Agents-Flex

Agents-Flex is a lightweight, open-source Java framework designed for building AI agent applications. It provides a modular architecture that simplifies the development of complex AI workflows, including chatbots, Retrieval Augmented Generation (RAG) systems, multi-agent orchestration, and image generation. The framework supports integration with major large language models (LLMs) such as OpenAI, Qwen, DeepSeek, and Ollama, and offers built-in connectors for HTTP, SSE, and WebSocket protocols. It also includes advanced features like prompt engineering with pre-built templates, memory management for conversation history, function calling via a built-in engine, and support for various vector databases (PGVector, Milvus, Chroma, Redis) for semantic search.

Beyond core AI capabilities, Agents-Flex incorporates enterprise-grade features such as LLM load balancing and high availability, full observability with OpenTelemetry integration, and a LLM Wiki knowledge tree for structured data understanding. It supports web search with built-in search engines (Bocha, Brave), Text2SQL for natural language database queries, and AI Skills for encapsulating reusable business logic. The framework also offers subagent support for hierarchical task decomposition, reranking for improved RAG accuracy, and seamless integration with Spring Boot via a starter module. With its emphasis on modularity and extensibility, Agents-Flex is suitable for both simple conversational agents and sophisticated, production-grade AI applications.