Generate File

This tool fetches a file the user has attached to the conversation by its id, returning the parsed markdown content when available (or a download link otherwise), so the agent can read uploaded PDFs and other files on demand. Injected automatically when a conversation has attached files.

This tool finds files in your library by name, your own description, or the text parsed out of them — so a PDF is findable by a phrase on page four even when its filename says nothing. Results are metadata (name, type, folder, parse status); the agent reads a file with Get File. You can narrow to images, PDFs, or everything else, or to a single folder.

This tool uses image generation models from Google or OpenAI to create images from text descriptions. Supports long descriptions. Choose a size (s, m, l, xl) and an aspect ratio (square, landscape, or portrait); on OpenAI models, small and medium always render as 1024x1024 (so pick large or extra large for non-square output) and large square is also capped at 1024x1024. Generates one to four images per request.