▌ honcho-memory.md

# Honcho Memory Integration ## What Honcho is a memory/context layer for AI agents — stores conversation context, user profile, and facts that persist across sessions. ## Setup (May 21, 2026) - Workspace: `default` - Peer: `nick-williams` - API key: stored in `~/.hermes/.env` as `HONCHO_API_KEY` - Python: Direct `requests` calls (not honcho SDK — SDK requires Python 3.10+, Nick's is 3.9.6) ## API Details - Base URL: `https://api.honcho.dev/v3` - Auth: Bearer header - Endpoints: - `POST /workspaces/{workspace}/sessions` → create session - `POST /workspaces/{workspace}/sessions/{id}/messages` → add messages with peer_id - `POST /workspaces/{workspace}/peers/{peer_id}/chat` → query with `{"query": "...", "top_k": 5}` ## Stored Profile Nick Williams — HK (UTC+8), short/direct communication, calls agent "Henry" Key projects: Comparative Data YT channel (daily World Bank videos 1080x1920), EPL betting AI (Poisson model), HKJC horse racing Wed/Sun, Beepbo SEO, AI CV on Railway (MiniMax-M2.7), GeBeCert anti-counterfeiting SaaS, LinkedIn job search (remote APAC, no Mandarin) Preferences: Adobe After Effects credit in description body (not just hashtags), rough human-sounding cover letters, OpenCode CLI for code reviews, independent data insights (not tipsters) ## Status Set up and tested (May 21, 2026) — queried back successfully. Not yet wired into cron jobs or daily workflow. Nick said "ok" to integrating — pending follow-through. ## Pending - Wire Honcho into daily cron jobs for auto-storing context - Use for session summaries after each conversation - Consider as primary memory layer (replacing or complementing MemPalace)