Skip to content

v0.7.0 — first PyPI release

Choose a tag to compare

View all tags
@kmadan kmadan released this 14 May 07:56
· 76 commits to main since this release
v0.7.0
f3eb042
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

The first AICertify release on PyPI.

pip install aicertify

⭐ If AICertify is useful to you, please star the repo — it helps AI governance and policy-as-code practitioners discover the project.

What AICertify does

Compliance-as-code for AI systems: evaluate AI applications against named regulatory frameworks (EU AI Act, NIST AI RMF, FERPA/COPPA, FAA/EASA aviation rules, fair lending, and more) using executable OPA/Rego policies from the gopal policy library. Produces audit-ready reports in PDF, Markdown, JSON, or HTML.

Quick start

pip install aicertify
git clone https://github.com/Principled-Evolution/aicertify.git
cd aicertify
python examples/quickstart.py

Or fork one of the three reference examples:

What's in this release

Core framework

  • Reporting subsystem producing audit-ready PDF / Markdown / JSON / HTML
  • Pluggable Python evaluators: FairnessEvaluator, ContentSafetyEvaluator, RiskManagementEvaluator, ComplianceEvaluator
  • OPA policy integration via the standalone gopal library (94 production Rego policies, 15+ frameworks, 5 industry verticals)
  • CLI (aicertify --help) — registered as a console-script entry point
  • Quickstart at examples/quickstart.py

Developer experience

  • AGENTS.md + CLAUDE.md — operational guides for AI coding agents (Claude Code, Cursor, Codex, Gemini CLI, Copilot)
  • Claude Code skills under skills/: /run-compliance-check, /evaluate-contract, /explain-regulation, /draft-policy
  • Three forkable application examples (each shipping README.md, input_contract.json, sample_interactions.json, policy_config.yaml, run.py, expected_report.md)
  • 10 starter contributor issues under good first issue and help wanted
  • SECURITY.md with a private vulnerability-disclosure flow
  • docs/why-aicertify.md — long-form positioning doc
  • docs/demo-report-eu-ai-act.pdf — view the deliverable without installing

Discoverability

  • README rewritten as a product page: centered hero, value-prop bullets, comparison table vs Fairlearn / AIF360 / MS RAI / Credo AI
  • 15 PyPI keywords: ai-governance, eu-ai-act, nist-ai-rmf, responsible-ai, policy-as-code, opa, rego, …
  • 15 GitHub topics mirroring the PyPI keywords
  • Translated READMEs for Simplified Chinese, Japanese, Korean, and Hindi

Security

This release clears ~95 of the ~96 Dependabot advisories that were open against the development line. Direct dep bumps include transformers >=4.53, huggingface-hub >=0.34, requests >=2.33, python-dotenv >=1.2.2, markdown >=3.8.1, black >=26.3.1, pytest >=9.0.3, pytest-asyncio >=1.0, fastapi >=0.119, starlette >=0.49.1. Explicit transitive lower bounds for aiohttp, urllib3, pillow, pypdf, nltk, langchain-core, langchain, langchain-openai, langchain-text-splitters, langsmith, pyasn1, protobuf, banks, sentencepiece, orjson, brotli, marshmallow, filelock, virtualenv, Pygments. CodeQL: 7 logging false-positives dismissed with rationale; 2 workflow alerts fixed by adding explicit permissions: contents: read. See CHANGELOG.md for the per-package list.

Remaining: 1 medium advisory (transformers <5.0.0rc3) — upstream-blocked behind a release candidate.

What's not yet in this release

  • An MCP server for Claude Code / Cursor — on the roadmap
  • A first-class aicertify install command that auto-registers skills + MCP + editor hooks — on the roadmap
  • Hiring as a dedicated gopal framework — currently using bfs/fair_lending as a proxy

Honest scope

A green AICertify report is necessary but not sufficient for production deployment of any AI system in a regulated context. AICertify produces the evidence; clinical evaluation, regulator notification, fundamental-rights impact assessments, and engagement with legal counsel remain mandatory. See docs/why-aicertify.md for the full positioning.

Thanks

Built by Principled Evolution. Policies you can read, run, and prove.

Looking for contributors: see the pinned umbrella issue and the good first issue label.

Assets 4
Loading