Skip to content

cli-reference.md

Latest commit

 

History

History
146 lines (99 loc) · 3.29 KB

cli-reference.md

File metadata and controls

146 lines (99 loc) · 3.29 KB

CLI Reference

The aegis CLI is built with Typer and organized into command groups.

aegis version

Print the Aegis version and exit.

aegis version

aegis eval

aegis eval run

Run an evaluation against an agent.

aegis eval run [OPTIONS]
Flag Short Type Default Description
--config -c Path — Path to eval.yaml configuration file
--dimensions -d str — Comma-separated dimension IDs to evaluate
--agent -a str — Agent REST endpoint URL
--domain str — Domain filter (e.g. legal, finance)
--fail-under float — Minimum composite score to pass (exit 1 if below)
--output -o Path — Output directory for results

CLI flags override values from the YAML config.

Examples:

# Run with config file
aegis eval run --config eval.yaml

# Run specific dimensions
aegis eval run --dimensions retention_accuracy,citation_validity

# Run with fail threshold
aegis eval run --config eval.yaml --fail-under 0.80

# Run against a REST agent
aegis eval run --config eval.yaml --agent http://localhost:8000/eval

aegis eval dimensions

List all registered evaluation dimensions.

aegis eval dimensions [OPTIONS]
Flag Short Type Default Description
--domain -d str — Filter by domain (legal, finance, safety)

Examples:

aegis eval dimensions
aegis eval dimensions --domain legal

aegis eval compare

Compare two evaluation runs side by side.

aegis eval compare --runs RUN_A --runs RUN_B
Flag Short Type Default Description
--runs -r str — Run ID (provide exactly twice)

aegis eval report

Generate an evaluation report for a completed run.

aegis eval report --run RUN_ID [OPTIONS]
Flag Short Type Default Description
--run -r str — Run ID to report on
--format -f str json Output format: json, text, html, pdf
--output -o Path — Output file path

Examples:

aegis eval report --run abc123 --format json --output report.json
aegis eval report --run abc123 --format html

aegis memory

aegis memory health

Display memory subsystem health status.

aegis memory health

aegis memory audit

Display the memory subsystem audit trail.

aegis memory audit

aegis train

aegis train start

Create a new training job.

aegis train start [OPTIONS]
Flag Short Type Default Description
--model -m str base-agent-v1 Base model name
--trainer -t str amir-grpo Trainer type: amir-grpo or grpo-sg
--lr float 1e-5 Learning rate

aegis train status

Show the status of a training job.

aegis train status --job-id JOB_ID
Flag Short Type Default Description
--job-id -j str — Training job ID