nano-brainlisted

# nano-brain — Persistent Memory + Code Intelligence A read-mostly knowledge graph + vector DB. Three call surfaces: HTTP, CLI (`npx nano-brain ...`), MCP tools. All operations require a `workspace_hash` that maps to one project root path. ## TL;DR ```bash # Bootstrap once per shell: eval "$(npx nano-brain workspaces current --export)" # Then query: npx nano-brain query "how did we solve auth caching" ``` All CLI commands read `$NANO_BRAIN_WORKSPACE` so you do not need to pass `--workspace=...` after bootstrap. For HTTP/MCP direct calls, see Phase 3. --- ## Phase 1 — DISCOVER (bootstrap) Run these steps once per shell session. They are idempotent — safe to re-run. ### 1.1 Verify server is reachable **From a container** (most agents): ```bash curl -fsS http://host.docker.internal:3100/api/status | jq -r .pg_status ``` **From the host directly:** ```bash curl -fsS http://localhost:3100/api/status | jq -r .pg_status ``` **Success criterion:** prints `healthy`. **Failure recovery:** If `curl` fails, the server is not running. Recovery in order: 1. On host: `npx nano-brain docker start` 2. Ask the user to start the server. **Do not** try to start the server yourself inside a container — see "nano-brain Server Rule" in the project AGENTS.md. ### 1.2 Resolve the current workspace hash **Recommended — CLI bootstrap one-liner:** ```bash eval "$(npx nano-brain workspaces current --export)" ``` This sets `NANO_BRAIN_WORKSPACE` for the rest of the shell. All subsequent