希望增加可识别文本的视觉大模型 API 接口

Hello, author! I tried to implement a general image-to-text function as an alternative to OCR functionality using some visual large model routes, and found that their recognition accuracy is quite good. I hope to add some interfaces for implementing general image-to-text functions that can be configured through BaseURL, API, etc. I made some initial attempts, and created a simple local general text-to-text system (of course, a relatively good graphics card configuration is required to run it) using Python flask + ollama + DeepSeek-ocr. I passed some simple OCR functionality tests by listening to the local 6666 port. This is my test image.