Module cleanup 

Post-transcription text cleanup ( “flow”).

Two backends:

• rules (default): pure on-device regex/heuristics — no network, no model
• openrouter: LLM rewrite via chat completions (optional, explicit)

Cleanup is opt-in at the CLI/library boundary so raw ASR stays available.

Re-exports

pub use openrouter::OpenRouterCleanup;

pub use rules::RulesCleanup;

Modules

openrouter

LLM-assisted cleanup via OpenRouter (JOE-1589 structured untrusted-data contract).

rules

On-device rule-based cleanup (JOE-1610).

Structs

CleanupReport

Structured report for a full-result cleanup transaction.

CleanupResult

Result of a cleanup pass over bare text.

Enums

CleanupProviderKind

Where cleanup runs.

CleanupStyle

How aggressively / in what shape to clean transcript text.

SegmentCleanupPolicy

How to treat ASR segments after cleaning the full transcript text.

Constants

MAX_PER_SEGMENT_CHARS

MAX_PER_SEGMENT_COUNT

Hard bounds for per-segment cleanup (failure semantics live here; remote batching belongs to a follow-up issue).

Traits

TextCleanup

Pluggable cleanup backend.

Functions

apply_cleanup

Apply cleanup to a full transcription result.

apply_cleanup_with_segments

Like apply_cleanup with explicit segment policy.

apply_cleanup_with_segments_op

Like apply_cleanup_with_segments with an optional OpContext for cancel, deadline, and progress (JOE-1831).

cleanup_text

Clean a bare string (stdin / text file path) without a full transcription result.