Add support for Kimi Code (Moonshot AI) by Krishnachaitanyakc · Pull Request #751 · git-ai-project/git-ai

Krishnachaitanyakc · 2026-03-18T21:21:57Z

Summary

Closes #639

Add agent detection for Kimi Code commits via email (noreply@moonshot.cn, noreply@kimi.ai) and GitHub username (kimi-code[bot])
Add KimiCodePreset checkpoint preset accepting standard hook input (session_id, cwd, model, hook_event_name, tool_input)
Wire up kimi-code subcommand in the checkpoint handler and update help text

Test plan

All existing agent_detection tests pass (20/20)
New tests for Kimi Code email and username detection pass
Doc tests pass
Full project builds successfully
CI passes

devin-ai-integration

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 2 additional findings.

Krishnachaitanyakc · 2026-03-21T16:13:21Z

@svarlamov can you review this?

svarlamov · 2026-03-24T03:08:34Z

Sorry for the late reply @Krishnachaitanyakc! This looks good, but is it possible to get transcripts for Kimi Code too?

Parse optional transcript field from hook_input JSON on PostToolUse events, enabling Kimi Code to pass conversation history using the standard AiTranscript format. Backward-compatible — omitting the field defaults to None as before. Addresses review feedback on git-ai-project#751.

svarlamov · 2026-03-27T21:54:38Z

@Krishnachaitanyakc Thanks for the PR! Was just reviewing -- does Kimi have hooks? The PR seems to be missing the logic to actually install hooks for Kimi (ref the logic under mdm/)

Krishnachaitanyakc · 2026-03-28T16:06:59Z

@svarlamov Good catch, sorry I missed the hook installer — should've been part of the original PR. Just pushed the KimiCodeInstaller that writes HOOK.md + shell scripts to ~/.config/agents/hooks/.

svarlamov · 2026-03-28T17:22:18Z

Ty! And did you have some time to test locally/verify the kimi extension? I'm trying to get it setup on my end locally as well.

Btw @Krishnachaitanyakc are you in our Discord? https://discord.gg/qMACzkVuN

Krishnachaitanyakc · 2026-04-02T05:15:00Z

@svarlamov yes, also added a comprehensive integration test suite covering preset parsing, error handling, and E2E checkpoint-commit-attribution

- Add email detection for noreply@moonshot.cn and noreply@kimi.ai - Add username detection for kimi-code[bot] - Add KimiCodePreset checkpoint preset with standard hook input format - Wire up "kimi-code" subcommand in checkpoint handler and help text

Parse optional transcript field from hook_input JSON on PostToolUse events, enabling Kimi Code to pass conversation history using the standard AiTranscript format. Backward-compatible — omitting the field defaults to None as before. Addresses review feedback on git-ai-project#751.

Cover preset parsing (AI/human checkpoint routing, model extraction, file path handling, inline transcript), error handling (missing fields, invalid JSON), and end-to-end checkpoint-commit-attribution flow.

- Remove stale tests/kimi_code.rs (moved to tests/integration/ but old file remained tracked, causing duplicate module errors) - Apply rustfmt to tests/integration/kimi_code.rs

Register kimi-code in AI_AUTHOR_NAMES and is_known_checkpoint_preset so E2E tests correctly identify Kimi Code as an AI agent.

The notes_sync_push_propagates_authorship_notes_to_remote test failed intermittently on windows-latest (wrapper mode). This is a known flaky test unrelated to the Kimi Code changes — the same test passes on main and on PR git-ai-project#851 which has an identical base.

Resolve conflict in git_ai_handlers.rs: include both KimiCode and Firebender presets added on each branch. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Remove unconditional println! that printed the full serialized TOML config to stdout on every hook installation. No other agent installer has debug output in its install_hooks method.

test_kimi_code_session_transcript_from_disk sets process-global KIMI_SHARE_DIR which can race with parallel tests. Adding #[serial] prevents flaky test failures.

- Move test_kimi_code_session_transcript_from_disk to reuse_tests_in_worktree_with_attrs! with #[serial] so the generated _in_worktree variant is also serialized - Fix clippy::collapsible_if in agent_presets.rs - Apply cargo fmt formatting

svarlamov · 2026-05-05T02:45:21Z

Hey @Krishnachaitanyakc -- the preset rewrite landed on main!! 🥳

I'm going to close this and ask that if you do have time, a new PR for the same feature, but based on the new agent preset structure (which is way simpler, wooohooo!!!!)

Adds first-class support for kimi-cli, Moonshot AI's Python coding assistant, against the new agent-preset framework that landed on main in May 2026. Replaces the closed PR git-ai-project#751, which was scoped to the old per-preset arms in git_ai_handlers and is structurally incompatible with the rewritten orchestrator. Preset (`presets/kimi_code.rs`): - Pure parser implementing `AgentPreset::parse`. Handles `PreToolUse` / `PostToolUse` for `WriteFile`, `StrReplaceFile`, and `Shell`. Accepts the alias keys (`tool_use_id` / `toolUseId`) in addition to kimi-cli's native `tool_call_id`. - Strict event handling: lifecycle events (`SessionStart`, `Stop`, `SubagentStart`, `PostToolUseFailure`, etc.) return `PresetError` rather than fabricating a phantom file-edit checkpoint. - `transcript_source: None` for now. kimi-cli stores sessions on disk under `$KIMI_SHARE_DIR/sessions/...` but the hook payload does not surface that path; a dedicated reader can land in a follow-up. Installer (`mdm/agents/kimi_code.rs`): - Writes `[[hooks]]` entries into `~/.kimi/config.toml` per the upstream hook documentation. No `matcher` field; the preset itself filters tool events. - Idempotent — repeat install with the desired command is a no-op. - Token-based ownership matching (`is_git_ai_kimi_command`) so that a hypothetical sibling preset like `kimi-code-pro` is never misidentified or clobbered. - Uninstall removes only kimi-code entries; user-defined hooks survive. - 16 unit tests covering fresh install, idempotency, dedup with same and differing paths, dry-run, parent-dir creation, mixed uninstall, missing-config and corrupt-TOML tolerance. Tool classification (`bash_tool.rs`): - New `Agent::KimiCode` variant. `WriteFile`/`StrReplaceFile` map to `FileEdit`, `Shell` maps to `Bash`, all else `Skip`. Wiring: - `presets/mod.rs::resolve_preset` accepts `kimi-code`. - `mdm/agents/mod.rs::get_all_installers` includes `KimiCodeInstaller`. - `git_ai_handlers` help text lists `kimi-code`. - `tests/integration/repos/test_file.rs::AI_AUTHOR_NAMES` and `test_repo.rs::is_known_checkpoint_preset` learn `kimi-code`. Tests: 14 preset unit tests, 16 installer unit tests, 11 integration tests including 3 full E2E flows (`WriteFile` attribution, pre+post isolation of human vs AI lines, `StrReplaceFile` attribution). `cargo fmt --check`, `cargo clippy --all-targets -- -D warnings`, `RUSTDOCFLAGS="-D warnings" cargo doc --no-deps`, and the full lib test suite (1467 tests) all pass locally. Closes git-ai-project#639

Krishnachaitanyakc marked this pull request as ready for review March 19, 2026 02:13

devin-ai-integration Bot reviewed Mar 19, 2026

View reviewed changes

svarlamov mentioned this pull request Mar 24, 2026

Kimi not supported #773

Closed

This comment was marked as resolved.

Sign in to view

Krishnachaitanyakc added 7 commits April 2, 2026 23:21

Fix rustfmt formatting in agent_detection test

9975ab2

Add integration tests for KimiCodePreset

6b9c924

Cover preset parsing (AI/human checkpoint routing, model extraction, file path handling, inline transcript), error handling (missing fields, invalid JSON), and end-to-end checkpoint-commit-attribution flow.

Fix CI: remove duplicate test file and fix formatting

a202626

- Remove stale tests/kimi_code.rs (moved to tests/integration/ but old file remained tracked, causing duplicate module errors) - Apply rustfmt to tests/integration/kimi_code.rs

Fix unused variable warning in kimi_code E2E test

9ed5bb7

Add kimi-code to test infrastructure

d35f750

Register kimi-code in AI_AUTHOR_NAMES and is_known_checkpoint_preset so E2E tests correctly identify Kimi Code as an AI agent.

Krishnachaitanyakc force-pushed the feat/kimi-code-support-639 branch from da05c9b to d35f750 Compare April 3, 2026 06:38

Krishnachaitanyakc and others added 4 commits April 3, 2026 08:00

Merge origin/main into feat/kimi-code-support-639

5496a8a

Resolve conflict in git_ai_handlers.rs: include both KimiCode and Firebender presets added on each branch. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

kimi

4d6110d

kimi support

38af66f

This comment was marked as resolved.

Sign in to view

fix: remove debug println in kimi_code install_hooks

3d9b952

Remove unconditional println! that printed the full serialized TOML config to stdout on every hook installation. No other agent installer has debug output in its install_hooks method.

This comment was marked as resolved.

Sign in to view

Krishnachaitanyakc added 2 commits April 10, 2026 17:53

fix: add #[serial] to test that mutates KIMI_SHARE_DIR env var

afd05e3

test_kimi_code_session_transcript_from_disk sets process-global KIMI_SHARE_DIR which can race with parallel tests. Adding #[serial] prevents flaky test failures.

svarlamov closed this May 5, 2026

Krishnachaitanyakc mentioned this pull request May 7, 2026

feat: add Kimi Code (Moonshot AI) preset and hook installer #1298

Open

7 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add support for Kimi Code (Moonshot AI)#751

Add support for Kimi Code (Moonshot AI)#751
Krishnachaitanyakc wants to merge 14 commits into
git-ai-project:mainfrom
Krishnachaitanyakc:feat/kimi-code-support-639

Krishnachaitanyakc commented Mar 18, 2026 •

edited by devin-ai-integration Bot

Loading

Uh oh!

devin-ai-integration Bot left a comment

Uh oh!

Krishnachaitanyakc commented Mar 21, 2026

Uh oh!

svarlamov commented Mar 24, 2026

Uh oh!

svarlamov commented Mar 27, 2026

Uh oh!

Krishnachaitanyakc commented Mar 28, 2026

Uh oh!

svarlamov commented Mar 28, 2026

Uh oh!

Krishnachaitanyakc commented Apr 2, 2026

Uh oh!

This comment was marked as resolved.

Uh oh!

This comment was marked as resolved.

Uh oh!

This comment was marked as resolved.

Uh oh!

svarlamov commented May 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

Krishnachaitanyakc commented Mar 18, 2026 • edited by devin-ai-integration Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

devin-ai-integration Bot left a comment

Choose a reason for hiding this comment

✅ Devin Review: No Issues Found

Uh oh!

Krishnachaitanyakc commented Mar 21, 2026

Uh oh!

svarlamov commented Mar 24, 2026

Uh oh!

svarlamov commented Mar 27, 2026

Uh oh!

Krishnachaitanyakc commented Mar 28, 2026

Uh oh!

svarlamov commented Mar 28, 2026

Uh oh!

Krishnachaitanyakc commented Apr 2, 2026

Uh oh!

This comment was marked as resolved.

Uh oh!

This comment was marked as resolved.

Uh oh!

This comment was marked as resolved.

Uh oh!

svarlamov commented May 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Krishnachaitanyakc commented Mar 18, 2026 •

edited by devin-ai-integration Bot

Loading