lightflow-flux-text-to-image

name	LightFlow FLUX Text To Image
description	Use this skill when working with the lightflow.flux_text_to_image workflow, syncing FLUX models, or generating images from prompts.
version	0.1.0

LightFlow FLUX Text To Image

Use lightflow.flux_text_to_image to generate images from text prompts through the LightFlow FLUX runtime.

Workflow

• Workflow id: lightflow.flux_text_to_image
• Inputs: prompt, negative, width, height, seed, count, steps, guidance, output_path, output_template, model.
• Outputs: image, image_path, images, image_paths.
• Runtime capability: lightflow.image.generate.
• Model requirements: flux_model, llm_model, and vae_model.
• Node Schema: image outputs are image artifacts; model is bound to flux_model; width, height, count, steps, and guidance include editor ranges.

Runtime

Run lfw sync lightflow.flux_text_to_image --auto-model --apply before a real generation run. The preferred runtime is LightFlow built with --features flux-native. Native text-to-image keeps a loaded FLUX/Qwen/VAE session in the LightFlow process, reuses it for later images with the same locked model paths, and sends count > 1 requests as one native batch call.

For ComfyUI-style residency across requests, use a long-lived LightFlow process such as lfw serve; one-shot lfw run commands release the native session when the process exits.

This project also includes a stable-diffusion.cpp adapter for builds without flux-native. Configure it only when using the external fallback:

export LIGHTFLOW_SD_CLI=/path/to/stable-diffusion.cpp/build/bin/sd-cli
export LIGHTFLOW_FLUX_RUNNER="$PWD/scripts/sd-cli-flux-runner"

The adapter validates sd-cli, model files, and source assets before loading models. The external fallback is process-per-call; use native execution for performance work.

Usage

lfw sync lightflow.flux_text_to_image --auto-model --apply
lfw run lightflow.flux_text_to_image \
  -i prompt='"a quiet lake at sunrise"' \
  -i width=768 \
  -i height=768 \
  -i output_path='"out/lake.png"'

Fast real-backend smoke test:

cargo run --manifest-path ../LightFlow/Cargo.toml --features flux-native --bin lfw -- \
  run lightflow.flux_text_to_image \
  -i prompt='"real FLUX smoke test, small red cube"' \
  -i width=128 \
  -i height=96 \
  -i seed=11 \
  -i steps=1 \
  -i guidance=1.0 \
  -i output_path='"/tmp/lightflow-flux-test/runner-real.png"'

Validation

lfw node test lightflow.flux_text_to_image

API Usage

Start lfw serve, then call the workflow through the shared HTTP run contract. Adjust inputs to match the workflow contract above.

curl -sS -X POST http://127.0.0.1:5174/workflows/lightflow.flux_text_to_image/run \
  -H 'content-type: application/json' \
  -d '{"inputs":{}}'