HKUDS/DeepTutor · error · RuntimeError

CodeBuddy CLI is required to sync account models.

Error message

CodeBuddy CLI is required to sync account models.

What it means

fetch_codebuddy_models() needs the CodeBuddy CLI executable (codebuddy or cbc) on PATH to query the account's model catalog via a subprocess; shutil.which() found neither, so it raises RuntimeError. The SDK alone isn't enough — model sync shells out to the CLI.

Source

Thrown at deeptutor/services/llm/provider_core/codebuddy_provider.py:788

        return
    async with _API_KEY_ENV_LOCK:
        previous = os.environ.get(_CODEBUDDY_API_KEY_ENV)
        os.environ[_CODEBUDDY_API_KEY_ENV] = api_key
        try:
            yield
        finally:
            if previous is None:
                os.environ.pop(_CODEBUDDY_API_KEY_ENV, None)
            else:
                os.environ[_CODEBUDDY_API_KEY_ENV] = previous


async def fetch_codebuddy_models(api_key: str | None = None) -> list[str]:
    """Return the model catalog currently available to the logged-in account."""
    _load_sdk()
    cli_path = shutil.which("codebuddy") or shutil.which("cbc")
    if not cli_path:
        raise RuntimeError("CodeBuddy CLI is required to sync account models.")

    cli_args = [
        cli_path,
        "--print",
        ".",
        "--model",
        "__deeptutor_list_models__",
        "--output-format",
        "json",
        "--max-turns",
        "1",
    ]
    process_kwargs: dict[str, Any] = {}
    if os.name == "nt":
        command = ["cmd.exe", "/d", "/s", "/c", subprocess.list2cmdline(cli_args)]
        # Windows-only constants; the stubs omit them off-Windows, and mypy
        # cannot narrow on os.name the way it narrows on sys.platform.
        process_kwargs["creationflags"] = (

View on GitHub (pinned to 3e82f13042)

Solutions

1. Install the CodeBuddy CLI and ensure `codebuddy` (or `cbc`) is on PATH (verify with `which codebuddy`).
2. In containers/CI, add the CLI install step to the image.
3. If PATH is the issue, launch the process with an environment that includes the CLI's install dir.

Example fix

# before
# CLI not installed
models = await fetch_codebuddy_models()

# after
# Dockerfile / CI
RUN npm install -g @codebuddy/cli
models = await fetch_codebuddy_models()

import shutil
if not (shutil.which("codebuddy") or shutil.which("cbc")):
    print("CodeBuddy CLI missing — install it and ensure it is on PATH")

def codebuddy_cli_on_path() -> bool:
    import shutil
    return shutil.which("codebuddy") is not None or shutil.which("cbc") is not None

try:
    models = await fetch_codebuddy_models()
except RuntimeError as e:
    if "CLI is required" in str(e):
        # skip model sync, use a static fallback list
        models = ["fallback-model"]
    else:
        raise

When it happens

Trigger: Calling fetch_codebuddy_models() on a machine without the CodeBuddy CLI installed, or where the binary isn't on PATH (installed via nvm-style isolated location, GUI session PATH, container image).

Common situations: Docker/CI images that pip-install the SDK but never install the CLI; SSH non-login shells with trimmed PATH; macOS app-launched processes missing /usr/local/bin.

Related errors

• codebuddy-agent-sdk is not installed. Install it with `pytho
• Installed codebuddy-agent-sdk does not expose query().
• The installed codebuddy-agent-sdk does not accept the turn l
• {output.strip() or "CodeBuddy did not return an account mode
• {_SIGN_IN_HINT}

AI-assisted analysis of HKUDS/DeepTutor@3e82f13042 (2026-08-27). Data as JSON: /api/errors/eac9f6d1aacd29cf. Report an issue: GitHub.