MemReview FINAL with Lahari Recording

MemReview: an AI code reviewer that remembers your team's decisions. Most AI code reviewers treat every pull request like the first one. They don't know your team's rules, so reviewers end up typing the same comments again and again. MemReview fixes that using Hindsight memory. In this demo: Without memory: generic feedback that ignores the team's decisions With memory: the agent recalls past PR feedback (for example, avoid inline CSS and use Tailwind instead, established in PR #42) and cites the earlier decision Reject and teach: when a reviewer rejects an AI comment, we save the correction as a new team memory that improves future reviews How it works: 1. A pull request diff is sent to our review endpoint 2. RECALL: we fetch relevant team memories from Hindsight 3. The Groq LLM (openai/gpt-oss-120b) reviews the code using that memory as context 4. RETAIN: reviewer feedback (accept/reject) is saved back into Hindsight 5. Repeated mistakes become team rules the agent applies automatically Tech stack: Python, FastAPI, Hindsight (retain/recall), Groq LLM GitHub repo: https://github.com/Lahari61111/pr-memory-r... Built with Hindsight: https://github.com/vectorize-io/hindsight Team: Lahari, Yasasri, Hansika, Jyothi #AI #CodeReview #Hindsight #FastAPI #Groq