Извлечение текста и данных из документов (self-hosted)

★ 3.2K

text-extract-api is a self-hosted API that converts any image, PDF, or Office document into Markdown text or structured JSON with high accuracy, including tables, numbers, and math formulas. Reach for it when you need to turn documents into text/data on your own server without sending them out: OCR scans and PDFs, pull tabular data, structure contracts and invoices for downstream processing, and even strip personally identifiable information (PII). It is built on FastAPI with a Celery task queue and Redis cache; recognition uses several OCR strategies (EasyOCR, llama3.2-vision, minicpm-v, marker-pdf), and a local Ollama model cleans up the result and produces JSON — all locally, no cloud. It has switchable storage strategies (local, Google Drive). Its focus is a private "document → Markdown/JSON" pipeline over an API, not a finished app with a UI or a cloud OCR service: it is a backend component to embed in your own pipelines.