DocuVerse
AI document workspace: CPU-first OCR turns PDFs and scans into searchable Markdown and extracts structured fields with per-field confidence.
Overview
DocuVerse ingests PDFs and images and runs CPU-first OCR (PaddleOCR PP-OCRv6; DeepSeek-OCR is an opt-in GPU profile) to deliver Markdown-first documents plus rule-based structured extraction — dates, totals, document numbers — inside a workspace model with role-based access. The FastAPI/PostgreSQL/Redis/Celery backend streams OCR job progress over WebSockets to the Vue web client.
The challenge
Self-hosted document intelligence has to run on modest hardware and still produce trustworthy output: scans without a text layer, mixed Vietnamese, Japanese and English pages, and structured fields — not just text — when downstream systems need to act on documents.
Our solution
A Clean Architecture FastAPI backend with Celery OCR jobs: a CPU-first PaddleOCR engine behind a swappable port, a PDF text-layer fast path, and a rule engine that matches templates to extract typed fields with per-field confidence — all streaming progress over WebSockets.
Key features
Results & impact
Related Projects
Explore more projects similar to this one
VietProStocks
VietProStocks combines Vietnamese market data, technical analysis and a retrieval-augmented LLM into a research assistant. The RAG stack — MiniLM embeddings, a FAISS index and a local llama.cpp model — is designed to run entirely on a Raspberry Pi 4 8GB, with an OpenAI-compatible backend as an escape hatch when more compute is available.
SentinelVision Edge
SentinelVision Edge turns low-cost Raspberry Pi cameras into an AI security fleet. Node agents run YOLO person and face detection at the edge, publish occupancy and dwell-time signals over MQTT, and stream video through MediaMTX. A FastAPI backend with PostgreSQL, Redis and Celery manages cameras, areas, recordings and roles, with a Vue 3 dashboard for real-time status.
Smart Virtual Meeting
Smart Virtual Meeting is a full conferencing stack: a NestJS + Prisma + Redis control plane, mediasoup 3.15 for selective forwarding, Socket.IO signaling and a React 19 client. Kubernetes, Helm and Terraform deployment assets cover the operational side, and a detailed scaling plan tackles the hard parts — TURN sizing, worker placement and explicit benchmark caveats.
Start Your Project
Tell us about your project and we'll get back to you within 24 hours.
Reference Project:
DocuVerse