Building a Local AI-Powered KPI Dashboard with Streamlit and Ollama⌗

In this project, I built a sleek, brand-aligned business intelligence dashboard powered entirely by local AI — no API keys, no cloud dependencies. Just Python, Streamlit, and Ollama running smart and fast right on my machine.

🎯 The Goal⌗

Create a lightweight yet powerful app that lets users:

• Upload or use built-in business KPI data
• Visualize revenue, profit, and conversion trends
• Generate AI-powered summaries and insights
• Ask follow-up questions using natural language
• Export clean PDF reports — charts included

🔧 The Tech Stack⌗

🧱 What the App Does⌗

📊 1. Upload or Simulate Business Data⌗

The app loads default revenue, expenses, leads, and client data — or allows the user to upload a custom CSV. It auto-generates calculated fields like:

• Profit = Revenue - Expenses
• Conversion Rate = Clients ÷ Leads

📈 2. Visualize KPIs⌗

Three clean, responsive charts are generated with Matplotlib:

• Revenue vs. Expenses (Line Chart)
• Monthly Profit (Bar Chart)
• Conversion Rate (Line Chart)

🧠 3. Generate AI Summaries⌗

Click a button and get a GPT-style summary of your data using Ollama + a locally-running Mistral model — no API required.

Summaries are cached in session_state, so they persist between interactions and don’t get overwritten by the chat module.

💬 4. Ask the Model Business Questions⌗

Type in a natural language question like:

“What’s the trend in profit over the past 3 months?”

The app feeds both the KPI data and the question to Ollama and returns a contextual, AI-generated answer.

📄 5. Export PDF Reports⌗

Generate branded PDF reports including:

• Title and summary section
• Embedded KPI charts
• Styled with your brand colors

📥 6. Download Raw Data⌗

Export your data as a CSV at any time.

🧠 Why It’s Cool⌗

• Offline AI: Runs entirely on your machine with Ollama (tested with Mistral, Gemma, and Phi)
• Customizable: Plug in any dataset or model
• Scalable: Easy to repurpose for client reporting, SaaS, or internal BI tooling
• Free and Private: No API usage, no cloud costs, no data leaving your machine

🚀 Next Steps⌗

I’ll continue refining it by:

• Adding markdown rendering in PDFs
• Supporting advanced formatting (headings, lists)
• Packaging it for internal use or monetization

Stay tuned — or reach out if you want help building something like this for your team.