Summary

Agentica is a lightweight, async-first Python framework for building AI agents. It simplifies the creation of complex agent systems by providing robust support for tool integration, Retrieval Augmented Generation (RAG), multi-agent collaboration, and workflow orchestration, targeting developers seeking efficient and scalable AI agent development.

How It Works

The framework employs an async-first architecture, enabling native asynchronous operations and parallel tool execution via asyncio.gather(). It abstracts over 20 different LLM providers, including OpenAI, DeepSeek, and ZhipuAI, through a unified API. Agentica integrates over 40 built-in tools for common tasks like search, code execution, and file manipulation, alongside advanced RAG capabilities and flexible multi-agent patterns (dynamic delegation and deterministic orchestration).

Quick Start & Requirements

• Installation: pip install -U agentica
• Prerequisites: API keys for desired LLM providers (e.g., export ZHIPUAI_API_KEY="your-api-key"). Supports OpenAI, DeepSeek, ZhipuAI, Claude, Qwen, Ollama, and others.
• Resources: Minimal setup time, dependent on LLM provider API access.
• Links: Full documentation at https://shibing624.github.io/agentica. Examples are available in the examples/ directory.

Highlighted Details

• Async-First Design: Native async APIs and efficient parallel execution for tools and agent tasks.
• Broad Model Compatibility: Unified interface for 20+ LLM providers and 40+ built-in tools.
• Integrated RAG: Features knowledge base management, hybrid retrieval, and reranking, with LangChain/LlamaIndex integration.
• Multi-Agent Frameworks: Supports both dynamic "Team" delegation and deterministic "Workflow" orchestration.
• Protocol Support: Implements Model Context Protocol (MCP) and Agent Communication Protocol (ACP) for interoperability.
• Extensibility: Markdown-based Skill system for model-agnostic functionality injection.
• Deployment Options: Includes a CLI tool and a Web UI via agentica-gateway.

Maintenance & Community

• Active development is indicated by GitHub Issues for bug reporting and feature requests.
• Community support is available via a WeChat group (contact xuming624 for an invite).

Licensing & Compatibility

• Licensed under the Apache License 2.0.
• Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

The provided documentation does not explicitly detail known limitations, alpha status, or specific unsupported platforms. Users should consult the project's issue tracker for potential caveats.