Repository navigation
Miffan 4.6.0 - #58
Merged
Merged
Miffan 4.6.0#58
Conversation
"Check computer status" on a saved computer now opens a status page instead of the first-time setup flow: one line says how the computer is, the checks fix things in place, and the computer's page shows the last result without connecting. The page asks the computer's cua-driver whether a newer release exists and offers the update. Updates run outside the page (CuaUpdates): they wait for a partner's turn to finish, keep partners off the computer while the driver restarts, and notify when the user is not watching. Helpers (sh 23, ps1 12) gain `cua check-update`; `cua install` now runs in the helper on every system, uses the computer's own proxy, and refuses a network too slow to download GitHub releases before changing anything. Windows follows the official stable release instead of pinning 0.34.0, checks the bridge after installing and restores the previous release when the new one fails. It also fixes two install bugs never hit before: any installer stderr line aborted the install, and a failed download looked like success. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Windows helper (now 13) only checked that GitHub's release host answered, so a proxy or network that connects but crawls passed and left the installer downloading for many minutes. Like the POSIX helper, it now requires 1 MiB of a release within 20 s, through the system proxy the installer uses, before changing anything. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Onboarding: - A conversational first run replaces the old onboarding pages: connect a service (OpenRouter login or a pasted key), pick a model, pick the first partner, who then says hello on its own. - The user picks the chat model; nothing is recommended. After OpenRouter login, free models are listed first. - Every step can be taken back: switch service while pasting a key, undo a connection, and partners are selected before they are confirmed. - Key errors are explained per cause (wrong key, unreachable, service error, no chat model); balance errors in a chat link to the AI service page. IM cards: - Shared ImActionCard, ImOptionList and ImChoiceQuiz (ui/im/ImCards.kt) now frame approval, question, bug report, computer and partner-message cards, and every onboarding card. - Ready answers under a new partner's hello (and quick messages on an empty chat) show as a multiple-choice quiz; once the owner picks one or types their own reply it folds into a read-only line. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…visible Abilities: - Asking the owner questions (ask_user) is on by default for new partners; partners made before get it once through a preferences migration. - The partner profile lists ask, calendar, clipboard, screen time and read aloud as switches; turning one on first gets the calendar permission or usage access it needs. - New request_ability tool: a partner asks in the chat to turn on an ability that is off. Allowing first gets the system permission, then the ability's tools join the same turn so the partner can carry on. - my_environment now lists the IM tools (profile, partners, groups, chat list) and the abilities that are off but can be requested, so a partner no longer concludes it has no tools. Questions: - Questions the owner skipped by replying in the chat used to vanish; they now fold into one line that opens to show what was asked, read-only. - Answers given on the card sit in the same white block as other cards. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Profile:
- New 「想多久再回」 row under the model row: four tiers instead of
reasoning effort levels (instant, depends, think more, think deeply),
each with a typical wait. The row is hidden when the model cannot think.
- Tiers map onto the assistant's reasoning level that professional mode
also edits: OFF / AUTO / HIGH / MAX (falling back to what the model
supports). Pro-mode LOW and MEDIUM read as "depends" and are only
overwritten when the owner picks a tier.
- On models that always think, the instant tier becomes "as fast as
possible" (LOW) and the sheet says why.
Chat:
- "Think hard this time" in the composer's + menu arms deep thinking for
the next message only, shown as a dismissible chip. The level travels
with the queued message and overrides the partner's setting for that
reply alone.
- While a long-thinking reply is pending the header counts the seconds
("thinking hard… 12s").
Strings in en, zh, ja, ko, ru and zh-TW.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The self-hosted runner is a desktop that is not always on and often lost contact mid-run, which held up release PRs. CI and both R2 jobs go back to the GitHub-hosted images they used before fcad730. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Ayuilos
enabled auto-merge
October 25, 2026 15:21
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
发布 Miffan 4.6.0（versionCode 178041）。更新日志见 docs/releases/4.6.0.md。
次版本：首次打开改成和 Miffan 聊天的引导（连服务、选模型、选第一个伙伴，可从备份恢复）；IM 卡片统一样式，新伙伴打招呼后附可选回答；伙伴可在聊天里申请打开能力，资料页新增能力开关，「提问」默认打开（一次性设置迁移）；「想多久再回」与「这次认真想想」；新的电脑状态页，可检查并更新操作小助手（cua-driver），更新前测网速，Windows 跟随官方稳定版并可回退。三份 README 已同步。新文案已补齐日、韩、俄、繁中翻译。无上游同步，无数据库结构变更。
🤖 Generated with Claude Code