autoglm-image-recognition

To generate an image description from an image URL, the Skill calls the AutoGLM Image Recognition API to convert picture contents into readable text. You provide a public http/https URL and the agent processes it to output a scene description.

Yes, you can extract visible text from an image using auto glm by providing a custom prompt like “识别图中文字” instead of the default “描述图片” prompt. This guides the recognition API to perform OCR-like text extraction tasks.

Yes, you need a local token service running at http://127.0.0.1:18432/get_token to run image recognition. The Skill uses this local endpoint to generate signed request headers required for POSTing to the image-recognition API endpoint.

To process a local image file for image understanding tasks, the file must first be uploaded to obtain a public oss_url. Once you have the public URL, the Skill passes it to the AutoGLM API to generate the description.

A key limitation of using a public URL for image description is that the URL must be publicly accessible. If the image is behind authentication or firewalls, the AutoGLM API cannot fetch it, and local files require manual uploading to generate a public oss_url first.