Autonomous Analytics Platform

A multi-agent analytics system that ingests CSV/Excel files and produces:

• data profiling and validation
• merge outputs
• statistical analysis
• EDA visuals
• predictive modeling
• executive reporting

Architecture

The backend is built using FastAPI and follows a modular multi-agent architecture under 'src/agents/':

• ingestion_agent: upload handling, profiling, merge orchestration
• statistical_agent: hypothesis/statistical tests
• eda_agent: automatic exploratory plots + insights
• predictive_agent: model selection and forecasting/prediction
• reporting_agent: executive report generation (HTML/PDF routes)

Project Structure

src/
  main.py
  core/
  agents/
    ingestion_agent/
    statistical_agent/
    eda_agent/
    predictive_agent/
    reporting_agent/
analytiq.html
requirements.txt
data/

Prerequisites

• Python 3.9+
• pip/virtualenv recommended

Installation

python -m venv .venv
# Windows PowerShell
.\.venv\Scripts\Activate.ps1

python -m pip install -r requirements.txt

Environment Variables

Create a .env file in project root (optional for Azure/OpenAI-powered decisions):

AZURE_OPENAI_ENDPOINT=
AZURE_OPENAI_API_KEY=
AZURE_OPENAI_DEPLOYMENT_NAME=gpt-4.1-mini
AZURE_OPENAI_API_VERSION=2024-12-01-preview

If these are missing, some AI-assisted parts fall back to deterministic/default behavior.

Run Backend

uvicorn src.main:app --reload

Service URLs:

• API home: http://127.0.0.1:8000/
• Swagger docs: http://127.0.0.1:8000/docs

Frontend

Main UI file:

• analytiq.html

Open it in a browser (or serve it via a static server). It calls backend endpoints at http://127.0.0.1:8000.

API Endpoints

Ingestion

• POST /ingest/
  Legacy synchronous endpoint (returns final payload after full run)

• POST /ingest/start
  Starts background ingestion job and returns job_id

• GET /ingest/status/{job_id}
  Poll incremental progress and partial results

Reporting

• GET /reporting/{dataset_name}/html
  Opens generated report in browser

• GET /reporting/{dataset_name}/pdf
  Returns generated PDF if available, otherwise fallback behavior

• GET /reporting/{dataset_name}/pdf/download
  Strict PDF attachment download (requires actual PDF to exist)

• GET /reporting/
  Lists available reports

Output Directories

Configured in src/core/config.py:

• data/bronze for ingestion artifacts
• data/pro for processed/merged/prediction outputs
• data/eda for EDA artifacts
• data/reporting for report artifacts

Notes

• Large CSV handling is enabled (sampling/streaming thresholds in config.py).
• CORS is enabled for localhost origins.
• If backend PDF generation dependencies are missing, use report open/print flow as fallback.

Troubleshooting

1. Backend not reachable

• Confirm uvicorn is running on 127.0.0.1:8000.

2. Upload works but AI-powered choices are weak

• Check Azure/OpenAI env vars.

3. PDF route returns missing/not found

• Verify reporting step completed and PDF dependencies are installed.

License

Add your preferred license here.