v0.0.12026-05-01
Core platform: MCP server, knowledge graph, ingestion pipeline
MCP Server
- 15 brain.* tools shipped: context pack, search, ingest, why (provenance), neighbors (graph traversal), propose_fact, annotate, save_memory, recall_memory.
- Hybrid vector + full-text search with BM25 scoring.
- Budget governor compaction logic for token-aware context management.
- SSL self-heal fallback for all database query paths.
Knowledge Graph
- Postgres 16 + pgvector schema with multi-tenancy, workspace-scoped RLS, and bi-temporal semantics (valid_from/valid_to + recorded_at).
- Deterministic entity extraction and graph construction — LLMs are advisors, not the source of truth.
- Full audit trail on every mutation via PostgreSQL triggers.
- Ingestion pipeline: raw documents → chunked → embeddings → entity extraction → graph edges.
Apps & Tooling
- Next.js dashboard for knowledge exploration: graph visualization, vault, chat, query.
- Admin onboarding wizard for workspace setup.
- CLI with cloud and local/offline modes for zero-config onboarding.
- Python workers for ingestion, advisory extraction, connectors, schema evolution, and agent sync.