LangFlow Streamlit Integration with RAGAS Evaluation

This project integrates LangFlow as a backend API with a Streamlit frontend for a chatbot interface. It also includes RAGAS evaluation for measuring the performance of RAG (Retrieval-Augmented Generation) pipelines.

Project Structure

project/
├── api/                  # FastAPI server that connects to LangFlow
├── chatbot/              # Streamlit application
├── evaluation/           # RAGAS evaluation tools
├── data/                 # Data storage for evaluation
├── .env                  # Environment variables
├── requirements.txt      # Project dependencies
├── docker-compose.yml    # Docker configuration

Getting Started

Langflow Graph

Prerequisites

• Docker and Docker Compose
• Python 3.10+

Installation

1. Clone this repository

2. Create a virtual environment and install dependencies:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   pip install -r requirements.txt

3. Create a .env file with the following content:

   LANGFLOW_API_URL=http://localhost:7860
   API_PORT=8000
   DEBUG=True
   

Running the Application

Using Docker

The easiest way to run the entire stack is with Docker Compose:

docker-compose up -d

This will start:

• LangFlow on port 7860
• The API server on port 8000
• The Streamlit UI on port 8501

Running Locally

1. Start LangFlow:

   docker run -p 7860:7860 logspace/langflow:latest

2. Start the API server:

   uvicorn api.app:app --reload

3. Start the Streamlit application:

   streamlit run chatbot/app.py

4. Start the evaluation dashboard:

   streamlit run evaluation/metrics.py

Using the Application

Setting Up LangFlow

1. Access LangFlow at http://localhost:7860
2. Create a new flow using the drag-and-drop interface
3. Set up your RAG pipeline with appropriate components:
   • Document loaders
   • Vector stores
   • LLM models
   • Chain components
4. Deploy your flow

Using the Chatbot

1. Access the Streamlit UI at http://localhost:8501
2. Select your flow from the dropdown in the sidebar
3. Start chatting with your LangFlow-powered application

Running Evaluations

1. Access the evaluation dashboard at http://localhost:8501/evaluation
2. Select the flow you want to evaluate
3. Click "Run Evaluation" to test your flow with RAGAS metrics
4. Review the results and optimize your flow accordingly

RAGAS Evaluation

This project uses RAGAS to evaluate the performance of your RAG pipelines with the following metrics:

1. Faithfulness: Measures how factually consistent the generated answer is with the retrieved context
2. Answer Relevancy: Evaluates whether the answer addresses the question
3. Context Relevancy: Assesses the quality of retrieval - how relevant the retrieved context is to the question
4. Context Recall: Measures how well the retrieved context covers the information needed to answer the question
5. Harmfulness: Evaluates the safety of the generated response

Customizing the Application

Adding Custom Evaluation Questions

Edit or replace the data/questions.json file with your domain-specific questions and ground truth answers.

Extending the API

The API is built with FastAPI, making it easy to add new endpoints:

1. Open api/app.py
2. Add new route functions using the FastAPI decorator syntax
3. Implement your endpoint logic

Customizing the UI

The Streamlit UI can be customized:

1. Edit chatbot/app.py to adjust the main application flow
2. Modify components in the chatbot/components/ directory
3. Add new utility functions as needed

Evaluation Insights

The evaluation dashboard provides valuable insights into your RAG pipeline performance:

• Metric comparison across different flows
• Historical performance tracking
• Detailed view of evaluation results
• Areas for improvement identification

Use these insights to iteratively improve your LangFlow pipelines:

1. Identify metrics with lower scores
2. Adjust relevant components in your flow
3. Re-run evaluations to measure improvement
4. Repeat until satisfactory performance is achieved