Running the Pipeline

This page covers day-to-day execution patterns using the CLI.

Commands overview

The CLI provides four main commands:

• validate-config: validate YAML and run pre-flight checks (including required env vars)

• list-inputs: show which CSV files will be processed

• run: execute the pipeline (one output directory per CSV)

• normalize-assets: copy standard assets into existing outputs and optionally rewrite index.html from run.json

Validate configuration

miyabi-qualys-ai-triage-pack validate-config --config config/config.yaml

List inputs

miyabi-qualys-ai-triage-pack list-inputs --config config/config.yaml

If this prints “No CSV files found”, confirm:

• input.input_dir points to the correct folder

• input.file_glob matches your files (for example *.csv)

Run for all inputs

miyabi-qualys-ai-triage-pack run --config config/config.yaml

Expected results:

• output/<csv_name>/index.html

• output/<csv_name>/assets/

• Optional: output/<csv_name>/run.json (when notifications.include_run_metadata: true)

• output/run_summary.json (summary across processed files)

Automatic mode (repo root entrypoint)

If you want the simplest “just run everything” mode (no CLI), run this from the repository root:

python main.py

Behavior:

• Loads configuration from config/config.yaml (or config/config.yml).

• Finds all input files under input.input_dir matching input.file_glob.

• Processes each CSV and generates output/<csv_name>/... just like the CLI.

Run for a single file

miyabi-qualys-ai-triage-pack run --config config/config.yaml --file path/to/my_export.csv

Overrides (without editing config)

The run command supports common overrides:

• --input-dir <path>: override input.input_dir

• --glob "<pattern>": override input.file_glob (example: "*.csv")

• --output-dir <path>: override output.output_dir

• --top-x <n>: override selection.top_x

Example:

miyabi-qualys-ai-triage-pack run --config config/config.yaml --input-dir Insumos/Reports --glob "*.csv" --top-x 15

Dry-run (ingestion only)

Dry-run executes parsing/aggregation and writes candidate artifacts without calling OpenAI:

miyabi-qualys-ai-triage-pack run --config config/config.yaml --dry-run

Expected results:

• output/<csv_name>/candidates.json

• No LLM-generated sections

Failure handling modes

• Default: continues processing multiple inputs even if one fails (--continue-on-error)

• Fail fast: stops on first error

miyabi-qualys-ai-triage-pack run --config config/config.yaml --fail-fast

Normalize existing outputs (assets + optional HTML rewrite)

If you already generated outputs and want to re-copy assets (visual standardization) and optionally rewrite index.html from run.json:

miyabi-qualys-ai-triage-pack normalize-assets --config config/config.yaml --output-dir output

To normalize assets without rewriting HTML:

miyabi-qualys-ai-triage-pack normalize-assets --config config/config.yaml --output-dir output --rewrite-html false

Why this exists:

• Re-copying assets helps keep old outputs visually consistent with updated styling.

• Rewriting index.html from run.json is a best-effort way to rebuild HTML using the current template/escaping rules.

Where to look after a run

• Per-file artifacts: output/<csv_name>/...

• Run summary: output/run_summary.json

See also:

• Outputs (detailed)