You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
- Fix brain_reflect/consolidation 502: Gemini 2.5 Flash thinking tokens
consumed maxOutputTokens:4096 budget — bumped to 65536 with 8192
thinking cap, added finishReason check and non-thinking part selection
- Fix Qdrant payload index creation: 'Keyword' → 'keyword' (4 places)
- Tighten NER extraction: reject verb phrases, digits, 4+ word phrases,
sentence starters in both isJunkPhrase and isJunkQuotedName
- Add /reconcile endpoint for Qdrant/Postgres/keyword data layer sync
- Add DELETE /entities/:name and POST /entities/:name/merge endpoints
- Fix query limit: listEvents/listFacts/listStatuses accept limit param
(was hardcoded at 50, now configurable up to 500)
- Add query param auth fallback (req.query.key) for dashboard SSE/graph
- Add 'general' catch-all category to client briefing endpoint
- Add temporal validity post-filter for keyword/graph search results
- Rebrand dashboard to Zengram, fix entity/graph auth in dashboard HTML
- Bump MCP server version to 2.5.1
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
0 commit comments