Full pipeline

run_full_pipeline() composes run_typo_generation() with run_autocorrect2_check() and returns a FullPipelineResult:

Full Pipeline Workflow

flowchart TD
    A["Source words and<br>generation settings"] --> B["Generate typo<br>mappings"]
    B --> C["TypoGenerationResult"]
    C -->|"candidate mappings"| D["Create B0X<br>AutoCorrect2 candidates"]
    D --> E["Check against<br>AutoCorrect2"]
    E --> F["AutoCorrect2CheckResult"]
    C --> G["FullPipelineResult"]
    F --> G

The adaptation step converts each surviving noisy-to-target mapping into an AutoCorrect2CandidateHotstring using the fixed generated option set B0X. Typo generation itself has no dependency on HotstringOptions or AutoCorrect2.

The full pipeline disables stage-specific report writing while composing the two stages and creates one combined report at the end when requested.

Command-line usage

Run the composed workflow with full-pipeline:

uv run python -m hotstring full-pipeline `
    --words-file .\words.txt `
    --single-attempts 100 `
    --multi-attempts 100 `
    --multi-min-length 5 `
    --workers 4 `
    --report .\reports\full-pipeline.txt

The AutoCorrect2 project directory is resolved from --project-dir, the process environment, an explicit dotenv file, or the project-root .env.

The command remains read-only unless --write-accepted is supplied. It does not accept manually supplied candidates because it creates B0X candidates from the typo-generation result.

See also: Command-line interface for all shared options.