iFlytek BigModel OCR Image Understanding

Problem It Solves

When you already have a local image, screenshot, or table capture and need machine-readable content, use this skill to pass the image path directly instead of assembling a one-off OCR call. It fits office workflows where the result will be written into documents, spreadsheets, ticket fields, or downstream automation.

How It Works and Where It Binds

The skill first checks that the input image path exists, then resolves credentials and model settings. Configuration precedence is: CLI flags such as --api-key, --model-id, and --base-url first, then environment variables (default IFLYTEK_API_KEY), then the JSON file passed via --config. assets/config.template.json is only a template and should not be treated as real credentials. After execution, it returns a JSON result, and the content field can be explained when needed.

Use it for local images containing text, tables, or screenshot content. If a custom prompt is needed, pass it through arguments rather than editing files inside the skill package.