tool.gpt

Name: vision
Description: Analyze a set of images using a given prompt and vision model.
Credential: github.com/gptscript-ai/credentials/model-provider
Args: detail: (optional) Fidelity to process images at. One of ["high", "low", "auto"]. Default is "auto".
Args: max-tokens: (optional) The maximum number of tokens. Default is 2048.
Args: model: (optional) The name of the model to use. Default is "gpt-4o".
Args: prompt: (required) The text prompt based on which the GPT model will generate a response.
Args: images: (required) Comma-separated list of image URIs to analyze. Valid URIs start with "http://" and "https://" for remote images, and "file://" for paths to local image files. Only supports jpeg and png. Commas in file paths should be escaped with a backslash.

#!/usr/bin/env node ${GPTSCRIPT_TOOL_DIR}/index.js