2024 — present · Solo build

Synapse Notes

Multi-tenant knowledge API that decomposes notes into atomic, vector-indexed parts via LLM.

Synapse Notes is a multi-tenant knowledge API built around one idea: a note is not an opaque blob. When a note is created or updated, the system decomposes it automatically into atomic semantic parts — each part is independently vector-indexed so search finds the right concept, not just the right document.

API surface

Synapse exposes a REST API, an MCP (Model Context Protocol) endpoint for agent tool-use, and OAuth 2.1 for authentication. Clients range from a web UI to coding-agent integrations that use Synapse as long-term memory for agent sessions.

Stack

Built in Go with pgvector for semantic search on PostgreSQL. MinIO handles file storage and Redis backs the job queue for async decomposition tasks. Deployed on a self-hosted Kubernetes cluster with automated rolling deploys.