- Frontend Delete button (kanban.js) called DELETE /api/kanban/tasks/{id}
but no such route existed; FastAPI returned 405 Method Not Allowed,
so tasks (including 'done' ones) could never be deleted from the board.
- Added the missing route: removes the task's JSON file, cleans up any
links on other tasks that pointed to it (no orphaned parent/child
pointers), audits the action, returns {status:deleted}.
- Verified end-to-end against the running server (create -> delete -> 404).
Also captures current working state across brain/, skills/, agents/,
data/, tests/, and docs.
- brain/graph/learn-loop.sh: rebuild+cluster -> (LLM-tier) deep-extract corpus
-> merge -> measure & heuristically optimize (orphan rate, community balance,
size) -> record metrics.json + learnings.md -> sync to Windows
- brain/corpus/{diagrams,designs,notes}/: drop sources for the brain to learn
- cron agentic-os-brain-rebuild now runs the learn-loop (compounding)
- README documents the learning tiers (offline structural vs LLM semantic)
Verified: 614 nodes, 1024 links, 0.2% orphans after optimization; no node loss.
- serve-brain-wsl.sh: serves WSL central brain over HTTP MCP :8090 (independent of Windows brain)
- keepalive-brain-wsl.sh: restarts serve if :8090 drops
- cron agentic-os-brain-keepalive-wsl every 5 min mirrors Windows keepalive
- README documents the two independent :8090 endpoints (WSL + Windows)
- brain/graph/sync-to-windows.sh: rebuild WSL brain + push merged graph to C:/Users/Austin/Brains/central
- cron wrapper now also syncs to Windows hourly
- README documents the two-brain layout
- Windows side (C:/Users/Austin/Brains/central): own venv + graphifyy[mcp],
serve-brain.bat (HTTP MCP :8090), keepalive-brain.bat — fully offline/standalone
- brain/graph/: build-central.sh (code graph + merge linked projects),
link-project.sh (register a project graph into the brain),
serve-brain.sh (HTTP MCP :8090), README.md
- server.py: mount /brain-graph so the brain is browsable from the dashboard
- venv: graphify [mcp] extra installed so HTTP MCP serve works
- hourly rebuild cron (agentic-os-brain-rebuild)
Topology: each project keeps its own graph; they roll up into one
central-graph.json that agents query over MCP. Memory notes become
queryable later via an LLM key (graphify . without --code-only).