Ask natural questions. Get real insights. Instantly.
Conversational Business Intelligence for your E-commerce data.
Experience the AI E-commerce Co-Pilot in action:
👉 Click here to open the live demo
AI E-commerce Analytics Co-Pilot is your intelligent data assistant. Built for business users and analysts, it lets you ask natural questions like:
“What was the click-through rate of my top 5 products last week?”
And instantly receive:
- Clean tabular results
- A visual chart
- A plain-English explanation from the AI
All without writing a single SQL query.
This tool brings real-time insights to your fingertips by connecting LLMs to your structured e-commerce data.
Ask questions in plain English. The system converts them into SQL, runs the query on your database, and returns rich, interactive results.
Built on LangChain, the agent uses either:
- Google Gemini Pro API (in production)
- Phi-3 via Ollama (locally)
...to generate SQL and explain results clearly.
User input is classified into:
- Queries (generate SQL)
- Casual chat (friendly replies)
- Off-topic input (graceful fallback)
This avoids unnecessary LLM calls and ensures a fast experience.
If a generated query fails (e.g., bad column name), the agent automatically:
- Analyzes the error
- Fixes the query
- Re-executes it
No need for user retries.
- Auto-generated bar/line/pie charts using Chart.js
- Full data tables with sorting
- Optional CSV export
- Only
SELECTqueries allowed — no mutations - Unsafe SQL is blocked at validation
- Prompt is hard-coded with schema & rules
- Metrics like RoAS use
NULLIFto avoid crashes
| Layer | Technologies |
|---|---|
| Frontend | React, TypeScript, Vite, Chart.js, Bootstrap, Lucide React |
| Backend | FastAPI, Python, LangChain, SQLAlchemy, Pandas |
| LLM Options | Phi-3 via Ollama (locally), Google Gemini API (deployed) |
| Database | SQLite |
| Deployment | Render, .env-based configuration |
