Getting Started¶

Install¶

pip install aipr-py

The distribution on PyPI is aipr-py. From source:

pip install git+https://github.com/yunaremaia/aipr.git

Requires Python 3.10+.

As a GitHub CLI extension:

gh extension install yunaremaia/aipr

After install, gh aipr OWNER/REPO works identically and exit codes are preserved for CI conditionals.

Verify:

aipr --version

Authentication¶

Policy fetches go through the GitHub API. Anonymous calls rate-limit fast, so set a token if you are scanning more than a couple of repositories:

export GH_TOKEN=ghp_xxx   # classic token with public repo read access

GITHUB_TOKEN is accepted as a fallback. Without a token the tool still works, it just makes more unauthenticated requests.

First Scan¶

aipr asciimoo/hister

Output:

aipr: asciimoo/hister
[BLOCKED] human-only policy
sources: CONTRIBUTING.md, README.md
confidence: 1.0  score: 19.0
autonomous contribution: NOT SAFE - require human co-authorship

A repository with no discoverable policy:

aipr: apache/maka
[UNKNOWN] no explicit AI policy found

and exits 2.

Reading the Output¶

sources - the files the verdict was drawn from.
confidence / score - how strongly the matched phrases agree.
The verdict line (human_only, restrictive, disclose_ok, permissive, unknown) is the classification itself.

Explain the Verdict¶

When you disagree with the classification, ask for the evidence:

aipr --explain asciimoo/hister

autonomous contribution: NOT SAFE - require human co-authorship
  [+5.0] ...Issues and PR descriptions must be fully human-written...
  [+5.0] ...AI should never be the main author of the PR...

Each line is a phrase that matched, with the weight it contributed.

Classify a Local File¶

To check a policy file before you commit it, or one you already have on disk:

aipr --text AI_POLICY.md

--text reads the file directly and makes no network calls.

Gate a Contribution¶

Because the verdict is in the exit code, a pre-submission check is one line:

aipr "$GITHUB_REPOSITORY" || echo "check the AI policy before contributing"

exit 0 - autonomous-safe
exit 1 - restricted or human-only, do not send a bot
exit 2 - unknown, read the policy yourself

See Usage for batch mode, SARIF output and CI integration.