🔍 RAG-EngineX

RAG-EngineX is a professional-grade, modular, and fully customizable Retrieval-Augmented Generation (RAG) framework for document-based Q&A. It features advanced chunking strategies, vector-based search, reranking with cross-encoders, LangSmith-powered observability, and detailed evaluation using ARES, faithfulness, and relevance metrics — all wrapped in an intuitive Streamlit UI.

---

🚀 Demo

🔗 Try it now: 👉 RAG-EngineX on Render

🎥 Watch Walkthrough Video (2 min)

---

✨ Features

• 📄 Multi-format Document Loader – Load PDFs and other formats with modular loader components.

• 🧩 Custom Chunking Strategies – Choose from character, recursive, or semantic chunking.

• 🧠 Pluggable Embedding Models – Support for sentence-transformers, OpenAI, or custom models.

• 📦 Vector Store Flexibility – Use FAISS, Chroma, or any LangChain-supported store.

• 🔁 Reranking Module – Improve retrieval relevance with cross-encoder rerankers.

• 🧪 Evaluation Metrics – Faithfulness, relevance, ARES, and custom scoring.

• 📊 LangSmith Integration – Trace every prompt-response interaction with latency tracking.

• 💻 Streamlit UI – Clean, interactive, and exportable frontend interface.

• 🛠️ Fully Modular Codebase – Easy to extend and integrate with other pipelines.

• 📤 Answer Export – Download results in JSON/CSV for audits or further analysis.

---

🧠 Architecture

RAG-EngineX follows a modular pipeline from loading documents to evaluating responses, with each component decoupled for full flexibility.

1️⃣ 📌 High-Level Pipeline Overview

Shows the flow of a document through the complete RAG-EngineX pipeline:

---

2️⃣ 🤖 Reranker Subsystem

The reranker improves retrieval quality using a BGE cross-encoder before passing results to the LLM:

---

3️⃣ 📊 LangSmith + Evaluation Module

Traces and evaluates generated answers using LangSmith and in-app scoring (ARES / Relevance / Faithfulness):

---

🖼 UI & Screenshots

The Streamlit UI is designed for interactivity and insight. It shows answers, sources, evaluation metrics, and supports exporting results.

🔍 LangSmith Trace Viewer

---

📈 Evaluation Metrics View (ARES / Relevance / Faithfulness)

---

🧠 Reranker Comparison View

---

🖥️ Main Streamlit App UI

---

⚙️ Tech Stack

## 📁 Project Structure

rag-enginex/
│
├── rag_enginex/             # Core pipeline logic
│   ├── loader.py            # Document loading
│   ├── chunker.py           # Chunking logic
│   ├── embedder.py          # Embedding generation
│   ├── vector_store.py      # Vector DB interaction
│   ├── reranker.py          # Cross-encoder reranker
│   ├── llm_answer.py        # LLM query and generation
│   ├── evaluator.py         # Automated eval (ARES etc.)
│   ├── evaluator_manual.py  # Manual scoring 
│   ├── llm_wrapper.py       # LLM abstraction
│   └── pipeline.py          # Orchestrates entire flow
│
├── ui.py                    # Streamlit UI
├── requirements.txt         # Python dependencies
├── runtime.txt              # Python version for Render
├── Procfile                 # Web server entry point
├── .streamlit/
│   └── config.toml          # Theme and server settings
└── assets/
    ├── LangSmith.png
    ├── Reranker.png
    ├── Evaluation Metrices.png
    ├── UI.png
    ├── pipeline_overview.png
    ├── reranker_architecture.png
    ├── langsmith_evaluation.png
    └── tech_stack.png

---

## 🛠️ Setup & Deployment

### 🔧 Local Development



 1. Clone the repo
git clone https://github.com/yourusername/rag-enginex.git
cd rag-enginex

 2. Create a virtual environment
python -m venv venv
source venv/bin/activate  # or venv\Scripts\activate on Windows

 3. Install dependencies
pip install -r requirements.txt

 4. Run locally
streamlit run ui.py

🚀 Deployment on Render.com

Full guide available here: Deploying RAG-EngineX to Render

Basic steps:

Push code to GitHub.

Create new Web Service on Render.

Set build command to :
    pip install -r requirements.txt.

Set start command to :
    streamlit run ui.py --server.port 10000.

Add runtime.txt with Python version.

Add Procfile with:

    web: streamlit run ui.py --server.port $PORT

(Optional) Set up .streamlit/config.toml to customize theme.

📤 Export & Evaluation Features:

    ✅ Faithfulness and relevance scores shown in collapsible sections.

    ✅ Evaluation results (JSON/CSV) downloadable via UI.

    ✅ Easily switch between auto-eval and manual evaluation.

✍️ Blogs & Write-ups
📘 Coming Soon on Hashnode

    “Building a Modular RAG Pipeline from Scratch”

    “How to Evaluate RAG Systems Without OpenAI”

    “Streamlit + LangChain: Perfect UI for Document QA”

🤝 Contributing
    Contributions are welcome! Feel free to submit issues or pull requests.

📜 License
    MIT License © Rahul Chauhan

🙌 Acknowledgments
    LangChain for modular building blocks

    LangSmith for observability

    HuggingFace & Sentence-Transformers for embeddings

    Papers like ARES & RAGAS for evaluation inspiration

---

---

✅ Next Steps (optional):

• I can save this as README.md and export it for your GitHub.
• Help you write the Render deployment guide mentioned above.
• Help you turn this into a blog post on Hashnode.