Morph MORPH
SIGN IN SIGN UP
dhruvhead / omlx UNCLAIMED
0 0 0 Python

ACTIVITY

Commits Reviews
week month year
COMMITS
20
in the last week
CONTRIBUTORS
3
active
STARS
0
total
FORKS
0
total
TOP CONTRIBUTORS
J
jundot
15 commits
A
Alex Fishlock
3 commits
C
Craig Tollifson
2 commits
RECENT COMMITS
J
formula: bump to v0.3.5
jundot
3af2e1e
J
Merge pull request #789 from fix/vlm-tool-message-passthrough
jundot
071f2a6
J
merge: resolve conflict with main (dflash-mlx comment)
jundot
07adf4c
C
fix: update dflash-mlx pin to valid commit (8e1df22 was force-pushed away)
Craig Tollifson
ed7a46a
C
fix: preserve tool_calls/tool_responses in VLM message formatting
Craig Tollifson
92aab5c
J
fix(dflash): lazy fallback to prevent double model loading
jundot
58b3ca5
J
fix(admin): hide reasoning parser dropdown when xgrammar is not installed (#774)
jundot
3d1ac71
J
deps: bump dflash-mlx to v0.1.3 (814c4a1)
jundot
dd1093d
J
fix: deep-copy tokenizer to prevent "Already borrowed" under concurrent load
jundot
a045b6a
J
fix: scale settle_tolerance with model size for large models (#768)
jundot
b10eaad
J
fix: revert TurboQuant KV conversion in external prefill (#771)
jundot
6731a2c
J
formula: bump to v0.3.5-rc1
jundot
b125bfc
J
fix: remove false-positive RotatingKVCache stale offset warning, skip specprefill system_end calc when disabled
jundot
62c8f5f
J
bump version to 0.3.5
jundot
fd55ab3
A
fix: address PR review — safer restore, consistent model detection, full endpoint coverage
Alex Fishlock
95ea04a
A
fix: rename colliding params for Gemma 4, restore in both streaming and non-streaming paths
Alex Fishlock
fec8dc0
A
fix: enrich Gemma 4 tool parameter descriptions to prevent dropped fields
Alex Fishlock
6fa0a77
J
docs: add DFlash-MLX integration guide
jundot
8baa0af
J
fix(benchmark): skip batch test for DFlashEngine (#759)
jundot
6b1029c
J
fix: remove VoiceDesign hasattr routing from TTS engine
jundot
aca2de1