Conversa AI — Enterprise Document Intelligence

Conversa AI is an enterprise-grade AI agent platform designed to solve two critical challenges in modern business: unstructured document processing and data accessibility. The first feature, OCR Docs, allows users to upload PDF or image files (JPEG, PNG, WEBP, HEIC, HEIF). Gemini Vision API extracts the full content — including text, tables, and key fields such as invoice numbers, dates, and contract values — and returns structured JSON. For scanned documents, Conversa AI automatically switches to a page-by-page OCR fallback strategy, rendering each page as a high-resolution PNG before sending it to Gemini for text extraction. Users can then chat with their documents using natural language questions. The second feature, Data Intelligence, enables users to upload CSV or Excel files. All sheets are parsed simultaneously using SheetJS, then passed to Gemini with full conversation history for multi-turn analysis. Users can ask questions like "What is the total revenue this quarter?" or "Are there any anomalies in this dataset?" and receive accurate, data-grounded answers in real time with word-by-word streaming responses. Both features include a 3x retry mechanism with exponential backoff (2s, 4s, 8s) to handle Gemini API rate limits gracefully. The application is fully deployed on Vercel with serverless API routes, a 60-second execution limit for heavy OCR tasks, and a single GEMINI_API_KEY environment variable for configuration. Live demo: https://conversa-ocr-docs.vercel.app GitHub: https://github.com/jefribulomakassar/conversa-ocr-docs