Codex Ops is a full-stack dashboard for orchestrating, monitoring, and reviewing AI coding-agent runs on GitHub repositories. It models a LangGraph-style operator workflow: clone and index a repo, retrieve context (RAG), plan edits, generate patches, validate changes, review diffs, and open a pull request.
Key features: • Live AI agent pipeline powered by OpenAI (gpt-4o-mini) with server-side API keys and a deterministic mock fallback when no key is set • End-to-end flow: create a task → persisted run → animated live pipeline view → plan, patches, validations, and PR simulation • Command Center dashboard and Analytics driven by real persisted runs (with sample data fallback) • Run history with search, filters, sorting, diff download (.patch), delete, and live refresh • Toast notifications, light/dark theme, and a polished operator UI built with React 19, TanStack Start/Router/Query, and shadcn/ui
Tech stack: TypeScript, TanStack Start (SSR + server functions), Cloudflare Workers/D1, Vite 7, Tailwind CSS 4, OpenAI Chat Completions API.
Built by Muhammad Zeeshan as a portfolio-grade demonstration of production-minded AI agent UX, secure env handling, and full-stack patterns on the edge.