ai-profile-cli 0.4.1

Evidence-backed, privacy-safe AI collaboration cards and dashboards generated locally from explicit Git provenance.

ai-profile

English · 繁體中文

Show how you collaborate with AI—not just how often you commit. aiprofile turns explicit AI provenance in your local Git history into evidence-backed, privacy-safe SVG cards and a self-contained interactive dashboard.

Use it when you want a credible AI portfolio without uploading source code, guessing attribution from coding style, or exposing repository identities. It recognizes AI-* commit trailers and verified AI co-author identities, aggregates activity across repositories, and keeps its database on your machine.

It is not an AI code detector. Commits without explicit evidence remain unknown; they are never guessed to be human or assigned to a provider.

• Credible: every AI attribution comes from explicit Git evidence.
• Private by default: repository identities stay hidden, and daily dates appear only when you explicitly publish them.
• Profile-ready: one local workflow produces light and dark summary, heatmap, and badge assets plus an interactive provider view.

Preview

These samples use synthetic data.

Install

Requires Python 3.11–3.14 and Git 2.17 or newer. Windows, macOS, and Linux are supported. Git repositories must use the SHA-1 object format.

pip install ai-profile-cli

The package name is ai-profile-cli; the installed command is aiprofile.

Quickstart

Run these commands from the root of one of your repositories:

aiprofile init
aiprofile scan .
aiprofile aggregate
aiprofile render

• init creates the local configuration and seeds your identity from git config user.email.
• scan records commits authored by your configured identities.
• aggregate shows the exact privacy-safe statistics that can be published.
• render writes six SVG files, dashboard.html, and profile.json to dist/.

Only commits authored by identities in ~/.aiprofile/config.json count. Review the email seeded by init and add any other addresses you commit with.

Scan additional repositories at any time:

aiprofile scan /path/to/another/repository

Repositories default to aggregate_only: their counts contribute to the summary, but their identity and daily dates stay private. --full marks that repository's aggregate activity as explicitly publishable so its dates can appear in date-based views. This is a policy choice, not a claim about the repository's GitHub visibility:

aiprofile scan --full /path/to/repository

That choice persists. Scanning the same repository later without --full does not downgrade it. To stop publishing its dates, change its publication_level in ~/.aiprofile/config.json to aggregate_only (keep identity-redacted aggregate totals) or excluded (omit it), then run aggregate and render again.

Only run one render at a time for a given output directory.

Add the cards to your GitHub profile

Render inside your GitHub profile repository (USERNAME/USERNAME), then commit the generated dist/ directory. Add this to its README:

<picture>
  <source media="(prefers-color-scheme: dark)" srcset="dist/summary-dark.svg">
  <img alt="AI collaboration summary" src="dist/summary-light.svg">
</picture>

Use heatmap-{light,dark}.svg or badge-{light,dark}.svg in the same pattern. To refresh your cards, run scan again, run render, and commit the updated files.

Interactive dashboard

dist/dashboard.html is a self-contained dashboard generated from the same validated aggregate data as the SVG cards. Open it directly in a browser to:

• switch between All AI and one provider;
• inspect provider-attributed commits, actor presences, and active days without mixing those units;
• inspect the publishable daily collaboration window;
• use light, dark, or system color themes.

The file loads no external script, font, tracker, or API. It contains the same public aggregate fields as profile.json, so review aiprofile aggregate before publishing either file.

GitHub READMEs do not execute JavaScript. Keep the SVGs in your Profile README and link them to dashboard.html on GitHub Pages or another static host. See the live profile example.

Declare AI participation

Known AI co-author identities are recognized automatically. For other tools, or when you want richer attribution, add an AI-* trailer block to the commit message:

feat: add aggregation service

AI-Provider: Anthropic
AI-Model: Claude-Sonnet
AI-Tool: Claude-Code
AI-Role: implementation, documentation
AI-Mode: AI-Assisted
AI-Reviewed-By: Human

Repeat the AI-Provider: key to declare another AI actor in the same commit. One commit can therefore equal one unique AI-attributed commit and multiple actor presences. Use AI-Mode: Human-Only only for an explicitly human-only commit.

Privacy

• No repository data is uploaded. The CLI makes no network calls and sends no telemetry.
• New repositories default to aggregate_only. Use scan --full as an explicit publication decision, or set a repository to excluded in ~/.aiprofile/config.json.
• Public assets contain aggregate counts, public provider names, evidence totals, and dates. They never contain repository names or paths, organization names, branches, commit SHAs or messages, raw trailers, or email addresses.
• aiprofile aggregate is the publication preview. Review it before committing generated assets.
• Aggregate-only output is identity redaction, not anonymity. Repeated publication can reveal when totals changed. See the full privacy model.
• Do not publish or sync ~/.aiprofile; it contains private repository paths, identities, and a local salt.

Metrics

• AI-attributed commits are unique commits with at least one explicit AI actor presence.
• Actor presences count distinct provider/tool participation within a commit. Provider totals can therefore exceed unique commit totals.
• Unknown stays separate from human.
• Evidence quality is reported as verified > declared > imported > inferred > unknown.

Help and support

Run aiprofile --help or aiprofile COMMAND --help for command options. Use GitHub Issues for bugs and feature requests. Report vulnerabilities through the security policy.

Contributing and license

Contributions are welcome; see CONTRIBUTING.md. Licensed under the MIT License.