V

VerifyAI AfriBa Labs

AfriBa Labs • Deterministic PR Verification Platform

Verify pull requests with mathematical certainty

VerifyAI tests code changes against structured contractual requirements in isolated sandboxes. Zero probabilistic guesses, zero flaky hallucinations — backed by tamper-evident cryptographic evidence.

VerifyAI Cockpit — afriba-labs/checkout-service #101
● Live Connected

Quality Gate: PASSED • Verdict: VERIFIED

All 6 contractual requirements verified across 4 isolated container sandboxes. Zero regressions.

🟢 MERGEABLE
CODE RISK SCORE
14.2 / 100 LOW
PREDICTIVE RISK
14% defect-rf-1.0
INVARIANTS TESTED
6 of 6 100%
EVIDENCE DIGEST
sha256:d8a9cafe...

How VerifyAI Works

A deterministic pipeline engineered from first principles for mission-critical software teams.

1

Connect GitHub Repository

Install the VerifyAI GitHub App with scoped permissions. Automatically maps code architecture, routes, database models, and service dependencies.

2

Verify Invariants & Contracts

On every pull request, VerifyAI extracts AST diffs, derives contractual requirements from the catalog, and executes checks in isolated process sandboxes.

3

Audit Cryptographic Evidence

Receive a definitive verdict (VERIFIED or FAILED) with SHA-256 evidence digests, alongside an isolated DefectPredict risk analysis to prioritize testing.

Engineered for Absolute Verification

Deterministic verification answers whether requirements are satisfied. Predictive risk estimates where defects are likely. Technologically separate.

Change Impact Graph

Calculates exact graph reachability, fan-in, fan-out, and blast radius from AST unified diffs across routes and models.

49-Requirement Catalog & DSL

Pre-built contract and invariant templates covering idempotency, authorization, schemas, state-transitions, and rate limits.

Isolated Process Sandbox

Executes test harnesses in stripped, scratch-isolated subprocesses with environment secret scrubbing and strict timeout enforcement.

Tamper-Evident Evidence Vault

Generates cryptographic SHA-256 digests for all test outputs, HTTP traces, and SQL queries. Verifiable, immutable audit trail.

DefectPredict Risk Layer

Pure TypeScript embedded random forest ensemble (defect-rf-1.0.0) prioritizes test queues without ever altering verdicts.

Flaky & Pre-Existing Attribution

Runs dual base-SHA vs. head-SHA executions. True regressions are isolated from pre-existing failures and transient environmental flakiness.

Enterprise Security & Execution Isolation

VerifyAI is designed to run in zero-trust environments. Your source code and credentials remain protected at all times.

Zero Source Code Retention

Code diffs and AST representations are processed in ephemeral memory. Source code is never retained or stored.

Server-Side Tokens Only

GitHub tokens and secrets are stored exclusively server-side and never exposed to the frontend or browser.

Automated Secret Redaction

All captured stdout logs, HTTP headers, and queries pass through an automated redactor scrubbing API keys and credentials.

Zero External AI/ML APIs

Verification and predictive risk models run 100% locally in embedded TypeScript. No source code is ever sent to external LLMs.

Documentation & Quickstart

Get up and running with deterministic verification in under two minutes.

1. Install GitHub App & Webhook

Authorize VerifyAI to receive pull request event deliveries. VerifyAI will inspect proposed unified diffs on every PR push.

2. Define Contract Invariants (Optional)

VerifyAI automatically infers applicable requirements from your AST diff. You can explicitly bind catalog templates in .verifyai.yml:

version: "1.0"
requirements:
  - template: ACC-001 # Strict Authentication Token
  - template: SEC-003 # Idempotency Key Validation

3. Review Deterministic Verdict

Check results directly in your PR checks or in the VerifyAI dashboard with cryptographic evidence hashes.

About AfriBa Labs

Building high-assurance verification tools for mission-critical software.

AfriBa Labs pioneers zero-LLM deterministic software verification. We reject probabilistic opinions where formal truth and empirical evidence are required. VerifyAI guarantees that critical contracts and invariants are never violated without detection.