Context-Hawk 🦅 — AI Context Memory Guardian

Enterprise-grade context memory management for AI agents. Works with any AI agent — universal, no business logic, open source.

What It Does

Unlike a memory database, Context-Hawk manages structured memory that survives sessions:

• Persistent task state: hawk resume continues from where you left off, even after restart
• Four-tier memory: Working → Short-term → Long-term → Archive with Weibull decay
• Structured JSON memories: importance, category, tier, L0/L1/L2 layers
• 5 context injection strategies: dynamically control how much memory enters context
• 5 compression strategies: summarize / extract / delete / promote / archive
• Self-introspection: knows when it's missing information or stuck
• LanceDB ready: vector search + BM25 hybrid retrieval (optional)
• Pure-memory fallback: works without LanceDB, JSONL persistence