🚀 Full Project Setup & Startup Guide (Backend + Streamlit Frontend)

This project powers the Power BI Insights Assistant — a dual-stack system built with:

• Backend: Node.js + Express server for Power BI embedding and Azure OpenAI communication
• Frontend: Streamlit app for conversational insights on Power BI data

---

Prerequisites

Before running the project, make sure you have:

Requirement	Description
Node.js	Version 18 or newer
npm	Comes with Node (check with npm -v)
Python 3.10+	For Streamlit frontend
Azure OpenAI access	Endpoint, API key, deployment name, and API version
Power BI Workspace access	For embedding and exporting report visuals
config.json	Contains your Power BI workspace & report IDs

---

1. Install dependencies

Backend (Node.js)

cd backend
npm install

Frontend (Streamlit)

pip install -r requirements.txt

💡 The node_modules/ folder is excluded from Git via .gitignore. If missing, just run npm install.

---

🔑 2. Create a .env file (for backend)

Create a .env file inside the backend folder:

# Azure OpenAI setup
AZURE_OPENAI_ENDPOINT=<your-endpoint>
AZURE_OPENAI_API_KEY=<your-api-key>
AZURE_OPENAI_DEPLOYMENT=<your-deployment-name>
AZURE_OPENAI_API_VERSION=<your-api-version>

# (Optional) Power BI configuration
WORKSPACE_ID=<your-workspace-id>
REPORT_ID=<your-report-id>

Note The endpoint must only contain the base URL — no /openai/deployments/... in it.

---

3. Run both servers

Start backend:

cd backend
npm start

Backend runs by default on http://localhost:5300.

Start Streamlit frontend:

cd streamlit_app
streamlit run main.py

Frontend runs on http://localhost:8501.

---

4. Verify backend connection

Check these endpoints in your browser:

http://localhost:5300/debug/azure/env
http://localhost:5300/debug/azure/deployments

✅ If they respond with JSON, your backend and Azure connection are working.

---

💬 5. How it works

1. Frontend (Streamlit) embeds your Power BI report inside a dashboard panel.
2. When Power BI loads or filters change, the frontend exports all visuals as CSV.
3. The frontend auto-sends them to the backend:

   POST /api/visual-data/upload
   

4. When the user asks a question (e.g., “Why did arrivals drop this week?”), Streamlit calls:

   POST /api/ask
   

5. The backend:
   • Retrieves cached CSVs
   • Summarizes them via chatHelper.js
   • Calls Azure OpenAI using your .env credentials
   • Returns a concise, data-grounded answer

---

📂 Project structure

├─ backend/
│  ├─ src/
│  │  ├─ server.js             # Main Express server
│  │  ├─ aiHelper.js           # Azure OpenAI integration + caching
│  │  ├─ chatHelper.js         # CSV summarization logic
│  │  ├─ embedConfigService.js # Power BI embed config
│  │  └─ utils.js              # Helper utilities
│  ├─ public/                  # Static JS/CSS for embedding
│  ├─ package.json             # Backend dependencies
│  ├─ .env.example             # Example Azure OpenAI vars
│  └─ .gitignore               # Ignores node_modules, env, logs
│
├
│─ main.py                  # Streamlit entrypoint  
│─ mockinsights.json        # Fallback mock data (if backend offline)
│─ requirements.txt         # Python dependencies
│  
│
├─ .gitignore
└─ README.md                   # (This file)

---

6. Useful commands

Command	Description
npm install	Install backend dependencies
npm start	Run backend server
pip install -r requirements.txt	Install Streamlit dependencies
streamlit run main.py	Run Streamlit frontend
git rm -r --cached node_modules	Remove tracked node_modules
nodemon src/server.js	Auto-reload backend on change

---

Common issues

Symptom	Cause	Fix
Error: Azure call failed [404]	Endpoint includes /openai/deployments/...	Use only base URL in .env
Error: Azure OpenAI env vars missing	.env not found	Restart after adding .env
401 Unauthorized	Wrong header type for API Management	Use Ocp-Apim-Subscription-Key (auto-handled)
Cannot find module 'node-fetch'	Dependency missing	Run npm install node-fetch@2
Streamlit shows “Backend error”	Backend not running or port mismatch	Check backend logs and backend_url in frontend

---

Development tips

• Auto-restart backend:

  npm i -g nodemon
  nodemon src/server.js

• Test Azure call manually:

  curl -i -X POST "$AZURE_OPENAI_ENDPOINT/openai/deployments/$AZURE_OPENAI_DEPLOYMENT/chat/completions?api-version=$AZURE_OPENAI_API_VERSION"     -H "Ocp-Apim-Subscription-Key: $AZURE_OPENAI_API_KEY"     -H "Content-Type: application/json"     -d '{ "messages": [{"role":"user","content":"ping"}] }'

---

✅ Summary

1. cd backend && npm install
2. Add .env with Azure credentials
3. npm start
4. cd streamlit_app && pip install -r requirements.txt
5. streamlit run main.py
6. Visit http://localhost:8501 and chat with your Power BI dashboard 🚀