eh evaluate

Evaluate source files against an eigenspace model.

Usage

eh evaluate [OPTIONS] [PATHS...]

Arguments

Argument	Description
PATHS	File or directory paths to evaluate. Discovers files recursively.

Options

Option	Default	Description
--language LANG	(auto)	Language override (required for stdin)
--model PATH	(bundled)	Path to .npz eigenspace model
--format {human,json,sarif}	human	Output format
--classify	off	Show accept/marginal/reject labels
--strict	off	Treat marginal as reject (exit code 2)
--lenient	off	Treat marginal as accept (exit code 0)
--diff RANGE	(none)	Evaluate only files changed in git revision range
--accept-threshold N	(model)	Override accept threshold
--reject-threshold N	(model)	Override reject threshold
--rank	off	Relative ranking mode: sort by score
--bottom N	(none)	Highlight bottom N files in rank mode
--bottom-pct N	(none)	Highlight bottom N% in rank mode
--top-n N	3	Top features per dimension in attribution
--directive-threshold N	0.3	Minimum score to generate directives
--scorecard	off	Per-repository scorecard (M1-M5, Q1-Q5)

Examples

# Evaluate a single file with classification
eh evaluate myfile.py --classify

# Evaluate changed files in a PR
eh evaluate --diff origin/main...HEAD --classify

# JSON output for scripting
eh evaluate src/ --format json

# SARIF for GitHub Code Scanning
eh evaluate src/ --format sarif > results.sarif

# Strict mode for CI
eh evaluate src/ --strict --classify

# Rank files, highlight worst 10%
eh evaluate src/ --rank --bottom-pct 10

Exit codes

Code	Meaning
0	All files accepted (or marginal in lenient mode)
1	At least one file marginal (default mode)
2	At least one file rejected