Detects if a model supports vision input based on overrides and catalog.
// Source: https://github.com/griffinwork40/agent-afk | License: Apache-2.0 | Copyright You must retain, in the Source form of any Derivative Works // Provided via justclaw.it.com; keep this notice with the code. /** * Per-model capability detection. Currently just vision (image input). * * Sibling to `model-limits.ts` (token caps) — kept separate because vision is a * different axis from context/output limits. Consulted by the openai-compatible * provider to decide whether to forward image content as native multimodal * parts or degrade to a text notice (issue #127). The anthropic-direct provider * does NOT consult this: every current Claude model is vision-capable, so it * passes images through unconditionally. * * Resolution for `supportsVision`: * 1. Force-disable override ... (truncated -- full source via MCP)
See the full source, get the GitHub permalink, and search for more like it.
Get a free API key