DeepSeek OCR Visor Agent: documents into structured data

Project

DeepSeek OCR Visor Agent

The situation

Turning invoices, contracts and scans into structured data by deploying an open-source OCR model directly means handling GPU compatibility, running out of VRAM and output formats yourself.

Solution

Published as the Python package deepseek-visor-agent and callable in a few lines of code: it detects the device automatically and steps down to smaller recognition modes when VRAM runs short; it outputs Markdown plus extracted fields, such as an invoice's amount and date, and splits multi-page PDFs automatically.

It plugs in as a tool for agent frameworks such as LangChain, LlamaIndex and Dify. The source is public under the Apache-2.0 license.

Sources

• GitHub project：https://github.com/YuanASI/deepseek-ocr-visor-agent ↗
• PyPI：https://pypi.org/project/deepseek-visor-agent/ ↗