Repository navigation
ci: release packages - #4220
ci: release packages#4220
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e724b08888
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| { | ||
| "name": "@moonshot-ai/kimi-code-sdk", | ||
| "version": "0.20.0", | ||
| "version": "0.21.0", |
There was a problem hiding this comment.
The 0.21.0 changelog explicitly removes the public createSessionWithKaos / resumeSessionWithKaos methods and kaos / persistenceKaos options, so SDK consumers upgrading from 0.20.0 can stop compiling. Publishing that API removal as a minor release conflicts with the repository's required major-bump path for breaking changes; either retain deprecated compatibility declarations or obtain the required major release approval.
AGENTS.md reference: AGENTS.md:L62-L62
Useful? React with 👍 / 👎.
|
|
||
| - [#4091](https://github.com/MoonshotAI/kimi-code/pull/4091) [`21406fb`](https://github.com/MoonshotAI/kimi-code/commit/21406fb4c805cc8c715e6d1f16ad3fb5f25f4fe3) Thanks [@sailist](https://github.com/sailist)! - Omit max_tokens for models without an explicit output limit, fixing repeated 400 errors on strict serving stacks (e.g. bare vLLM) when the model's max output size is unknown. | ||
|
|
||
| - [#4184](https://github.com/MoonshotAI/kimi-code/pull/4184) [`cbd7540`](https://github.com/MoonshotAI/kimi-code/commit/cbd754049a114ee36d1b27675c82819e4963c692) Thanks [@7Sageer](https://github.com/7Sageer)! - Remove the never-implemented `executor = "kaos"` option from MCP stdio server configuration; configs using it now fail validation at load time. |
There was a problem hiding this comment.
For an existing configuration containing a disabled stdio MCP entry with executor = "kaos", the previous release could load the configuration because the value was only rejected when that client was constructed; 2.2.0 now rejects the entire configuration during schema validation. This entry names the loss but provides no config, environment variable, flag, or migration path to restore startup behavior, so preserve the legacy value as deprecated or add the required escape hatch/sign-off before publishing.
AGENTS.md reference: AGENTS.md:L106-L108
Useful? React with 👍 / 👎.
| - Updated dependencies [[`5b93669`](https://github.com/MoonshotAI/kimi-code/commit/5b936697670ed15444bcb4720bd91e6ee776a941), [`5b93669`](https://github.com/MoonshotAI/kimi-code/commit/5b936697670ed15444bcb4720bd91e6ee776a941)]: | ||
| - @moonshot-ai/agent-core-v2@0.5.0 | ||
| - @moonshot-ai/remote-control@0.0.1 |
There was a problem hiding this comment.
The release parent contains the new remote-control device list, rename, deactivate, and delete routes plus new status/conflict response fields from c69bd6c and c1ec372, but this changelog describes 0.2.4 only as dependency updates and the generated release description never names the desktop/web consumers or previous-client compatibility. Because the API-surface snapshot is an explicit external contract tripwire, add a release entry that declares those consumers and compatibility guarantees before consuming the release changesets.
AGENTS.md reference: AGENTS.md:L114-L116
Useful? React with 👍 / 👎.
| @@ -1,5 +1,63 @@ | |||
| # @moonshot-ai/kimi-code | |||
|
|
|||
| ## 2.2.0 | |||
There was a problem hiding this comment.
The 2.2.0 artifact also ships observable changes that are absent from this public CLI changelog: KIMI_CODE_LLM_HEADERS_TIMEOUT_MS from c151d5a, relay-triggered shutdown for kimi web --remote-control from 7f0ff53, and the changed fs:list glob semantics from 5b93669 (listed only under the private engine package). CLI users therefore receive new inputs and changed behavior without the release declaring them; add CLI changesets or equivalent 2.2.0 entries before publication.
AGENTS.md reference: AGENTS.md:L102-L104
Useful? React with 👍 / 👎.
| { | ||
| "name": "@moonshot-ai/kimi-code-sdk", | ||
| "version": "0.20.0", | ||
| "version": "0.21.0", |
There was a problem hiding this comment.
This release consumes the SDK changesets and advances the repository to 0.21.0, but the manifest still has private: true and .github/workflows/release.yml:131 publishes exclusively through changeset publish. The pinned Changesets 2.30.0 implementation filters publish candidates to packages where packageJson.private is false, so 0.21.0 will be skipped while the changesets are permanently consumed (Changesets publish implementation); remove private or otherwise make this explicitly public before merging the release commit.
AGENTS.md reference: AGENTS.md:L22-L22
Useful? React with 👍 / 👎.
|
|
||
| ### Minor Changes | ||
|
|
||
| - [#4159](https://github.com/MoonshotAI/kimi-code/pull/4159) [`0a3a2fb`](https://github.com/MoonshotAI/kimi-code/commit/0a3a2fbafab473df9e9f6609f825af78d2c1ce6d) Thanks [@RealKai42](https://github.com/RealKai42)! - Add an opt-in NotifyUser progress update feature (off by default): new sessions can show the agent's mid-turn progress with an optional one-line title, enabled in the TUI `/experiments` or the web/Desktop settings' Lab section; set `[experimental] notify_user = true` or `KIMI_CODE_EXPERIMENTAL_NOTIFY_USER=1` to turn it on. |
There was a problem hiding this comment.
For kimi web users who enable notify_user, the server now advertises the update_panel capability and the changelog promises a Lab toggle and progress display, but apps/kimi-code/dist-web is byte-identical to the previous 2.1.1 release (its last sync predates #4159) and its main bundle contains no NotifyUser or update_panel handling. The agent can therefore emit updates that the shipped browser UI cannot present; sync and commit the code-app bundle with this release, or remove the web capability and claim until that bundle is available.
AGENTS.md reference: AGENTS.md:L18-L18
Useful? React with 👍 / 👎.
|
|
||
| - [#4145](https://github.com/MoonshotAI/kimi-code/pull/4145) [`242ac23`](https://github.com/MoonshotAI/kimi-code/commit/242ac230006447f8601e1678c49182bfbae89fe9) Thanks [@wbxl2000](https://github.com/wbxl2000)! - Add `kimi app [path]` to open a new desktop chat in the current or specified directory. | ||
|
|
||
| - [#4061](https://github.com/MoonshotAI/kimi-code/pull/4061) [`20a2cea`](https://github.com/MoonshotAI/kimi-code/commit/20a2cea72f5d2c4be3a3845e4ff24ce302a6630e) Thanks [@chengluyu](https://github.com/chengluyu)! - Cap the agent's in-turn wait for background tasks at 90 seconds (longer timeouts are no longer accepted), discourage repeated waits, and let a new message end the wait immediately. |
There was a problem hiding this comment.
Custom agents and persisted tool-call flows that previously supplied a WaitFor.timeout between 91 and 600 seconds now fail schema validation because WAIT_FOR_MAX_TIMEOUT_S is hard-coded to 90. The release entry names that loss, but there is no config, environment variable, or feature flag that lets the affected population retain the old accepted range, nor a guard for such a compatibility path; provide the required escape hatch or explicit maintainer sign-off before publishing the narrowed contract.
AGENTS.md reference: AGENTS.md:L106-L108
Useful? React with 👍 / 👎.
e724b08 to
5a00484
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5a00484bb3
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
|
|
||
| ### Minor Changes | ||
|
|
||
| - [#4184](https://github.com/MoonshotAI/kimi-code/pull/4184) [`cbd7540`](https://github.com/MoonshotAI/kimi-code/commit/cbd754049a114ee36d1b27675c82819e4963c692) Thanks [@7Sageer](https://github.com/7Sageer)! - Remove the `kaos` / `persistenceKaos` session options and the `createSessionWithKaos` / `resumeSessionWithKaos` client methods, which were already ignored at runtime. |
There was a problem hiding this comment.
The 0.21.0 section lists only the kaos removal and user-turn counts, but this release also ships public additions and type changes including Session.getTowerStatus(), the steerIfActive option for promptWithSkills, and new required WorkspaceTrustInfo disclosure fields. SDK consumers—especially those constructing trust-info values or implementing test doubles—need these contract changes called out, so add corresponding entries and compatibility guidance rather than publishing the release as if it contained only the two listed API changes.
AGENTS.md reference: AGENTS.md:L102-L104
Useful? React with 👍 / 👎.
|
|
||
| ### Patch Changes | ||
|
|
||
| - Updated dependencies [[`cbd7540`](https://github.com/MoonshotAI/kimi-code/commit/cbd754049a114ee36d1b27675c82819e4963c692), [`e5b4b88`](https://github.com/MoonshotAI/kimi-code/commit/e5b4b88ba93d5a27aa9a36cf847eed4f1aa39c48)]: |
There was a problem hiding this comment.
Version 0.8.2 contains a direct apps/vscode/src/runtime/replay-adapter.ts change that strips expanded skill instructions when resuming marked or legacy bundled-skill sessions, with new tests pinning that visible behavior, but this section claims only dependency updates. VS Code users upgrading and reopening those sessions therefore receive an undeclared transcript-rendering change; add a VS Code release entry for it.
AGENTS.md reference: AGENTS.md:L102-L104
Useful? React with 👍 / 👎.
|
|
||
| - [#4031](https://github.com/MoonshotAI/kimi-code/pull/4031) [`33e368a`](https://github.com/MoonshotAI/kimi-code/commit/33e368a1eff8d958856af8e2fb64f4db122b2247) Thanks [@Grapedge](https://github.com/Grapedge)! - Replace the user message bullet ✨ with ❯. | ||
|
|
||
| - [#4221](https://github.com/MoonshotAI/kimi-code/pull/4221) [`2e6821d`](https://github.com/MoonshotAI/kimi-code/commit/2e6821dcbd8034f128b2326a73d4704201b7c108) Thanks [@kimi-agent-bot](https://github.com/kimi-agent-bot)! - web: Improved interactions and fixed known bugs. |
There was a problem hiding this comment.
For browser users this is the only entry covering the newly synchronized code-app bundle, but “Improved interactions and fixed known bugs” identifies neither an interaction nor a bug. Because the web source is outside this repository and the committed bundle changes thousands of generated lines, this wording leaves the shipped behavior impossible to review or understand; replace it with concrete user-facing sentences describing what changed.
AGENTS.md reference: AGENTS.md:L89-L90
Useful? React with 👍 / 👎.
|
|
||
| - [#4183](https://github.com/MoonshotAI/kimi-code/pull/4183) [`c15638b`](https://github.com/MoonshotAI/kimi-code/commit/c15638b0a1234e5aa97659fbc625d7dc4d60d93f) Thanks [@RealKai42](https://github.com/RealKai42)! - Cancel in-flight video uploads when a media read is interrupted. | ||
|
|
||
| - [#4081](https://github.com/MoonshotAI/kimi-code/pull/4081) [`06ebfc8`](https://github.com/MoonshotAI/kimi-code/commit/06ebfc821e304bd4ca3f828ae739954166463d65) Thanks [@sailist](https://github.com/sailist)! - Fix shifted turns and lost replies when a UserPromptSubmit hook injects text; hook JSON output without a "message" field is no longer added to the conversation. |
There was a problem hiding this comment.
For an existing UserPromptSubmit hook that emits valid structured JSON without a message field, such as {"continue":true}, the previous implementation fell back to injecting stdout into the conversation, while the new structuredOutput guard silently returns no hook part. This entry names the loss, but the hook receives no error and there is no config, environment variable, flag, compatibility path, or explicit release sign-off allowing affected hook scripts to retain their previous behavior; preserve the fallback behind an escape hatch or obtain the required sign-off before publishing.
AGENTS.md reference: AGENTS.md:L106-L108
Useful? React with 👍 / 👎.
|
|
||
| - [#4031](https://github.com/MoonshotAI/kimi-code/pull/4031) [`33e368a`](https://github.com/MoonshotAI/kimi-code/commit/33e368a1eff8d958856af8e2fb64f4db122b2247) Thanks [@Grapedge](https://github.com/Grapedge)! - Add sticky user message summaries in fullscreen mode; click a summary to jump to the original message. | ||
|
|
||
| - [#4074](https://github.com/MoonshotAI/kimi-code/pull/4074) [`c9f01f0`](https://github.com/MoonshotAI/kimi-code/commit/c9f01f07388abc2cdca40c9c5267f58c65086710) Thanks [@tpoisonooo](https://github.com/tpoisonooo)! - Tower mode: deterministic orchestration with TowerRebase and LLM-free status, goal/tower mutual exclusion, and workspace protection. |
There was a problem hiding this comment.
Tower workers and reviewers previously received the Agent tool plus the explore and plan subagent allowlist, but this release removes both, so Tower missions that delegated investigation from inside a worker can no longer run that workflow. The entry mentions orchestration and workspace protection without naming this lost capability, and disabling Tower does not preserve the old Tower behavior; name the loss and provide an escape hatch or explicit maintainer sign-off before publishing.
AGENTS.md reference: AGENTS.md:L106-L108
Useful? React with 👍 / 👎.
* feat(kap-server): add remote control device deletion (#4192) (#4195) * feat(kap-server): add remote control device deletion * fix(kap-server): coordinate with the local tunnel before deleting this machine's device Co-authored-by: liruifengv <liruifeng1024@gmail.com> * ci: enable recoverable publishing for the October release (#4218) * ci: publish core releases from isolated release branches * ci: simplify release recovery to the CLI package * ci: publish VS Code from approved release commits * test: await release test registration for type-aware lint * test: avoid assuming order of independent tower events * ci: limit release migration to existing publishing flow * chore: sync web dist from code-app (#4221) * chore: sync web dist from code-app code-app: f828bac439ab5b9266ff8387c2afb96ecfcb59ae * ci: retrigger (flaky rpc-events teardown ENOTEMPTY) --------- Co-authored-by: kimi-agent-bot <kimi-agent-bot@users.noreply.github.com> * ci: release packages (#4220) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> --------- Co-authored-by: qer <wbxl2000@outlook.com> Co-authored-by: liruifengv <liruifeng1024@gmail.com> Co-authored-by: kimi-agent-bot <kimi-agent-bot@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
* feat(kap-server): add remote control device deletion (#4192) (#4195) * feat(kap-server): add remote control device deletion * fix(kap-server): coordinate with the local tunnel before deleting this machine's device Co-authored-by: liruifengv <liruifeng1024@gmail.com> * ci: enable recoverable publishing for the October release (#4218) * ci: publish core releases from isolated release branches * ci: simplify release recovery to the CLI package * ci: publish VS Code from approved release commits * test: await release test registration for type-aware lint * test: avoid assuming order of independent tower events * ci: limit release migration to existing publishing flow * chore: sync web dist from code-app (#4221) * chore: sync web dist from code-app code-app: f828bac439ab5b9266ff8387c2afb96ecfcb59ae * ci: retrigger (flaky rpc-events teardown ENOTEMPTY) --------- Co-authored-by: kimi-agent-bot <kimi-agent-bot@users.noreply.github.com> * ci: release packages (#4220) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * feat(tui): queue /compact behind the running turn (#4276) * feat(tui): queue /compact behind the running turn * docs: note /compact queueing while the agent is busy * fix(tui): harden queued /compact edge cases - keep /compact blocked while a compaction holds the turn (busy reason normalized to compacting) - treat the first queued /compact as a FIFO barrier for Ctrl-S manual steer - gate the failed-fire retry timer on the session actually going idle * fix(tui): cover remaining /compact queue review edges - advertise Ctrl-S in the queue pane only for items ahead of the compact barrier - queue /compact when the awaited lazy session creation lands on a running turn - document the auto-compaction skip for queued /compact in docs and the changeset * fix(tui): guard queued /compact against concurrent compactions - requeue without firing when another compaction already started (the SDK's begin() swallows this as a silent no-op; declared residual window needs the contract to expose the result) - restore the input when a duplicate /compact is refused after lazy session creation * fix(tui): render requeued /compact and mark it always available in docs - request a render after restoring a compact item to the queue so it does not appear vanished during backoff - the reference table's availability column now matches the queueing behavior (/compact is accepted during streaming and runs in order) * fix(tui): steer messages ahead of a queued /compact automatically The automatic (tower/WaitFor) steer batch now slices at the first queued /compact just like the manual path, so pre-barrier prompts still end a WaitFor promptly instead of being blocked by an unrelated compact item. Batches containing bash or skill items ahead of the barrier keep their existing all-or-nothing semantics. * test(tui): drive the tower steer barrier case through the WaitFor event path * chore: downgrade the /compact queue changeset to patch (#4279) * feat: add cloud-gated model switch confirmation (#4286) * feat(cli): add cloud-gated model switch confirmation dialog * feat(server): expose named client_configs endpoint for cloud-controlled clients * fix(server): keep broadcasting compaction events to transcript subscribers * Revert "fix(server): keep broadcasting compaction events to transcript subscribers" This reverts commit e489915. * fix(cli): address review findings on the model switch confirmation * test(cli): fix the auth-flow gate test typing * fix: close the login-path holes in the model switch gate - install the refreshed model catalog before matching the cloud rule, so an alias introduced by /login no longer bypasses the confirmation - return 'gated' from refreshConfigAfterLogin and forward the picked effort to the dialog; setDefaultModel skips its follow-up activation while the gate owns the switch - ignore Escape once a dialog action was chosen so the input lock survives the new-session creation * fix: keep displaced panels alive when the new-session switch settles The async finalizer of a settled dialog action now only restores the editor while the model-switch dialog still owns the editor replacement, so an approval or question panel mounted over it is never wiped out with its reverse-RPC request still pending. * feat: support id lists and always-fresh model switch confirmation configs (#4292) * feat: accept id lists in model switch confirmation rules * feat: fetch client configs fresh on every request The /client_configs endpoint now proxies the cloud per request instead of serving a copy cached at server startup, so config changes take effect without a server restart. The startup warm-up and the in-memory store are removed; region is resolved per request. --------- Co-authored-by: qer <wbxl2000@outlook.com> Co-authored-by: Kimi Agent <zhangkaiyi@moonshot.ai> Co-authored-by: kimi-agent-bot <kimi-agent-bot@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and publish to npm yourself or setup this action to publish automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to release/2026-10-13, this PR will be updated.
Releases
@moonshot-ai/kimi-code@2.2.0
Minor Changes
#4159
0a3a2fbThanks @RealKai42! - Add an opt-in NotifyUser progress update feature (off by default): new sessions can show the agent's mid-turn progress with an optional one-line title, enabled in the TUI/experimentsor the web/Desktop settings' Lab section; set[experimental] notify_user = trueorKIMI_CODE_EXPERIMENTAL_NOTIFY_USER=1to turn it on.#4145
242ac23Thanks @wbxl2000! - Addkimi app [path]to open a new desktop chat in the current or specified directory.#4061
20a2ceaThanks @chengluyu! - Cap the agent's in-turn wait for background tasks at 90 seconds (longer timeouts are no longer accepted), discourage repeated waits, and let a new message end the wait immediately.#3998
09af3b4Thanks @tpoisonooo! - Tower mode: a message sent while the agent is running aborts the in-progress turn and runs as its own turn.#4031
33e368aThanks @Grapedge! - Add sticky user message summaries in fullscreen mode; click a summary to jump to the original message.#4074
c9f01f0Thanks @tpoisonooo! - Tower mode: deterministic orchestration with TowerRebase and LLM-free status, goal/tower mutual exclusion, and workspace protection.#4059
1f6f0b1Thanks @7Sageer! - Add the KIMI_CODE_TRUST_WORKSPACE environment variable to mark the workspace as trusted.#4165
afd95f6Thanks @Grapedge! - Replace user skill cards with highlighted prompts and support Ctrl-S for skill prompts.Patch Changes
#4183
c15638bThanks @RealKai42! - Cancel in-flight video uploads when a media read is interrupted.#4081
06ebfc8Thanks @sailist! - Fix shifted turns and lost replies when a UserPromptSubmit hook injects text; hook JSON output without a "message" field is no longer added to the conversation.#4154
419acedThanks @7Sageer! - Fix media file reading on the first turn after resuming a session.#4083
f409caaThanks @sailist! - Cron tasks from the source session no longer fire inside a forked session.#4152
f0bc988Thanks @7Sageer! - Forked subagents now inherit the parent's available agent types, including custom agents.#4057
a940f2fThanks @huangzheng2016! - Fix the permission mode switch in the goal-start approval prompt not taking effect on the v2 engine.#4081
06ebfc8Thanks @sailist! - Fix messages rendering out of place in sessions with a UserPromptSubmit hook.#4153
7091138Thanks @RealKai42! - Let the agent write one- or two-sentence progress notes between tool calls instead of bare status lines; in the experimental Updates panel (KIMI_CODE_EXPERIMENTAL_NOTIFY_USER=1), stop asking the agent to repost its mid-turn text as an update and remind it to post updates less often.#4054
4fbe065Thanks @RealKai42! - Stop prompting the agent to send NotifyUser updates when a session is opened in a host that does not display them.#4091
21406fbThanks @sailist! - Omit max_tokens for models without an explicit output limit, fixing repeated 400 errors on strict serving stacks (e.g. bare vLLM) when the model's max output size is unknown.#4184
cbd7540Thanks @7Sageer! - Remove the never-implementedexecutor = "kaos"option from MCP stdio server configuration; configs using it now fail validation at load time.#2919
0f052feThanks @duyphan154! - Sanitize foreground Bash tool output so captured terminal escape sequences no longer change the host terminal state.#4164
3395f5bThanks @huangzheng2016! - Show a "Shutting down" status hint while shutdown hooks run during exit, and let a Ctrl+C during that window force an immediate exit instead of waiting for the hook timeout.#4033
eec9bbdThanks @Grapedge! - Feedback surveys now appear less often.#4160
72fc1d0Thanks @kimi-agent-bot! - Show "Working…" in the status spinner while thinking.#4056
395d537Thanks @7Sageer! - Replace inline MCP commands in the workspace trust prompt with activation summaries and configuration paths for manual inspection.#4076
e3bf50cThanks @7Sageer! - Remove interruption reminders from the conversation when their interrupted turns are undone.#4031
33e368aThanks @Grapedge! - Replace the user message bullet ✨ with ❯.#4221
2e6821dThanks @kimi-agent-bot! - web: Improved interactions and fixed known bugs.@moonshot-ai/agent-core-v2@0.5.0
Minor Changes
5b93669Thanks @liruifengv! - Add anallow_ignored_globsoption to the workspacefs:listrequest: matching entries are listed even when gitignored, ignored ancestor directories of a match stay traversable, and dot-path allowances requireshow_hidden: true.Patch Changes
5b93669Thanks @liruifengv! - Match workspace fs globs (include_globs,exclude_globs,allow_ignored_globs) with the standard picomatch engine:a/**/bno longer matches paths likea/xxb, anda/**now also matchesaitself.@moonshot-ai/kimi-code-sdk@0.21.0
Minor Changes
#4184
cbd7540Thanks @7Sageer! - Remove thekaos/persistenceKaossession options and thecreateSessionWithKaos/resumeSessionWithKaosclient methods, which were already ignored at runtime.#4156
e5b4b88Thanks @Grapedge! - Add cumulative main-agent user-turn counts to session recovery snapshots.@moonshot-ai/pi-tui@0.85.0
Minor Changes
33e368aThanks @Grapedge! - Add fullscreen layout effects that settle render requests before terminal output.kimi-code@0.8.2
Patch Changes
cbd7540,e5b4b88]:@moonshot-ai/acp-server@0.0.2
Patch Changes
5b93669,5b93669]:@moonshot-ai/kap-server@0.2.4
Patch Changes
5b93669,5b93669]:@moonshot-ai/klient@0.1.3
Patch Changes
5b93669,5b93669]:@moonshot-ai/migration-legacy@0.1.17
Patch Changes
5b93669,5b93669]:@moonshot-ai/remote-control@0.0.1
Patch Changes
5b93669,5b93669]: