Automated Financial Intelligence and Risk Analysis Platform
Nebras transforms raw financial data into executive-ready insights and professional PDF reports. The platform combines deterministic data processing with AI-assisted narrative generation to deliver CFO-level reporting in seconds.
- Intelligent Ingestion Automatically detects transaction ledgers or P&L statements and normalizes column names
- Deterministic Analysis Calculates KPIs with zero tolerance for computational error
- Risk Detection Identifies anomalies such as negative margins and expense spikes
- Executive Reporting Generates board-ready PDF reports with AI-drafted recommendations
High-level system architecture (Frontend → API → Processing → Output).
See Architecture Documentation for details.
End-to-end processing flow from upload to PDF delivery.
| Step | Description |
|---|---|
| 1. Upload | User submits CSV or Excel file |
| 2. Validate | File type, size, and rate limit checks |
| 3. Detect | Automatic schema identification |
| 4. Analyze | KPI calculation and trend analysis |
| 5. Scan | Risk pattern detection |
| 6. Generate | PDF report with AI narrative |
| 7. Deliver | JSON response with download URL |
See System Flow for the complete processing pipeline.
| Layer | Technology |
|---|---|
| Frontend | Vue 3, Vite, SCSS |
| Backend | FastAPI, Python |
| Analysis | Pandas, NumPy |
| AI Layer | LLM-based narrative generation |
| Reporting | Playwright, Jinja2 |
cd server
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -r requirements.txt
python -m playwright install chromium
uvicorn main:app --reload --host 0.0.0.0 --port 8000npm install
npm run devThe application will be available at:
- Frontend: http://localhost:5173
- Backend API: http://localhost:8000
| Schema | Required Columns |
|---|---|
| Transactions | Date, Amount, Category |
| P&L | Month, Revenue, Expenses |
Column names are flexible the system recognizes common variants automatically.
| Endpoint | Method | Description |
|---|---|---|
/api/analyze |
POST | Upload and analyze financial data |
/reports/{filename} |
GET | Download generated PDF |
/health |
GET | API status check |
See API Reference for full documentation.
- Ephemeral Processing Data processed in memory, discarded after analysis
- Rate Limiting IP-based request throttling
- Input Validation Strict file type and size enforcement
This project is licensed under the MIT License.