Cnapy AI
I-powered data analysis platform — upload a CSV/Excel file, ask questions in plain English, and get instant charts, insights, and business recommendations. No SQL or Python required.
About This Project
Cnapy is a full-stack SaaS web app that removes the barrier between non-technical users and their data. Users upload a spreadsheet and chat with an AI agent that reasons about the dataset, runs sandboxed Pandas/NumPy analysis, generates Plotly charts, and returns plain-English explanations with business recommendations — all in one conversational turn.
The AI agent is built on the Claude API using a ReAct (Reason → Act → Observe) loop with six tools (dataset profiling, sandboxed code execution, chart generation, data cleaning, recommendation generation, and report export). Analysis code runs inside a RestrictedPython sandbox with no file I/O, imports, or shell access.
The app ships with full account infrastructure: email/password and Google OAuth login (django-allauth), a guided onboarding flow with a pre-loaded demo dataset, Stripe-powered subscriptions ($12/mo or $99/yr) with webhook-driven tier enforcement, and PDF/CSV/Excel report export. Async work (agent calls, exports) runs on Celery + Redis, with PostgreSQL for persistence and AWS S3 for file storage. The frontend uses HTMX and Alpine.js over Django templates with Tailwind CSS, so query results stream in without a SPA framework.
I built this solo end-to-end: architecture, backend, agent design, sandboxing/security hardening, CI (GitHub Actions), and deployment tooling (Gunicorn + Caddy).
