Description
In real-world systems, the same application often behaves differently across environments. Local development rarely reflects production reality. As applications move through staging, beta, and production, they accumulate additional layers of security, networking, and policy enforcement. These differences frequently result in unexpected latency, failed requests, or production-only bugs that are difficult to diagnose quickly.
While developers can manually investigate these issues using browser tools, logs, and metrics, doing so requires fragmented context and prior knowledge of what to look for. Environment Drift Analyzer solves this by providing an automated AI agent that identifies why application behavior differs across environments and explains the root causes in digestible terms.
Features
- Designed an edge-native system to detect and explain behavioral drift between application environments using deterministic diffing combined with LLM-generated, schema-validated explanations.
- Architected a contract-first pipeline with normalized observable signals, idempotent workflow orchestration, and SQLite-backed Durable Objects for persistent, bounded historical memory across deploys.
- Implemented manual redirect probing, header whitelisting, SSRF safeguards, and retry-safe workflows to ensure correctness under edge execution constraints and strict network policies.
Info
- live demo
- source code
- Cloudflare (Workers, Workflows, Durable Objects, Pages), SQLite, Llama 3.3, TypeScript, React, Zod
System Architecture
Workflow Sequence Diagram
