What it does
RepoLens is an evidence-first review workspace for JavaScript and TypeScript repositories. It helps reviewers decide where to look first without executing repository code.
Import a public GitHub repository, choose a local ZIP, or use the bundled demo. RepoLens selects eligible source files, then creates a deterministic review queue from static, AST-derived evidence. Reviewers can inspect every finding's measurement and highlighted source range, select functions for focused metrics, and export a Markdown review brief.
Optional Deep Analysis uses the OpenAI API to generate source-grounded review hypotheses, questions, failure modes, test ideas, and simplification opportunities. AI output never changes the static score and is not presented as a verdict.
Why it matters
The first pass of code review is expensive: reviewers must follow imports, assess complexity, and separate meaningful risk from noise before they know where to focus. RepoLens makes that pass legible by preserving the evidence behind every signal.
How Codex and GPT-5.6 accelerated the work
I used Codex with GPT-5.6 throughout OpenAI Build Week as a hands-on engineering collaborator. It accelerated exploration, implementation, debugging, tests, deployment, and documentation, while I directed and reviewed the final product decisions.
Key decisions accelerated with Codex:
- Keep deterministic AST-derived evidence separate from optional AI hypotheses, so no AI result is treated as a verdict.
- Build the review flow around a queue, source evidence, function detail, and review desk.
- Filter local ZIPs and public GitHub repositories to eligible source before analysis, excluding dependencies, build output, generated files, lockfiles, secrets, and private keys.
- Replace whole-archive transport with source-only intake so the product is governed by its actual 100-file / 1M-character review limits.
- Keep Deep Analysis server-side, source-grounded, structured, and rate-limited.
Try it
Open the live demo, choose Try the demo, paste a public GitHub URL, or select a local ZIP. Static analysis works immediately. Deep Analysis is enabled in the deployed demo.
For local development: Node.js 20+, npm install, copy .env.example to .env.local, and run npm run dev. Add a server-only OPENAI_API_KEY only if you want optional Deep Analysis locally.
Built With
- codex
- github-rest-api
- gpt-5.6
- jszip
- next.js
- openai-api
- react
- typescript
- typescript-compiler-api
