GoRAG is a production-ready Retrieval-Augmented Generation (RAG) framework built for high-scale AI engineering. It features a three-layer architecture that serves developers at all skill levels.

🏗️ Three-Layer Architecture

graph TB
    subgraph Pattern层["Pattern Layer (Application)"]
        P1["NativeRAG<br/>Vector Retrieval"]
        P2["GraphRAG<br/>Knowledge Graph"]
    end
    
    subgraph Pipeline层["Pipeline Layer (Assembly)"]
        I["Indexer"]
        R["Retriever"]
        D["Repository"]
    end
    
    subgraph Step层["Step Layer (Function)"]
        S1["Independent Modules<br/>Rewrite/Chunk/Embed..."]
    end
    
    Pattern层 --> Pipeline层
    Pipeline层 --> Step层

✨ Key Features

• 🚀 Performance First: Concurrent workers and streaming parsers with O(1) memory efficiency
• 🏗️ Pipeline-Based Architecture: Every step is explicit, traceable, and pluggable
• 🧠 Three-Phase Enhancement: Query enhancement → Retrieval → Result enhancement
• 🕸️ Advanced GraphRAG: Native support for Neo4j, SQLite, and BoltDB
• 🔭 Built-in Observability: Comprehensive distributed tracing
• 📦 Zero Dependencies: Pure Go implementation with auto-download models

🚀 Quick Start

NativeRAG (Vector Retrieval)

Best for document QA and semantic search:

import "github.com/DotNetAge/gorag/pkg/pattern"

// Create a NativeRAG with auto-configuration
rag, _ := pattern.NativeRAG("my-app",
    pattern.WithBGE("bge-small-zh-v1.5"),
)

// Index documents
rag.IndexDirectory(ctx, "./docs", true)

// Retrieve
results, _ := rag.Retrieve(ctx, []string{"What is GoRAG?"}, 5)

GraphRAG (Knowledge Graph)

Best for complex relationship reasoning:

rag, _ := pattern.GraphRAG("knowledge-graph",
    pattern.WithBGE("bge-small-zh-v1.5"),
    pattern.WithNeoGraph("neo4j://localhost:7687", "neo4j", "password", "neo4j"),
)

// Add nodes and edges
rag.AddNode(ctx, &core.Node{ID: "person-1", Type: "Person", ...})
rag.AddEdge(ctx, &core.Edge{Source: "person-1", Target: "company-1", ...})

// Query neighbors
neighbors, edges, _ := rag.GetNeighbors(ctx, "person-1", 1, 10)

📚 Documentation

Getting Started

• Quick Start Guide - Pattern layer in 15 minutes
• NativeRAG Details - Three-phase enhancement
• GraphRAG Details - Knowledge graph reasoning
• Options Reference - All configuration options

Advanced Topics

• Development Guide - Pipeline layer development
• Indexer Development - Build custom indexers
• Retriever Development - Build custom retrievers

Step Layer

• Step Development Guide - Create new steps

🔭 Built-in Observability

idx, _ := indexer.DefaultAdvancedIndexer(
    indexer.WithZapLogger("./logs/rag.log", 100, 30, 7, true),
    indexer.WithPrometheusMetrics(":8080"),
    indexer.WithOpenTelemetryTracer(ctx, "jaeger:4317", "RAG"),
)

⚡ Technical Standards

• Go 1.24+: Latest language features
• Zero-CGO SQLite: Painless cross-compilation
• Clean Architecture: Strict separation of interfaces and implementations
• Modular Steps: Reuse steps in any custom pipeline

🤝 Contributing

We aim to build the most robust AI infrastructure for the Go ecosystem.

• Check our Contributing Guidelines

📄 License

GoRAG is licensed under the MIT License.