Skip to content

RAGents Documentation

Welcome to RAGents - an advanced agentic RAG framework with multimodal processing and type-safe LLM interactions.

What is RAGents?

RAGents is a comprehensive framework for building intelligent agents that combine:

  • Retrieval-Augmented Generation (RAG) with multimodal processing capabilities
  • Type-safe LLM interactions using the instructor package
  • Extensible agent architectures including Decision Trees, ReAct, and Graph Planners
  • Production-ready deployment with Docker, Kubernetes, and Kubeflow support

Key Features

🤖 Multiple Agent Types

  • Decision Tree Agents: Structured reasoning with branching logic
  • ReAct Agents: Reasoning and Acting with tool integration
  • Graph Planner Agents: Complex multi-step planning and execution

📚 Advanced RAG Engine

  • Multimodal Processing: Handle text, images, PDFs, and more
  • Pluggable Vector Stores: ChromaDB, Weaviate, pgvector, Elasticsearch
  • Smart Reranking: Multiple reranking strategies for better relevance
  • Query Rewriting: DSPy-inspired query optimization

🧠 Type-Safe LLM Integration

  • Structured Outputs: Pydantic models for reliable responses
  • Multiple Providers: OpenAI, Anthropic with unified interface
  • Async Support: Built for high-performance applications

🔬 Evaluation & Observability

  • Built-in Metrics: RAGAS-style evaluation framework
  • OpenInference Tracing: Comprehensive observability
  • Structured Logging: Debug and monitor your agents

🚀 Production Ready

  • Docker Containerization: Easy deployment and scaling
  • Kubernetes Integration: Cloud-native orchestration
  • Kubeflow Pipelines: ML workflow management
  • CI/CD Ready: GitHub Actions for automated testing and deployment

Quick Start

# Install RAGents
pip install ragents

# Set your API key
export OPENAI_API_KEY="your-api-key-here"

# Run the demo
python -m ragents.demo

Architecture Overview

graph TB
    User[User Input] --> Agent[RAGents Agent]
    Agent --> LLM[LLM Client]
    Agent --> RAG[RAG Engine]
    RAG --> VS[Vector Store]
    RAG --> Proc[Document Processors]
    LLM --> Providers[OpenAI/Anthropic]
    Agent --> Tools[Tool Registry]
    Agent --> Response[Structured Response]

Getting Started

Ready to build intelligent agents? Check out our installation guide and quick start tutorial.

Community & Support

  • GitHub: ragents repository
  • Issues: Report bugs and request features
  • Discussions: Ask questions and share ideas

RAGents - Building the future of intelligent agents, one conversation at a time.