Greptile Summary

This PR adds Kimi Code CLI as a new non-interactive LLM backend by introducing KimiAdapter, registering it in the provider registry and wizard config, forwarding KIMI_* env vars through the subprocess sanitizer, and shipping 7 unit tests. All previous review concerns (false auth on OPENAI_API_KEY, hollow no-assertion test, dead tomli fallback, silent empty-parse, and ValueError vs RuntimeError contract mismatch) appear to be fully addressed in this revision.

Confidence Score: 5/5

Safe to merge — all prior P0/P1 findings from earlier review rounds have been addressed; remaining comments are P2 style nits only.

No new P0 or P1 issues found. Auth check uses KIMI_API_KEY correctly, parse() raises RuntimeError on empty output (consistent with runner.py's error contract), the hollow test was replaced with one that has real assertions, and the dead tomli fallback was removed. Only minor style issues remain.

app/integrations/llm_cli/kimi.py — misleading _ = stderr / _ = returncode suppressions in parse()

Important Files Changed

Sequence Diagram

sequenceDiagram
    participant Caller
    participant CLIBackedLLMClient
    participant KimiAdapter
    participant _check_kimi_auth
    participant kimi_CLI

    Caller->>CLIBackedLLMClient: invoke(prompt)
    CLIBackedLLMClient->>KimiAdapter: detect()
    KimiAdapter->>kimi_CLI: kimi --version
    kimi_CLI-->>KimiAdapter: version string
    KimiAdapter->>_check_kimi_auth: check KIMI_API_KEY / config.toml
    _check_kimi_auth-->>KimiAdapter: (logged_in, detail)
    KimiAdapter-->>CLIBackedLLMClient: CLIProbe
    alt not installed or not logged in
        CLIBackedLLMClient-->>Caller: raise RuntimeError
    end
    CLIBackedLLMClient->>KimiAdapter: build(prompt, model, workspace)
    KimiAdapter-->>CLIBackedLLMClient: CLIInvocation(argv, stdin)
    CLIBackedLLMClient->>kimi_CLI: subprocess.run(argv, stdin=prompt, env=KIMI_* filtered)
    kimi_CLI-->>CLIBackedLLMClient: stdout / returncode
    alt returncode != 0
        CLIBackedLLMClient->>KimiAdapter: explain_failure()
        CLIBackedLLMClient-->>Caller: raise RuntimeError
    end
    CLIBackedLLMClient->>KimiAdapter: parse(stdout)
    alt stdout empty
        KimiAdapter-->>Caller: raise RuntimeError
    end
    KimiAdapter-->>CLIBackedLLMClient: content string
    CLIBackedLLMClient-->>Caller: LLMResponse(content)

_{Reviews (5): Last reviewed commit: "fix(llm-cli): final restoration of Claud..." | Re-trigger Greptile}

working on greptile suggestion

@greptile review

@greptile full-review

@greptile review

@muddlebee @yashksaini-coder PTAL!!

@gitsofaryan

missing

also check other Acceptance criteria, many missing gaps

missing entry for env vars at .env.example

refer app/integrations/llm_cli/AGENTS.md for specs

@greptile-apps review

also @gitsofaryan an e2e demo

pls igore my previous reviews. it was outdated.

@muddlebee Hey, facing a payment issue with the Kimi API CLI login works but API calls fail with exit code 75 (quota/billing error). Can you help check if the account billing is sorted? Thanks

@muddlebee please suggest fixes if any!

@gitsofaryan unfortunately team can't provide kimi cli subscription or auth.

and this error ideally shouldn't come

Hey @gitsofaryan, took a look at the merge conflict locally. There's just one conflict in app/integrations/llm_cli/runner.py, caused by PR #1295 (merged yesterday) which extracted the subprocess env allowlist into a new app/integrations/llm_cli/subprocess_env.py module while your branch was modifying the inline list to add KIMI_.

Fix is two steps. Run these from your branch:

git fetch origin
git merge origin/main

# Conflict in runner.py — take main's version (the new shape uses subprocess_env.py)
git checkout --theirs app/integrations/llm_cli/runner.py

# Add KIMI_ to the prefix list in the new module
sed -i 's/("LC_", "CODEX_", "CLAUDE_")/("LC_", "CODEX_", "CLAUDE_", "KIMI_")/' app/integrations/llm_cli/subprocess_env.py

# Stage and finish the merge
git add app/integrations/llm_cli/runner.py app/integrations/llm_cli/subprocess_env.py
git commit
git push

The KIMI_ prefix is what makes KIMI_API_KEY, KIMI_BASE_URL, etc. forward through to the subprocess env, which is what your test at tests/integrations/llm_cli/test_kimi_adapter.py:220-243 already asserts. After the merge, that prefix lives in subprocess_env.py instead of runner.py, but the behavior is identical.

I verified locally: ruff check and ruff format --check are clean after the resolution, no imports break, all your Kimi-specific changes survive the auto-merge of the other files.

Separately on the Kimi API quota error (exit code 75) — that's an account-side billing thing, not something the merge fix addresses. Worth checking your Moonshot AI billing dashboard or rotating to an account with credit. Good luck!

@gitsofaryan, ended up pushing the resolution to your branch directly since you've got Allow edits from maintainers enabled — commit a4aa14f5 is the merge with main. The PR should flip from CONFLICTING to MERGEABLE shortly once GitHub re-evaluates it.

You can ignore the previous comment with the manual steps. To sync your local clone, just:

git fetch origin
git reset --hard origin/feat/kimi-cli-1112  # if you have no local changes ahead
# or
git pull --rebase  # if you do have local work to keep

Same fix as I described before: took main's runner.py (since PR #1295 extracted the env list into subprocess_env.py), then added "KIMI_" to the prefix tuple in the new module so your env var threading still works. Lint, format, and the structure of your Kimi tests all check out locally. Sorry for any toes stepped on — happy to revert if you'd rather redo it yourself.

Heads up — main moved between my first push and now (PR #1149 landing the Cursor Agent CLI added the same shape of changes you did, in the same files). Pushed a second merge commit 4225c28f to clear the new conflicts. PR is now MERGEABLE (state was UNSTABLE last I checked because of CI status, not merge conflicts — worth a glance at any failing checks before you merge).

Resolutions in this second merge were all "combine both additions":

• app/integrations/llm_cli/subprocess_env.py — prefix tuple now has both CURSOR_ and KIMI_
• app/config.py — provider early-return now lists both cursor and kimi
• .env.example — kept both Cursor and Kimi env-var blocks

If you've made any local changes since this morning, git pull --rebase should be a clean rebase. Otherwise git fetch origin && git reset --hard origin/feat/kimi-cli-1112 to sync.

@muddlebee fixed (utf encoding issue on windows) , im unable to generate kimi API not able to find free/dev API usage.

@yashksaini-coder thanks, also addressed the changes.

@gitsofaryan thats not an issue, given the e2e test cases are covered, I;m taking this up, will review and add necessary changes if required :)

assigning to myself

🎊 Achievement unlocked: PR Merged. @gitsofaryan passed code review, survived CI, and shipped. Respect. 🤝

👋 Join us on Discord - OpenSRE : hang out, contribute, or hunt for features and issues. Everyone's welcome.

@muddlebee Thanks man.