🚀 GoRAG - Your Local RAG Assistant in Go

GoRAG is a simple yet powerful RAG (Retrieval-Augmented Generation) application written in Go. It allows you to create a smart "knowledge base" from your documents (PDF and Markdown) and interact with it through a modern web interface or API. Everything runs locally using Ollama, ensuring your data privacy.

✨ Key Features

• Multi-Format Extraction: Automatically processes PDF and Markdown files from your documents directory.
• Modern Web UI: Beautiful, responsive chat interface with real-time response streaming.
• Persistent Vector Store: Powered by BadgerDB, providing high-performance local storage (no more massive JSON files).
• Source Citations: The assistant cites specific sources (e.g., [Source 1]) in its answers so you can verify the information.
• Smart Chunking: Text is split into chunks with configurable overlap to preserve context across boundaries.
• Docker Ready: Complete docker-compose.yml included for easy deployment with Ollama.
• Graceful Shutdown: Cleanly handles termination signals to ensure data integrity.
• Index Management CLI: Build your index on demand with a dedicated flag.
• Structured Logging: Uses Go's slog for clean, professional logging.

🏁 Getting Started

Prerequisites

• Go: Version 1.25 or higher.
• Ollama: Installed and running with your preferred model (e.g., llama3).

  ollama pull llama3

Installation and Setup (Local)

1. Clone the repository:

   git clone https://github.com/podanypepa/gorag.git
   cd gorag

2. Add your documents: Place your .pdf or .md files into the docs/ directory.

3. Build the Index:

   go run . --index

4. Run the Server:

   go run .

   Visit http://localhost:9090 in your browser.

Installation and Setup (Docker)

1. Start the containers:

   docker-compose up -d

2. Download the model (first time):

   docker exec -it gorag-ollama-1 ollama run llama3

3. Index your documents:

   docker-compose run gorag ./gorag --index

⚙️ Configuration

Variable	Description	Default Value
OLLAMA_URL	The URL of the Ollama API instance.	http://localhost:11434
MODEL_NAME	The name of the model to use.	llama3
INDEX_DIR	The directory for the BadgerDB store.	index_db
PDF_DIR	The directory containing documents.	docs/
SERVER_PORT	The port for the web server.	9090

🔌 API Usage

Ask questions via simple GET requests:

curl "http://localhost:9090/ask?q=What are the key findings?"

🖥️ CLI Client

A simple streaming CLI client is available in bin/client.

cd bin/client
go build -o ollama_stream .
./ollama_stream "Explain quantum computing."

---

This project explores the power of local RAG systems using Go. Secure, fast, and fully local.