GitSage
Engine Report

Confidence Score

Every GitSage suggestion includes a Confidence Score — a trust metric that ensures your Git history remains pristine and accurate.

Confidence Animator (Live Demo)

Scroll into view to see the bar animate from 0 to 85%:

AI Confidence Score
feat(auth): add JWT token expiry validation
0%
[░░░░░░░░░░░░░░░░░░░░]
0%High Confidence ≥85%100%

Score Tiers

85–100%High Confidence
[██████████████████░░]

GitSage auto-commits if auto_commit: true. Otherwise, shows the suggestion with 1-click approve.

50–84%Medium Confidence
[████████████░░░░░░░░]

Always prompts for review. You can approve, edit inline, or regenerate.

25–49%Low Confidence
[███████░░░░░░░░░░░░░]

GitSage warns you and requests contextual input (e.g. --context flag) before proceeding.

0–24%Very Low Confidence
[███░░░░░░░░░░░░░░░░░]

GitSage declines to auto-suggest. You are prompted to write the message manually.

What Affects the Score?

Diff clarityLarger, well-named function changes score higher than cryptic binary patches
Semantic signalsMeaningful variable names, docstrings, and comments improve accuracy
Scope detectionSingle-concern commits with clear module boundaries score higher
Model certaintyThe AI's own token probabilities are factored into the score

GitSage Docs

View on GitHub