able bootstrap installer; mark project stable
• docs: improve repository discoverability
• docs: update verified eval metrics
• feat(search): document search upgrade — BGE-M3 doc embedder, reranker, enrichment (SEARCH_SPEC Sprints 0-3)
• feat(api): add stateless embedding and completion endpoints
• feat(eval): add scale/live recall modes, id-based ranking, and a memory audit script
• feat(memory): document versioning, recency decay, context budgeting, trace, admission gate
• docs: mark Phase 1 (Memory Reliability) complete with verified baseline
• feat(connectors): Google Drive connector — first connector (Phase 2)
• feat(setup): auto-detect and offer fix for doc-embed mismatch
• feat(security): optional shared-secret auth, tiered rate limits, write/destructive audit
• fix(tests): add repo root to sys.path so dotted hooks.* imports work in CI
• feat(ui): add cleanup action for superseded document chunks
• fix(codex): persist LONGBRAIN_API_KEY across logins
• docs: update test count to 303