Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit history

Commits on Apr 7, 2026

• bump version to 0.3.5.dev1

  jundot

  committed
  0103c2bView commit details

  Copy full SHA for 0103c2b

  View code at this pointBrowse repository at this point

• feat: add seed parameter for reproducible generation (#640)

  Show description for d745c31
  jundot

  committed
  d745c31View commit details

  Copy full SHA for d745c31

  View code at this pointBrowse repository at this point

• fix: skip quantizing modules without to_quantized() in oQ streaming (#625)

  Show description for 9905b8b
  jundot

  committed
  9905b8bView commit details

  Copy full SHA for 9905b8b

  View code at this pointBrowse repository at this point

• fix: use activeTheme for navbar theme picker styling in auto mode

  jundot

  committed
  dfd9c00View commit details

  Copy full SHA for dfd9c00

  View code at this pointBrowse repository at this point

• fix: add missing TTS/STT/STS deps to audio extra (#590)

  Show description for f41c8d7
  jundot

  committed
  f41c8d7View commit details

  Copy full SHA for f41c8d7

  View code at this pointBrowse repository at this point

• Merge pull request #621 from Stv-X/auto-color-scheme

  Show description for 4e63ae2
  jundot

  authored
  4e63ae2View commit details

  Copy full SHA for 4e63ae2

  View code at this pointBrowse repository at this point

• fix: unwrap tuple layer outputs in oQ sensitivity measurement (#627)

  Show description for 6973ced
  jundot

  committed
  6973cedView commit details

  Copy full SHA for 6973ced

  View code at this pointBrowse repository at this point

• bump mlx-lm/mlx-vlm and enable Gemma 4 native tool calling

  Show description for 82654b9
  jundot

  committed
  82654b9View commit details

  Copy full SHA for 82654b9

  View code at this pointBrowse repository at this point

Commits on Apr 6, 2026

• fix: adjust navbar theme picker width to prevent clipping

  Stv-X

  committed
  6b50ad8View commit details

  Copy full SHA for 6b50ad8

  View code at this pointBrowse repository at this point

• feat: add auto theme mode and system appearance sync

  Stv-X

  committed
  6a4e404View commit details

  Copy full SHA for 6a4e404

  View code at this pointBrowse repository at this point

Commits on Apr 5, 2026

• fix: TurboQuant SSD cache reconstruction crash (#577)

  Show description for 898b98f
  jundot

  committed
  898b98fView commit details

  Copy full SHA for 898b98f

  View code at this pointBrowse repository at this point

• fix: dark mode heading visibility and mobile UX in chat page (#586)

  jundot

  committed
  e8151f9View commit details

  Copy full SHA for e8151f9

  View code at this pointBrowse repository at this point

• feat: skip admin auth when skip_api_key_verification is enabled on localhost (#587)

  jundot

  committed
  d3dd8c5View commit details

  Copy full SHA for d3dd8c5

  View code at this pointBrowse repository at this point

• feat: add 6-bit and 8-bit turboquant options (#594)

  jundot

  committed
  7acdd58View commit details

  Copy full SHA for 7acdd58

  View code at this pointBrowse repository at this point

• refactor: unify max_num_seqs and completion_batch_size into max_concurrent_requests

  Show description for d0d317c
  jundot

  committed
  d0d317cView commit details

  Copy full SHA for d0d317c

  View code at this pointBrowse repository at this point

• bump 0.3.4

  jundot

  committed
  8924e61View commit details

  Copy full SHA for 8924e61

  View code at this pointBrowse repository at this point

• fix: VLM decode model memory duplication (#582)

  Show description for 47df15a
  jundot

  committed
  47df15aView commit details

  Copy full SHA for 47df15a

  View code at this pointBrowse repository at this point

• fix: apply gemma4 message extractor in anthropic handler

  Show description for d8a51bb
  jundot

  committed
  d8a51bbView commit details

  Copy full SHA for d8a51bb

  View code at this pointBrowse repository at this point

• fix: pad undersized RotatingKVCache on SSD restore to prevent merge crash

  Show description for f3f9695
  jundot

  committed
  f3f9695View commit details

  Copy full SHA for f3f9695

  View code at this pointBrowse repository at this point

Commits on Apr 4, 2026

• bump 0.3.3

  jundot

  committed
  9181b4aView commit details

  Copy full SHA for 9181b4a

  View code at this pointBrowse repository at this point

• fix: vlm sanitize proxy missing audio_tower attribute

  jundot

  committed
  a24ea52View commit details

  Copy full SHA for a24ea52

  View code at this pointBrowse repository at this point

• fix: preserve image_url parts in gemma4 message extractor

  Show description for 9ee9a03
  jundot

  committed
  9ee9a03View commit details

  Copy full SHA for 9ee9a03

  View code at this pointBrowse repository at this point

• fix: inherit prefill_attention from TurboQuantKVCache instead of forcing dequantize fallback

  jundot

  committed
  e9846ccView commit details

  Copy full SHA for e9846cc

  View code at this pointBrowse repository at this point

• fix(chat): add horizontal margin to thinking container

  jundot

  committed
  9216edaView commit details

  Copy full SHA for 9216eda

  View code at this pointBrowse repository at this point

• fix(chat): hide _ui:false messages and remove stray </think> on abort

  Show description for b5f06a4
  jundot

  committed
  b5f06a4View commit details

  Copy full SHA for b5f06a4

  View code at this pointBrowse repository at this point

• Merge pull request #509 from rayone/feat/chat-mcp-tools

  Show description for 9b6ce99
  jundot

  authored
  9b6ce99View commit details

  Copy full SHA for 9b6ce99

  View code at this pointBrowse repository at this point

• fix: handle null num_experts in oQ for Gemma 4 dense models (#554)

  Show description for f33275e
  jundot

  committed
  f33275eView commit details

  Copy full SHA for f33275e

  View code at this pointBrowse repository at this point

• fix: handle bfloat16 audio in TTS wav conversion (#551)

  Show description for 1464be6
  jundot

  committed
  1464be6View commit details

  Copy full SHA for 1464be6

  View code at this pointBrowse repository at this point

• fix: VLM batched decode degeneration via mlx-lm decode model

  Show description for e301d66
  jundot

  committed
  e301d66View commit details

  Copy full SHA for e301d66

  View code at this pointBrowse repository at this point

• Merge pull request #565 from TipKnuckle/feat/gemma4-reasoning-parser-clean

  Show description for 6d84ed6
  jundot

  authored
  6d84ed6View commit details

  Copy full SHA for 6d84ed6

  View code at this pointBrowse repository at this point

• oq: improve bit allocation policy for better quantization quality

  jundot

  committed
  7d790deView commit details

  Copy full SHA for 7d790de

  View code at this pointBrowse repository at this point

• refactor: rewrite BatchTurboQuantKVCache as TurboQuantKVCache subclass

  Show description for 89624d9
  jundot

  committed
  89624d9View commit details

  Copy full SHA for 89624d9

  View code at this pointBrowse repository at this point

• add vision feature cache for multi-turn image reuse

  Show description for 2d1a6cd
  jundot

  committed
  2d1a6cdView commit details

  Copy full SHA for 2d1a6cd

  View code at this pointBrowse repository at this point

• fix grammar constrained generation for new BatchGenerator pipelining

  Show description for 638bebd
  jundot

  committed
  638bebdView commit details

  Copy full SHA for 638bebd

  View code at this pointBrowse repository at this point

• upgrade mlx-lm to 4469ad4 (BatchGenerator refactor + gemma4)

  Show description for 2dc7cc8
  jundot

  committed
  2dc7cc8View commit details

  Copy full SHA for 2dc7cc8

  View code at this pointBrowse repository at this point