Commit Graph

6 Commits

Author SHA1 Message Date
Jeffrey Morgan
c01608b6a1 x/imagegen: add image edit capabilities (#13846) 2026-01-22 20:35:08 -08:00
Jeffrey Morgan
634c416645 Add experimental image generation fields to /api/generate (#13753)
Request fields (experimental):
- width: image width (max 4096)
- height: image height (max 4096)
- steps: denoising steps
- seed: random seed

Response fields (experimental):
- images: base64-encoded generated images
- completed: current step progress
- total: total steps

Other changes:
- Fix lifecycle bug where image models wouldn't unload (refCount issue)
- Fix "headers already written" error on Ctrl+C during streaming
- Add gin middleware for OpenAI /v1/images/generations compatibility
- Update CLI to use /api/generate with progress bar
- Add preload support in interactive mode
2026-01-17 18:27:41 -08:00
Devon Rifkin
e51dead636 preserve tool definition and call JSON ordering (#13525)
* preserve tool definition and call JSON ordering

This is another iteration of
<https://github.com/ollama/ollama/pull/12518>, but this time we've
simplified things by relaxing the competing requirements of being
compatible AND order-preserving with templates (vs. renderers). We
maintain backwards compatibility at the cost of not guaranteeing order
for templates. We plan on moving more and more models to renderers,
which have been updated to use these new data types, and additionally
we could add an opt-in way of templates getting an order-preserved list
(e.g., via sibling template vars)

* orderedmap_test: remove testify
2026-01-05 18:03:36 -08:00
Jeffrey Morgan
565b802a6b openai: fix tool call ID mapping (#12988) 2025-11-06 15:26:25 -08:00
Jeffrey Morgan
30fcc71983 api: add omitempty to required tool function parameter type (#12989) 2025-11-06 14:08:55 -08:00
Devon Rifkin
2c2f4deaa9 openai: refactor to split compat layer and middleware
This makes the core openai compat layer independent of the middleware
that adapts it to our particular gin routes
2025-10-05 14:18:56 -07:00