-
Notifications
You must be signed in to change notification settings - Fork 924
Pull requests: jundot/omlx
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
fix: fall back to openai-harmony rendering for gpt-oss models without chat_template
#852
opened Apr 18, 2026 by
FaisalFehad
Contributor
Loading…
5 tasks done
fix(discovery): fall back to directory name for audio model type detection
#849
opened Apr 18, 2026 by
ryancee
Loading…
fix(specprefill): add qwen3_5_moe / qwen3_6 query extractor
#846
opened Apr 18, 2026 by
mrtkrcm
Loading…
4 tasks done
fix(server): Anthropic stream emission corner cases around tool_use
#845
opened Apr 18, 2026 by
mrtkrcm
Loading…
4 tasks done
fix(tool_calling): recover naked <function=...> calls emitted by Qwen3-Coder
#844
opened Apr 18, 2026 by
mrtkrcm
Loading…
4 tasks done
feat(chat): add file attachment support for PDF, Markdown, JSON, and plain text
#842
opened Apr 18, 2026 by
fxops-ai
Loading…
Adds 5 new intelligence benchmarks with bundled JSONL data and admin …
#837
opened Apr 17, 2026 by
SheeJiaWei
Loading…
fix: clean errors for non-LLM engines and missing STT processors (#507, #800)
#826
opened Apr 17, 2026 by
ethannortharc
Contributor
Loading…
5 tasks done
2
fix(admin): handle symlinked models in delete_hf_model endpoint (v2)
#824
opened Apr 17, 2026 by
Bahtya
Loading…
fix: honor response_format in /v1/audio/speech (#753)
#823
opened Apr 17, 2026 by
ethannortharc
Contributor
Loading…
4 tasks done
fix: SpecPrefill RoPE AttributeError on retry (#766) + Qwen3 MoE VLM misdetection (#812)
#822
opened Apr 17, 2026 by
Chuhan1112
Loading…
3 tasks
Adds JANG quantized model support to oMLX's batched engine.
#820
opened Apr 17, 2026 by
yohann-bearzi
Contributor
Loading…
fix(dflash): route image requests to VLM fallback; fix(tts): honour response_format
#818
opened Apr 17, 2026 by
Chuhan1112
Loading…
7 tasks
feat: preserve thinking across turns for Qwen 3.6+ incl. external clients
#814
opened Apr 16, 2026 by
latent-variable
Contributor
Loading…
feat: DDTree tree-based speculative decoding (rides DFlash)
#805
opened Apr 16, 2026 by
sooth
Loading…
8 tasks
refactor(integration): use [profiles.omlx] instead of top-level model override for codex integration
#776
opened Apr 15, 2026 by
yguilai
Loading…
feat: virtual model profiles via .virtual.yaml files
#773
opened Apr 15, 2026 by
TipKnuckle
Contributor
Loading…
feat(download): add concurrent model downloads and per-download worker settings
#758
opened Apr 14, 2026 by
uncle9x9
Loading…
5 tasks done
feat: add single_model_mode to force unload before load
#730
opened Apr 12, 2026 by
jroth1111
Loading…
[Performance] add hot cache only mode and optimize memory usage
#701
opened Apr 10, 2026 by
RepublicOfKorokke
Loading…
4 of 14 tasks
feat(admin): add hotswappable engine package management
#679
opened Apr 9, 2026 by
0xClandestine
Loading…
Previous Next
ProTip!
Updated in the last three days: updated:>2026-04-15.