framepipe-dev/media-inference-worker

media‑inference‑worker

What it is – A minimal test client for a media‑processing pipeline that calls an image‑generation model (e.g., qwen‑image‑3). The repository contains a small Python script (generate.py) and a RUNBOOK.md that documents the available HTTP endpoints and raw request formats.

How it works – After installing the Python dependencies listed in requirements.txt, you run the script with two arguments: the model name and a text prompt. The script sends a request to the (presumably internal) inference service using the service‑account credentials that were copied with the code. The service returns a generated image, which the script saves or prints.

Typical usage

pip install -r requirements.txt          # install deps
python generate.py qwen-image-3 "Editorial portrait, hard flash, 35mm grain"

For other operations (e.g., different endpoints, batch requests) see the RUNBOOK.md file that lists the raw request payloads.

Scope & limitations – This repo is a test client; it does not contain the inference server itself, nor does it manage authentication or credit‑usage monitoring. The service account key may be expired or rotated at any time, so the client is primarily useful for developers who already have access to the corresponding media pipeline.

Who might find it useful – Engineers working on the broader media pipeline who need a quick way to verify that the image‑generation endpoint is reachable and behaving as expected.