comfy_wrap git:20261014.71d43e0A

Immutable. This exact content never changes and is served at /api/v1/blob/902c65aae858ccef.

---
name: comfy_wrap
description: >
  Comfy Workflow Wrapper — generate images by injecting a text prompt into
  a ComfyUI workflow JSON and running it via vibecomfy. Single-action pack
  for parameterized ComfyUI workflow execution.
---

# Comfy Workflow Wrapper

The comfy_wrap pack provides one action that wraps a ComfyUI workflow JSON:
inject a text prompt, execute via vibecomfy, and collect the output image.

## Action

| Executor | What it does |
|---|---|
| `comfy_wrap.run` | Load a ComfyUI workflow JSON from `/tmp/example_comfy.json`, inject a user prompt into the positive CLIPTextEncode node, execute through vibecomfy, and copy the first result to the declared output. |

See the [action guide](../actions/run/STAGE.md) for its input, output, workflow
node, dependency, and recovery details.

## When to use

- Use `comfy_wrap.run` when you have a pre-built ComfyUI workflow JSON and want
  to parameterize it with a text prompt.
- Use `comfy_wrap.run` when you need a quick txt2img path through ComfyUI without
  writing a full pipeline.

## When NOT to use

- Do not use for arbitrary model selection or multi-modal generation — use
  `generation.generate_image` for the standard happy path.
- Do not use for LoRAs, IP-adapter, ControlNet, or custom samplers — use
  `vibecomfy.run` directly (the escape hatch).

## Quick-start

```python
from astrid.sdk import AstridClient
import astrid.sdk as sdk

with AstridClient.open_from_launcher() as client:
    result = sdk.invoke(
        "comfy_wrap.run",
        kind="action",
        project="demo",
        client=client,
        inputs={"prompt": "a serene mountain lake at dawn"},
        wait=True,
    )
```

The action's output is declared by the pack and managed by the host. This
documentation records the supported route; live ComfyUI inference is not
required for pack layout validation and was not exercised during migration.

## Requirements

- `vibecomfy` package installed
- ComfyUI workflow JSON at `/tmp/example_comfy.json`
- GPU recommended for ComfyUI model inference