Pre-requisites

• Install Python 3.11 (3.12/13 should work but not tested)
• Windows: [https://www.python.org/downloads/windows/]
• MacOS: [https://www.python.org/downloads/macos/], or via Homebrew
• We'll use virtual env for easy setup and portability
• Install Pycharm IDE (Community)
• If you only use it for the workshop, consider applying a Free Education License

Key Libraries

The following libraries will be used in our RAG implementation:

• FastAPI: For building APIs.
• OpenAI: To integrate with OpenAI's API.
• Azure Search Documents: For document retrieval in Azure.
• PyPDF: To parse PDF documents.

Where do we start?

• Clone the RAG template repo
• Backend API Blueprint with fastapi
• Fully functional Angular frontend
• Azure service integrations in from of env. variables
• All necessary services are ready, you just have to connect

Task 1: Clone repo and setup environment

Clone the RAG template repo

1. Checkout project README and follow the setup instructions
2. Run the backend and checkout the Swagger UI

git clone https://xxx@dev.azure.com/ZuhlkeAsia/RAG%20Workshop/_git/rag-template

Task 2: Hello World API

Your first task is to implement a simple Helloworld API using FastAPI.

• Add a new route /hello that returns a JSON response with a message.
• Check your implementation by running the FastAPI server

# Use fastapi-cli to run the server in dev
fastapi dev main.py

# Alternatively, use uvicorn directly
uvicorn main:app --reload

# Or run main via PyCharm (no hot-reload)

Task 3: Create a new backend/.env file for Azure integration

Reminder Everything you need to build your API are included!

PYTHONPATH=./backend
# Azure OpenAI api-version. See https://aka.ms/azsdk/azure-ai-inference/azure-openai-api-versions
OPENAI_API_VERSION=2024-06-01
AZURE_OPENAI_ENDPOINT=https://oai-swex-rag-workshop-eastus.openai.azure.com/
AZURE_OPENAI_API_KEY=xxk
AZURE_OPENAI_EMBEDDING_MODEL=text-embedding-3-small
AZURE_OPENAI_CHAT_MODEL=gpt-4o-mini
CHUNK_SIZE=500
AI_SEARCH_ENDPOINT=https://srch-swex-rag-workshop.search.windows.net
AI_SEARCH_API_KEY=xxx
AI_SEARCH_INDEX_NAME=kevin-lin-rag-workshop # TODO: ADAPT THIS WITH YOUR NAME

🏆 Bonus Task: Add type definition to your API

• Use Pydantic to define request/response model for your API.

from pydantic import BaseModel


class HelloWorldResponse(BaseModel):
    message: str


@app.get("/", response_model=HelloWorldResponse)
def read_root():
    return {"message": "Hello, World"}