RAGBENCH

Evaluate · Compare · Improve your RAG pipeline

View on GitHub
MIT License · Open Source · Free

What You Get

ORGANIZATION

Projects

Organize evaluations by project. Keep configs, documents, and runs grouped.

KNOWLEDGE

Documents

Upload PDFs and text files. Chunk, embed and index into Qdrant vector store.

INTERACTIVE

Chat

Chat with your indexed documents. Test retrieval quality interactively.

METRICS

Evaluate

Run evaluation with RAGAS metrics: faithfulness, answer relevancy, contextual precision & recall.

INSIGHTS

Compare

Radar chart comparison of two eval runs. See which config wins on each metric.

TUNING

Pipeline Config

Tune chunking, retrieval mode, LLM model, and embedding model. Enter your API key in settings.

Quick Start

Get running in under 60 seconds

RECOMMENDED

Option A · Instant Run with Docker Hub

Backend Image Frontend Image

No cloning, Python, or Node needed. Download the pre-configured compose manifest and launch prebuilt containers immediately:

curl -O https://raw.githubusercontent.com/paramjeetn/RagBench/main/docker-compose.hub.yml
docker compose -f docker-compose.hub.yml up -d
Backend on :8000 Frontend on :3000 Postgres on :5432 Qdrant on :6333

Option B · Clone & Build from Source

01

Install Docker

Docker Desktop is the only requirement. No Python, no Node — Docker handles everything.

Get Docker Desktop
02

Clone & Configure

Clone the repo and copy the example env file. Add at least one LLM API key.

git clone https://github.com/paramjeetn/RagBench
cd RagBench
cp .env.example .env

Then open .env and add your API key (Gemini is free at aistudio.google.com)

Gemini (Free)OpenAIAnthropic
03

Launch

One command starts everything: Postgres, Qdrant, backend, and frontend.

make up

Then open localhost:3000 in your browser

Backend on :8000 Frontend on :3000 Postgres on :5432 Qdrant on :6333

API Keys & Pipeline Config

Once the app is running, click Pipeline Settings in the sidebar to configure your LLM, embedding model, and enter API keys. Keys are stored in your browser — never on any server.

Browser-Level Storage · 100% Private
  • ⚙
    Chunking strategy
    recursive / fixed / semantic
  • 🔍
    Retrieval mode
    hybrid / dense / sparse
  • 🤖
    LLM model
    Gemini · OpenAI · Anthropic · Ollama
  • 📐
    Embedding model
    local fastembed, OpenAI, Gemini
  • 🔑
    API keys
    browser-only, never stored on server

Useful Commands

$make up# Start all services
$make down# Stop all services
$make logs# Tail all logs
$make logs-backend# Backend logs only
$make clean-data# Wipe DB + Qdrant (fresh start)
$make clean-slate# Full rebuild from scratch