Greptile Summary

Adds ProgressOutput::Quiet to fully suppress all progress output (spinners, text-mode lines, println writes) for --quiet/--silent CLI use cases. The implementation is complete and consistent: early-return guards are placed in update(), println(), refresh_once(), and the background thread's start() function, and output() correctly gives Quiet precedence over the CLX_TEXT_MODE environment variable.

Confidence Score: 5/5

• This PR is safe to merge; the changes are minimal, focused, and all code paths that produce output are correctly guarded.
• No P0 or P1 issues found. All four output-producing paths (update, println, refresh_once, background thread start) are correctly guarded. The output() function's precedence logic is sound and matches the documented intent. The #[non_exhaustive] attribute is a deliberate, appropriate design choice. Only P2 observations remain (no behavior test for Quiet mode).
• No files require special attention.

Important Files Changed

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["output() called"] --> B["Read OUTPUT mutex copy"]
    B --> C{stored == Quiet?}
    C -- Yes --> D["return Quiet\nenv vars cannot override"]
    C -- No --> E{CLX_TEXT_MODE=1?}
    E -- Yes --> F["return Text"]
    E -- No --> G["return stored"]

    H["update() called"] --> I{is_disabled OR STOPPING\nOR output==Quiet?}
    I -- Yes --> J["return early"]
    I -- No --> K{output==Text?}
    K -- Yes --> L["render_text_mode()"]
    K -- No --> M["notify() triggers start()"]

    M --> N{started OR disabled\nOR Text OR Quiet\nOR stopping?}
    N -- Yes --> O["return, no thread spawned"]
    N -- No --> P["spawn background refresh thread"]

    R["refresh_once() called"] --> S{is_disabled OR output==Quiet?}
    S -- Yes --> T["return Ok immediately"]
    S -- No --> U["render + write frame to terminal"]

_{Reviews (6): Last reviewed commit: "feat(progress): add ProgressOutput::Quie..." | Re-trigger Greptile}

One concern: ProgressOutput is a public enum in a 1.x crate without #[non_exhaustive]; adding a new variant is technically a semver-breaking change for downstream users who exhaustively match on it

Addressed. I don't know if a major version bump is mandated here.

jdx/hk#782 uses this. If this approach is ok, I will follow up with a similar approach for mise.