New Bot/PA Setup with Cognee Memory Mapping

# New Bot/PA Setup with Cognee Memory Mapping ## Overview When setting up a new bot instance on the OpenClaw server, follow this guide to enable Cognee memory integration. ## Prerequisites - OpenClaw instance installed - User account created (e.g., ada, kimi, mabel, clarence) - Memory folder created at `/home/[username]/.openclaw/workspace/memory/` ## Cognee Setup ### 1. Memory Folder Create the memory directory: ```bash mkdir -p /home/[username]/.openclaw/workspace/memory ``` ### 2. Cognee Virtual Environment Install Python venv and Cognee: ```bash python3 -m venv /home/[username]/.cognee-venv /home/[username]/.cognee-venv/bin/pip install cognee ``` ### 3. API Key Setup The Cognee API uses OpenAI (get key from Simon). Environment variables needed: - LLM_API_KEY - LLM_MODEL=gpt-4o-mini - LLM_PROVIDER=openai ## Cron Jobs Set up two cron jobs for each user: - Daily Worklog at 5pm - Cognee Ingest every 4 hours ## Querying Memory Use the Cognee API at http://192.168.50.230:18888 curl http://192.168.50.230:18888/query?instance=USERNAME&q=your+question ## Testing 1. Add markdown files to memory folder 2. Run: curl http://192.168.50.230:18888/add?instance=USERNAME 3. Query: curl http://192.168.50.230:18888/query?instance=USERNAME&q=test ## Notes - Each instance has separate memory dataset in Cognee - Memory files should be .md format - Cognee indexes content every 4 hours via cron