phoenix

Phoenix RAG Implementation

This directory contains a Retrieval Augmented Generation (RAG) implementation using Phoenix for tracing and monitoring.

Features

• Document loading from text files
• Vector storage with ChromaDB
• OpenAI integration for embeddings and text generation
• Phoenix tracing with OpenTelemetry
• Interactive query interface
• Persistent vector storage
• OpenInference instrumentation for OpenAI calls

Setup Guide

1. Environment Setup

Choose one of the following methods:

Using Conda (Recommended)

# Create environment from yml file
conda env create -f environment.yml

# Activate the environment
conda activate phoenix-rag

Using Pip

# Create conda environment
conda create -n phoenix-rag python=3.10 -y

# Activate the environment
conda activate phoenix-rag

# Install requirements
pip install -r requirements.txt

2. Environment Configuration

1. Create a copy of the example .env file:

cp .env.example .env

2. Update the .env file with your credentials:

OPENAI_API_KEY=your_openai_api_key
PHOENIX_CLIENT_HEADERS='api_key=your_phoenix_api_key'
PHOENIX_COLLECTOR_ENDPOINT='https://app.phoenix.arize.com'

3. Prepare Documents

1. Create a data directory if it doesn't exist:

mkdir -p data

2. Add your text documents to the data directory:

• Files should be in .txt format
• Each file should contain relevant content for your RAG system
• Example structure:

  data/
  ├── document1.txt
  ├── document2.txt
  └── about_phoenix.txt
  

Running the Project

1. Initial Document Loading

Before first use, load your documents into the vector store:

python rag_phoenix.py --load

This command will:

• Read all .txt files from the data directory
• Create embeddings using OpenAI
• Store vectors in the ChromaDB database
• Save the database in the chroma_db directory

2. Interactive Query Mode

Start the interactive query interface:

python rag_phoenix.py

3. Available Commands

In interactive mode:

• Type your question and press Enter to get an answer
• Special commands:
  • exit: Quit the program
  • reload: Reload documents from the data directory
  • Empty line: Skip and continue

4. Monitoring

1. View traces in Phoenix:

   • Visit https://app.phoenix.arize.com
   • Navigate to your project (phoenix-rag-demo)
   • View traces, embeddings, and LLM calls

2. Monitoring features:

   • Track OpenAI API calls
   • Monitor embedding generation
   • Analyze retrieval performance
   • View full conversation traces

Project Structure

phoenix/
├── rag_phoenix.py      # Main implementation
├── requirements.txt    # Pip dependencies
├── environment.yml     # Conda environment
├── .env               # Environment variables
├── data/              # Text documents
│   └── *.txt
└── chroma_db/         # Vector database

Troubleshooting

1. OpenAI API Issues:

   • Verify your API key in .env
   • Check API usage limits
   • Ensure internet connectivity

2. Document Loading Issues:

   • Verify text files are in UTF-8 format
   • Check file permissions
   • Ensure files are in .txt format

3. Vector Store Issues:

   • Delete chroma_db directory and reload if corrupted
   • Check disk space
   • Verify ChromaDB installation

4. Phoenix Tracing Issues:

   • Verify Phoenix API key
   • Check collector endpoint
   • Ensure proper environment variable setup

Dependencies

Major dependencies and their versions:

• arize-phoenix-otel >= 0.6.1
• openai >= 1.0.0
• langchain >= 0.1.0
• chromadb >= 0.4.0
• openinference-semantic-conventions >= 0.1.12

For a complete list, see requirements.txt.