buer103/kairos

Kairos is a Python AI agent framework for building stateful, multi-turn agents and production integrations. It combines ideas from Hermes and DeerFlow and adds a RAG engine, structured knowledge, evidence chains, confidence/citation, credential pooling, tiered memory and a plugin system. Who uses it: developers and teams who need configurable agents, multi-platform gateways, tool integration, or RL/observability in production. Key features: - ReAct agent loop with 20-layer middleware pipeline and 24 lifecycle hooks - 23 built-in tools (file, terminal, browser, vision, RAG, delegate) and custom tool registration - 17 provider profiles (OpenAI, Anthropic, Gemini, DeepSeek, etc.) with credential failover - Tiered memory (profile/timeline/facts) with FTS5 search and confidence filtering - Gateway adapters (11 platforms), CLI/TUI, sandboxing, security policies, observability (Prometheus), sub-agent orchestration, RL training, Docker deployment - Install: pip install kairos-agent. Beta v0.16.0 with tests and examples included.

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.