CLI Reference¶
Global¶
ai-act --help
ai-act --version
classify¶
Classifies a system descriptor into a risk tier.
ai-act classify <system.yaml>
ai-act classify <system.yaml> --json
Options:
--json: machine-readable output
check¶
Runs requirement-level compliance checks.
ai-act check <system.yaml>
ai-act check <system.yaml> --json
Options:
--json: machine-readable output
JSON includes summary, findings, transparency, gpai_summary, and audit_trail.
Each successful run also attempts a best-effort history append to .eu_ai_act/history.jsonl.
checklist¶
Generates checklist artifacts.
ai-act checklist <system.yaml> --format json
ai-act checklist <system.yaml> --format md -o checklist.md
ai-act checklist <system.yaml> --format html -o checklist.html
Options:
--format [json|md|html]-o, --output PATH
report¶
Renders report artifacts from classification + checker + transparency + GPAI + checklist context.
ai-act report <system.yaml> --format json
ai-act report <system.yaml> --format md -o report.md
ai-act report <system.yaml> --format html -o report.html
ai-act report <system.yaml> --format pdf -o report.pdf
Options:
--format [json|md|html|pdf]-o, --output PATH
PDF notes:
--output/-ois required for--format pdf- install optional dependency:
pip install -e ".[reporting]"
report runs also append a history event (best effort). History write failures emit warning output but do not fail the command.
dashboard build¶
Builds a multi-system dashboard by scanning descriptor files in a directory.
ai-act dashboard build <descriptor_dir>
ai-act dashboard build <descriptor_dir> --recursive
ai-act dashboard build <descriptor_dir> --include-history --history-path .eu_ai_act/history.jsonl
ai-act dashboard build <descriptor_dir> --output-dir dashboard_artifacts
Outputs:
<output-dir>/dashboard.json<output-dir>/dashboard.html
Options:
--recursive--include-history--history-path PATH--output-dir PATH(default: current working directory)
Error policy:
- Invalid descriptors are skipped and included in the dashboard
errorssection. - Command exits
0if at least one valid system is processed. - Command exits
1if no valid systems are found.
history¶
Inspect persisted audit history events and diffs.
ai-act history list
ai-act history list --event-type check --limit 10 --json
ai-act history show <event_id>
ai-act history diff <older_event_id> <newer_event_id> --json
Options:
history list--system <name>--event-type [check|report]--limit N--json--history-path PATHhistory show--json--history-path PATHhistory diff--json--history-path PATH
Default history path is project-local .eu_ai_act/history.jsonl, resolved from the nearest parent directory containing pyproject.toml. If no project root is found, the current working directory is used.
transparency¶
Evaluates transparency obligations from system descriptors.
ai-act transparency <system.yaml>
ai-act transparency <system.yaml> --json
Options:
--json
gpai¶
Evaluates GPAI obligations from GPAI model descriptors.
ai-act gpai <model.yaml>
ai-act gpai <model.yaml> --json
Options:
--json
validate¶
Schema validation for AI system descriptors.
ai-act validate <system.yaml>
articles¶
Prints article mappings by tier.
ai-act articles
ai-act articles --tier minimal
ai-act articles --tier limited
ai-act articles --tier high_risk
ai-act articles --tier unacceptable
Options:
--tier [minimal|limited|high_risk|unacceptable]
Exit Behavior¶
- Non-zero exits are used for invalid file/schema or command-level runtime failures.
- In CI gating, failure logic is implemented by the composite action policy.