AgentAssert Documentation Portal
Complete reference manuals, Python API definitions, YAML ContractSpec schemas, and mathematical proofs maintained in the official GitHub repository.
Agent Reliability Engineering: From Prompt Hope to Formal Bounds
An end-to-end video walkthrough by author Varun Pratap Bhardwaj explaining:
- Why uncontracted agents suffer compounding drift under long multi-turn sessions.
- How the 4-tuple $C = (P, I, G, R)$ evaluates in <10ms inline without blocking token streams.
- The 18,000 mission study and why paired agents fail together 90.0% of the time.
- How to generate anytime-valid compliance certificates for enterprise AI governance.
Core Architecture & Specification
Formal ContractSpec Grammar
Formal EBNF grammar, AST structure, and YAML schema specification for $C = (P, I, G, R)$.
Python Runtime API Reference
Class definitions and docstrings for Contract, Monitor, PolicyEvaluator, and TelemetryCollector.
Mathematical Metrics & Theorems
Formulations for Lyapunov drift $D(t)$, recovery rate $\gamma$, and non-parametric moment certificates.
Integrations & Practical Guides
Framework Interceptors Guide
Detailed middleware hooks for LangGraph, CrewAI, Microsoft AutoGen, and OpenAI Assistants SDK.
Runnable Code Examples
Production-ready scripts demonstrating PII redaction, spend caps, and automated escalation.
Domain Contracts Catalog
YAML templates across Financial Advisory, Clinical Triage, Code Gen, Legal, and Multi-Agent pipelines.
Empirical Benchmarks & Datasets
AgentContract-Bench v1 Spec
Experimental protocol, 200 evaluation scenarios, deterministic test oracles, and statistical methodology.
18,000 Preregistered Missions Dataset
Open dataset and statistical analysis replication scripts for multi-agent co-failure study.