SpecTrace AI

SpecTrace AI is an evidence intelligence platform for safer software releases. Engineering teams often make release decisions with fragmented evidence: application logs, screenshots, videos, QA notes, API traces, and customer-impact reports. SpecTrace brings this evidence into one traceable investigation and helps teams understand what happened, why it happened, what should be fixed, and whether the release should be approved, changed, or blocked. The demo focuses on a checkout regression where customers complete payment successfully but remain stuck on the processing screen without receiving an order confirmation. SpecTrace detects a payment response-contract mismatch: the frontend expects paymentIntentId, while the payment service returns transactionId. From that evidence, the system generates a root-cause hypothesis, evidence timeline, AI reviewer summary, engineering recommendation, release decision rationale, Playwright regression test draft, GitHub issue draft, and Given/When/Then acceptance criteria. SpecTrace uses a Next.js frontend, .NET API backend, Python FastAPI AI worker, Fireworks AI serverless enrichment, and deterministic fallback safety. The current live demo runs on Vercel and Render, with Fireworks AI used safely through serverless inference. The architecture is prepared for AMD Developer Cloud, ROCm, and Gemma-style inference when AMD GPU pod access is allocated. SpecTrace is not just an AI summary tool. It is a release intelligence workflow from evidence to engineering execution to human decision.