Skip to main content

Can I send images or files to the LLM Gateway, or use it to generate images or embeddings?

You can send images and documents to models that accept them, but the Gateway only returns text, so it can't generate images, audio, video or embeddings.

You can send images and documents to a model that lists them in capabilities.input_modalities at https://llm.wiro.ai/v1/models/{owner}/{model}. Replies are text only.

  • How to send: a public HTTPS link, base64 or data content (up to 12 MiB per item after decoding; larger fails with payload_too_large), or a file-... ID from an upload to /v1/files.

  • Types: images must be GIF, JPEG, PNG or WebP; documents must be PDF or plain text.

  • Sizes: a request body can be up to 16 MiB, and a /v1/files upload up to 50 MiB, one file per upload and no ZIP archives.

Not offered: image, audio or video generation (use the Run API or MCP instead), embeddings, reranking, the legacy /v1/completions endpoint, the native Gemini API, provider-hosted web search or code interpreter, bring-your-own-key, and automatic fallback to another model.

Details: Files and what the Gateway doesn't do in our docs, and MCP vs. the LLM Gateway.

Did this answer your question?

Related Articles