Memoh-v2 offers a powerful, self-hostable AI Agent platform emphasizing true isolation, advanced memory, continuous evolution, and team collaboration. It aims to provide a superior, flexible alternative for users seeking robust AI agent capabilities without vendor lock-in.

How It Works

The platform isolates each Bot within its own containerd container, ensuring distinct environments. Memory utilizes a three-layer approach: Qdrant for semantic search, BM25 for keywords, and LLMs for extraction/summarization. Bots autonomously evolve via reflection, experimentation, and review cycles. An overseer Bot orchestrates AI teams, delegating tasks and consolidating outputs for complex workflows.

Quick Start & Requirements

Installation is simplified via a Docker script (curl -fsSL https://raw.githubusercontent.com/Kxiandaoyan/Memoh-v2/main/scripts/install.sh | sh). Docker is the primary requirement. Access the web UI at http://localhost:8082. Dependencies include PostgreSQL, Qdrant, and Containerd; Ollama is supported for local models.

Highlighted Details

• True Container Isolation: Dedicated containerd containers per Bot provide independent file systems, shells, and network access, supporting snapshot rollback.
• Multi-Layer Memory: Integrates Qdrant (vector), BM25 (keyword), and LLM (extraction) for nuanced recall, featuring automatic fact/solution extraction, time decay, and multi-language support.
• Autonomous Self-Evolution: Bots refine capabilities through a traceable reflection-experimentation-review cycle.
• AI Team Orchestration: Enables overseer-managed AI teams for complex task delegation and collaborative problem-solving.
• Zero-Cost Infrastructure: Core services like web search (SearXNG) and vector database (Qdrant) are self-hosted and free.
• Multi-Platform Integration: Connects with Telegram, Feishu, WeChat, Discord, and Web interfaces, supporting cross-platform identity and synchronization.
• Extensible Skillset: Includes 21 built-in skills and supports community skill installation.

Maintenance & Community

This project is noted as a secondary development based on "Memoh". Specific details on active maintainers, community channels, or a public roadmap are not provided in the excerpt.

Licensing & Compatibility

Licensed under GNU AGPL v3, a strong copyleft license requiring derivative works to be shared under the same terms, potentially impacting commercial integration.

Limitations & Caveats

The README mentions a "Known Limitations" section, but specifics are absent from the provided text. The AGPL v3 license itself may pose adoption constraints for certain commercial use cases.