What is deepseekimg?

This project is a small add-on, called a skill, for the Claude Code coding assistant. It gives another AI model, DeepSeek, the ability to understand pictures, something DeepSeek cannot do on its own since it only reads text. The way it works is simple. When you show DeepSeek an image and ask a question about it, this skill quietly sends that image to a separate vision model running on your own computer through a tool called Ollama. That vision model looks at the picture, works out what is in it or reads any text inside it, and sends a written answer back. DeepSeek then uses that answer to reply to you, so from your side it feels like DeepSeek can see, even though it never actually looks at the image itself. To use it, you need Node.js version 18 or newer, plus Ollama installed and running on your machine along with a vision model such as gemma4:e2b. Setup involves installing Ollama, pulling that model, and then either copying this skill into a global folder so Claude Code always finds it, copying it into just one project instead, or calling the included script directly by hand without registering it as a skill at all. Once installed, you talk to DeepSeek normally and mention an image path along with your question, for example asking how many people are in a photo, asking it to read text out of a screenshot, or asking what a product in a picture is. There is also a plain describe mode that asks for a full description of an image with no specific question attached. A few settings can be changed through environment variables, including which vision model to use, the address of the Ollama server if it runs somewhere other than your own machine, and whether to turn on an extended thinking mode. This makes it possible to point the skill at a different model or a remote Ollama instance without changing any code.