New
qsp image "PROMPT" — text-to-image via the OpenAI-compatible /v1/images/generations, saved to disk. Default model flux.2-pro; -o, --size, -n, --json. -n generates N images (each a separate request/charge, written as soon as it succeeds); files always use the real returned format.
Changed
qsp models now splits chat vs image-generation tables and ships a refreshed pre-signin catalog + prices.- Default chat model is now
deepseek-v4-flash (cheaper, faster, 1M context). New QSP_IMAGE_MODEL env var.
- Cleaner failures across the board (
json.JSONDecodeError, image save errors).
pip install -U quicksilverpro
qsp image "a fox in the snow" -o fox.jpg
