assay-ai 1.2.0

Receipt-native AI safety toolkit

Assay

Stop filling out the AI compliance spreadsheet. Add one CI step. Every merge produces portable, signed evidence that anyone can verify.

When someone asks "prove what your AI did," you need more than logs. Logs live on your infrastructure, under your control. Assay produces signed evidence bundles they can verify independently -- no access to your systems required.

Install

pip install assay-ai

2-Minute Demo

pip install assay-ai
assay demo-incident

This runs a two-act scenario with synthetic data (no API key needed):

• Act 1: Agent uses gpt-4 with a guardian check. Result: integrity PASS, claims PASS.
• Act 2: Someone swaps the model and drops the guardian. Result: integrity PASS, claims FAIL.

That second result is an honest failure -- authentic evidence proving the run violated its declared standards. Not a cover-up. Not theater. Exit code 1.

assay demo-challenge    # CTF-style: spot the tampered pack
assay demo-pack         # build + verify from scratch

The Golden Path

# 1. Find uninstrumented LLM calls
assay scan .

# 2. Instrument (one line)
#    from assay.integrations.openai import patch; patch()

# 3. Produce a signed proof pack
assay run -c receipt_completeness -c guardian_enforcement -- python my_app.py

# 4. Verify + explain
assay verify-pack ./proof_pack_*/
assay explain ./proof_pack_*/

# 5. Lock the verification contract
assay lock write --cards receipt_completeness,guardian_enforcement -o assay.lock
assay verify-pack ./proof_pack_*/ --lock assay.lock --require-claim-pass

How It Works

Assay separates two questions on purpose:

• Integrity: "Were these bytes tampered with after creation?" (signatures, hashes, required files)
• Claims: "Does this evidence satisfy our declared governance checks?" (receipt types, counts, field values)

Integrity	Claims	Exit Code	Meaning
PASS	PASS	0	Evidence checks out, behavior meets standards
PASS	FAIL	1	Honest failure: authentic evidence of standards violation
FAIL	--	2	Evidence has been tampered with

The split is the point. Systems that can prove they failed honestly are more trustworthy than systems that always claim to pass.

Commands

Command	Purpose
assay demo-pack	Generate demo packs (no config needed)
assay demo-incident	Two-act scenario: passing run vs failing run
assay demo-challenge	CTF-style good + tampered pack pair
assay scan	Find uninstrumented LLM call sites
assay run	Wrap command, collect receipts, build signed pack
assay verify-pack	Verify a Proof Pack (integrity + claims)
assay explain	Plain-English summary of a proof pack
assay lock write	Freeze verification contract to lockfile
assay lock check	Validate lockfile against current card definitions
assay doctor	Preflight check: is Assay ready here?

Documentation

• Quickstart -- install, golden path, command reference
• Deep Dive -- architecture, trust model, honest failure
• Decision Log -- every locked decision and why

Related Repos

Repo	Purpose
assay	Core CLI, SDK, conformance corpus (this repo)
assay-verify-action	GitHub Action for CI verification
assay-ledger	Public transparency ledger

License

Apache-2.0