[AZ OpenAI] AzDocRAG Demo — Document-based RAG Model Platform

A lightweight demo platform for document-based Retrieval-Augmented Generation (RAG) using Azure OpenAI, LangChain, and Weaviate.

AzDocRAG Demo is an experimental platform that demonstrates how to:

• Embed and index text documents into a vector database (Weaviate)
• Retrieve relevant chunks for user queries
• Generate contextual responses using Azure OpenAI or OpenAI models
• Run end-to-end document Q&A via FastAPI backend and optional frontend

Overview

Note

Built with Azure OpenAI by default, but also works with your own OpenAI key. Easily switch between local Weaviate, cloud vector DBs, MongoDB, or CosmosDB — your setup, your choice.

---
config:
  theme: dark
---
flowchart TD
    subgraph FE[Frontend]
        ST[Streamlit UI]
    end
    subgraph BE[Backend]
        FA[FastAPI Server]
        AZO[Azure OpenAI]
    end
    subgraph DB[Database / Vector Store]
        W[Weaviate]
    end
    ST -->|HTTP Request / User Input| FA
    FA -->|Calls| AZO
    FA -->|Query / Store Data| W
    AZO -->|Response - Generated Text| FA
    W -->|Return Vectors / Metadata| FA
    FA -->|Response| ST

Loading

You must have OpenAI, and Azure OpenAI is also acceptable, as shown in the image below.

Supports RAG functionality for documents saved for testing purposes.

Development

Dependencies

Python version: 3.11 Backend: FastAPI Frontend: Streamlit

Conda (env)

# If not exists miniconda env
conda create -n rag python=3.11

# Run the environment
conda activate rag

# Initialize setup
pip install -r requirements.txt

Quickly run a demo

Prepare OpenAI and configure the .env file

You must enter your OpenAI key, model, and other required information in .env.
The .env environment file for the backend must be located within backend/src.
If you want to use an Azure OpenAI, sets AZURE_* environment, too.

Warning

The current service code only supports Azure OpenAI. For the standard OpenAI API, you must make changes directly within the service and Route.

AZURE_OPENAI_ENDPOINT=https://<Your-OpenAI_Resource-Name>.openai.azure.com/
AZURE_OPENAI_API_KEY=input-openai-api-key
OPENAI_KEY=input-openai-api-key
OPENAI_API_VERSION=20xx-01-01-preview
OPENAI_GPT_MODEL=gpt-4o-mini--type-your-gpt-model-name
OPENAI_EMBEDDINGS_DEPLOYMENT=text-embedding-3-large--type-your-deployment-name

Important

A text embedding model is also required for vectorizing sentences.

Start with make up

make up     # up docker, fast-api, streamlit

# Stop
make down

Weaviate (VectorDB)

# In root directory:
docker compose up -d

FastAPI

# Development
cd ./backend/src
python main.py

# Local
uvicorn main:app --reload

Streamlit

cd ./frontend/src
streamlit run app.py

Contact

Email: histigma01@gmail.com