Generate high-quality images using FLUX models. Two models available: FLUX.1-dev (faster, 40 steps default) and FLUX.2-dev (highest quality, 50 steps default). Both models are optimized for quality output. Images are saved to the configured output directory with embedded metadata. IMPORTANT: This tool is LONG-RUNNING and NON-BLOCKING. flux1-dev takes ~4-8 min, flux2-dev takes ~30-40 min. The MCP client may receive a timeout error - this is EXPECTED and NORMAL. Do NOT treat a timeout as a failure. Inform the user that generation is running in the background and they should upload the result image when ready.

Immediately unload the FLUX model from GPU memory. Use this to free up VRAM when you're done generating images. The model will be automatically reloaded on the next generation request.

Get current status of the FLUX generator. Shows whether the model is loaded, time until auto-unload, and current VRAM usage.

Set the auto-unload timeout for the FLUX model. The model will automatically unload after this many seconds of inactivity. Set to 0 to disable auto-unload.

Retrieve a preview (thumbnail) of a generated image. Use this after generate_image completes — especially useful for FLUX.2-dev background generations where the MCP client timed out. Pass the image_id from generate_image output, or omit to get the last generated image. Returns the thumbnail inline and the full-size image path.