Spaces:
Sleeping
Sleeping
Commit History
-
Render mode/upsampler as radio choices on the canvas
Add the choices array (per NodeWidget.svelte) to the mode and upsampler input ports so they show as radio selections instead of free text, matching ideogram-ai/ideogram4 /generate's Literal types. Co-Authored-By: Claude <noreply@anthropic.com>6b2a1f3 -
Switch to Space node calling ideogram-ai/ideogram4 /generate
A kind:'space' node authenticates with each visitor's HF OAuth token (gradio@6.19.0 call_space passes the resolved token to gradio_client), so no server-side key is needed and the fal bound-function workaround is unnecessary. The /generate endpoint already exposes all controls (prompt, mode, upsampler, width, height, seed, randomize_seed) and returns (image, seed, caption), mapped to input/output ports by order. Co-Authored-By: Claude <noreply@anthropic.com>891e4b6 -
Remove committed bytecode, add .gitignore
Co-Authored-By: Claude <noreply@anthropic.com>2f7efaf -
Expose fal's Ideogram 4 controls via a bound function node
gradio's built-in model node only forwards the prompt to text_to_image (no extra_body), so fal-specific fields would be silently dropped. Bind ideogram_v4() instead, which calls InferenceClient(provider='fal-ai').text_to_image(prompt, extra_body=...) with rendering_speed, image_size, expansion_model, acceleration, num_images, seed, output_format and enable_safety_checker as typed canvas ports. Wire Prompt -> generator -> Output Gallery. Co-Authored-By: Claude <noreply@anthropic.com>ed42479 -
Link ideogram-ai/ideogram-4-fp8 model in Space metadata
Co-Authored-By: Claude <noreply@anthropic.com>50f53ca -
Add Gradio Workflow app wired to ideogram-ai/ideogram-4-fp8
Co-Authored-By: Claude <noreply@anthropic.com>21ae5bf -
initial commit
Verified signature1013165