self-hosted 
Agent Memory Part 2: Hybrid Search, Graph Recall, and Memory Consolidation
We built the 'What's Next' — hybrid BM25+vector search, graph-aware recall, auto-ingest, and extractive memory consolidation. 107 tests, zero frameworks.
Zero-Cost AI Agent Memory: Local GPU Embeddings with SurrealDB and OpenClaw
How I gave my AI agent semantic memory using a local GPU, SurrealDB, and Qwen3 embeddings — zero API costs, zero data leaving my machine.
How To: Claude Opus 4.6 1 Million Token Context Running On OpenClaw
Step-by-step guide to enabling 1 million token context with Claude Opus 4.6 on a self-hosted OpenClaw gateway. Includes the 10 issues I hit and how to fix each one.