07akioni
d743267388
Merge pull request #5946 from deepseek-harness/release-0.2.1-alpha.2
...
release(dsh): 0.2.1-alpha.2
2026-10-09 17:31:32 +08:00
Turtle
5876567dfc
Merge pull request #5915 from deepseek-harness/turtle/exp-badge-align-42f8ebeac170
...
fix(web): align and localize Auto review permission option
2026-10-09 16:50:08 +08:00
07akioni
65b8b7e853
Merge pull request #5557 from deepseek-harness/fix/prune-python-runtime-tests
...
Prune NumPy/pandas upstream tests from bundled Python runtimes
2026-10-09 16:48:52 +08:00
07akioni
78be1e5e46
Merge master into fix/prune-python-runtime-tests
2026-10-09 16:37:22 +08:00
Yudong Han
18643292ae
Merge pull request #5369 from deepseek-harness/fix/python-wheel-size-5250
...
fix: download optional Python runtime resources explicitly
2026-10-09 16:32:27 +08:00
07akioni
d911abce6b
Merge pull request #5523 from deepseek-harness/refactor/desktop-pnpm-dedupe
...
refactor(desktop): deduplicate bundled pnpm
2026-10-09 16:31:58 +08:00
Ruilin Geng
5f3c7b98e2
Merge pull request #4945 from deepseek-harness/fix/4931-composer-focus
...
fix(client): preserve cursor and selection on composer whitespace clicks
2026-10-09 16:22:17 +08:00
imccyu
311fc54705
Merge pull request #5955 from deepseek-harness/fix-chat-height-reset
...
fix(ui-chat): reset layout on mode changes and running session returns
2026-10-09 16:13:23 +08:00
kingwl
db5c4aaab9
Merge master and resolve integration conflicts for #5523
2026-10-09 16:12:44 +08:00
gengruilin
18678f4851
Merge origin/master into fix/4931-composer-focus
2026-10-09 16:10:14 +08:00
Yichen Jiang
ba41fbdf1d
Merge pull request #5949 from deepseek-harness/worktree/fix-model-catalog-visibility
...
fix(models): hide unconfigured models and unavailable selection IDs
2026-10-09 16:09:53 +08:00
Turtle
a1a43cbb4e
test(web): restore access mode after narrow Auto check; clarify EXP docs
2026-10-09 16:07:04 +08:00
yudshj
0616cbc77b
test: keep locale regression commands ASCII on Python 3.10
2026-10-09 16:06:27 +08:00
Turtle
e7c4ef0426
fix(web): raise Auto review icon A by 0.4px
2026-10-09 16:03:26 +08:00
Yichen Jiang
e03a21166d
test(models): update unavailable selection browser flow
2026-10-09 15:58:42 +08:00
yudshj
4fd85be130
fix: validate optional Python resources and isolate downloads
2026-10-09 15:52:59 +08:00
imccyu
7ee4b46782
fix(ui-chat): reset layout on mode changes and running session returns
2026-10-09 15:51:46 +08:00
Turtle
29b89df1b8
feat(web): add Auto review shield-A permission icon
2026-10-09 15:40:16 +08:00
Turtle
84f6a99514
Merge pull request #5942 from deepseek-harness/turtle/plan-mode-reject-33e1a9fe3ad2
...
fix(plan-mode): end the turn when the user dismisses a plan review
2026-10-09 15:28:21 +08:00
Ziya
f197d71940
Merge pull request #5921 from deepseek-harness/fix/voice-input-component-labels
...
fix: localize voice input component names and descriptions
2026-10-09 00:14:36 -07:00
07akioni
e2c410ef84
release(dsh): 0.2.1-alpha.2
2026-10-09 15:13:40 +08:00
Turtle
0155017b73
fix(plan-mode): exit plan mode only on approval in plan guidance
...
A dismissed review now returns a successful exit_plan_mode result, so
the plan policy must not treat any success as leaving plan mode. Also
update the user-facing description of a dismissed review.
2026-10-09 14:58:20 +08:00
Yichen Jiang
02f5423070
fix(models): hide unconfigured DeepSeek models and unavailable selections
2026-10-09 14:56:27 +08:00
Ziya
640e28cd94
feat(web): reorganize plugin settings and add experimental catalog page ( #5922 )
...
Group built-in settings before optional and installed plugins, and open the full experimental catalog on a separate page.
Preserve detail navigation and unsaved configuration when installation finishes, and use the visible More label as the accessible name.
2026-10-08 23:42:56 -07:00
ZiyaZhang
799fede7bc
fix: align speech recognition service terminology
2026-10-08 23:41:37 -07:00
Turtle
18a4bc4093
fix(plan-mode): end the turn when the user dismisses a plan review
...
Choosing "Request changes" on a plan review rejected the question, and
exit_plan_mode turned that into a failed tool result asking the model to
stop. The loop then spent one more model request only to have the model
wait. A dismissed review now returns { approved: false } and concludes the
turn, so the user's next message starts the following request.
2026-10-09 14:39:43 +08:00
ZiyaZhang
e3909dcc6e
Merge master into fix/voice-input-component-labels
2026-10-08 23:35:20 -07:00
Yichen Jiang
aad5ab5067
Merge pull request #5937 from deepseek-harness/worktree/fix-stale-build-workspaces
...
feat(build): add make clean shortcut
2026-10-09 14:30:15 +08:00
winewill
d16abc2745
Merge pull request #5898 from deepseek-harness/build/windows-ci
...
build(windows): add manual installer CI and harden Desktop runtime patches
2026-10-09 14:06:31 +08:00
imccyu
d6692ac626
Merge pull request #5305 from deepseek-harness/worktree-flownew
...
feat(web): 增加工作过程收起时机实验开关
2026-10-09 14:03:54 +08:00
Yichen Jiang
d3454a1aa3
chore(build): limit PR to the make clean shortcut
2026-10-09 14:01:12 +08:00
Ziya and ZiyaZhang
398f3f7ee3
fix(web): 弱化并左对齐思考翻译切换按钮 ( #5911 )
...
* fix(web): soften and align the reasoning translation toggle
* fix(web): scope translation toggle styles explicitly
---------
Co-authored-by: ZiyaZhang <[email protected]
2026-10-09 05:57:02 +00:00
Yichen Jiang
0b5671d956
fix(build): exclude stale package directories from workspace bundles
2026-10-09 13:46:34 +08:00
yudshj
305021fcd6
fix: preserve Linux SEA layout and stable snapshot references
2026-10-09 13:46:00 +08:00
imccyu
37d9dadd4e
fix(ui-chat): retain fold surplus as bottom space
2026-10-09 13:35:20 +08:00
imccyu
9db5308d9b
fix(ui-chat): combine fold displacement with input following
2026-10-09 13:35:19 +08:00
imccyu
f1dcb15b9c
fix(ui-chat): tolerate subpixel rounding at the content floor
2026-10-09 13:35:18 +08:00
imccyu
7a504ac3e0
fix(ui-chat): settle fold following and apply review cleanups
2026-10-09 13:35:18 +08:00
imccyu
bde4bc78a6
fix(ui-chat): avoid pathological flow selector matching
2026-10-09 13:35:17 +08:00
imccyu
00d9dd1b81
fix(ui-chat): preserve flow layout through slot anchors
2026-10-09 13:35:16 +08:00
imccyu
a0a9aef5ec
feat(ui-chat): add independent collapse timing and viewport-scoped flow
2026-10-09 13:35:16 +08:00
yudshj
1b8532b7c2
fix: keep Python executable packaging inside the deployed closure
2026-10-09 13:34:56 +08:00
Turtle
039fe73c55
Merge pull request #5917 from deepseek-harness/docs/error-handling-style
...
docs: define error handling responsibilities and failure guarantees
2026-10-09 13:28:43 +08:00
07akioni
544fe59fbd
Merge pull request #5909 from deepseek-harness/07akioni/electron-44-7-addon-019
...
chore(deps): upgrade Electron, electron-builder, and native loader
2026-10-09 13:26:18 +08:00
Yudong Han
ea4f2d2929
Merge pull request #5929 from deepseek-harness/fix/sensevoice-vad-onset-clipping
...
fix(speech-to-text-sensevoice): restore the onset a cold VAD drops
2026-10-09 13:17:10 +08:00
turtle1999
1875ec4ab8
docs: keep error handling guidance concise and clarify ownership
2026-10-09 13:09:59 +08:00
winewill
f10cfbad34
docs(desktop): clarify Windows runner routing and pending builds
2026-10-09 13:00:41 +08:00
Yudong
d6042ca0d7
fix(speech-to-text-sensevoice): restore the onset a cold VAD drops
...
Silero cannot back-date the first segment of a cold detector, so it
reports a start index several frames after speech actually began. Every
recording therefore lost its opening audio: measured across 50
utterances the median loss was 102 ms and the worst 230 ms — a full
syllable, which surfaced as dropped leading words ("麻烦把发票…" →
"把发票…", "用 Python…" → "写一个 Python…").
Only the first segment is affected; later segments aligned within 3 ms
because the detector retains history by then. Tuning cannot fix it:
lowering minSpeechSeconds from 0.25 s to 0.05 s left the median loss at
110 ms, and lowering the threshold to 0.2 still lost up to 174 ms while
risking false triggers.
Restore the dropped onset from the recording itself, bounded by the new
vadOnsetPaddingSeconds (default 0.3 s, zero disables). Measured on the
same corpus with identical DSH settings, this recovers the full no-VAD
accuracy and in the technical domain exceeds it, since segmentation also
trims silence:
语料 条件 无VAD VAD丢起音 VAD+回填 改善
audio clean 12.50% 16.33% 12.50% 23.5%
audio noisy10 17.74% 19.15% 17.94% 6.3%
audio_tech clean 10.34% 15.65% 9.81% 37.3%
audio_tech noisy10 25.73% 30.50% 26.26% 13.9%
Inference cost is unchanged (2.6 s vs 2.6 s on both clean corpora).
2026-10-09 12:51:54 +08:00
Turtle and Yifffan
e9292ba398
fix(web): show Auto review EXP only in menu options
...
Co-authored-by: Yifffan <[email protected]
2026-10-09 12:23:02 +08:00
yudshj
89f35fe2e9
fix: download optional Python runtime resources explicitly
2026-10-09 12:06:50 +08:00
07akioni