Reverse engineer any repo into it's original prompt
Reverse engineer any repo into it's original prompt
https://github.com/user-attachments/assets/f0cdb7b2-c6f0-4483-8a01-153170479f2e
Turn a public GitHub repository into a single synthetic user prompt that someone might paste into Cursor, Claude Code, Codex, etc. to vibe code the project from scratch.
The app pulls repo metadata, a root file tree (depth 1), and the README, then uses an LLM to produce one short, conversational prompt grounded in that context.
Paste a GitHub URL or owner/repo on the home page. You can also open /owner/repo (e.g. /vercel/next.js) for a shareable link that runs the same flow.
GitHub-style /owner/repo/tree/... URLs on this site redirect to /owner/repo so they do not 404. The reverse flow still uses the whole repo for now; subfolder-aware context (scoped to that path) is planned for a later change.
Next.js 16 (App Router), React 19, TypeScript, Tailwind CSS 4, GitHub API, Supabase (optional), Stripe (optional).
Copy .env.example to .env.local and fill in at least one LLM API key.
The quick reverse endpoint supports five providers. Set GITREVERSE_QUICK_LLM to pin one, or leave it unset (auto) to let the app use whichever key it finds first:
| Provider | Key env var | Model env var | Default model |
|---|---|---|---|
| Grok (xAI) | XAI_API_KEY |
XAI_MODEL |
grok-3 |
| OpenRouter | OPENROUTER_API_KEY |
OPENROUTER_MODEL |
google/gemini-2.5-pro |
| Azure OpenAI | AZURE_OPENAI_API_KEY + AZURE_OPENAI_BASE_URL |
AZURE_OPENAI_MODEL |
gpt-5.4 |
| Google AI Studio | GOOGLE_GENERATIVE_AI_API_KEY |
GOOGLE_AI_STUDIO_MODEL |
gemini-2.5-pro |
| ApiSmart | APISMART_API_KEY |
APISMART_MODEL |
DEEPSEEK_V4_FLASH |
In auto mode the order of preference is: Grok → OpenRouter → Azure → Google → ApiSmart.
ApiSmart (sponsored) is an OpenAI-compatible gateway. Copy the exact Model ID from the ApiSmart models page (IDs differ from display names). Pin it with GITREVERSE_QUICK_LLM=apismart.
Azure quick reverse uses gpt-5.4 by default with AZURE_OPENAI_REASONING_EFFORT=medium. Title generation also uses Azure and defaults to gpt-5.4-mini with reasoning disabled.
Prompt-cache embeddings and /library hybrid search now use Azure embeddings first:
AZURE_OPENAI_EMBEDDING_MODEL — defaults to text-embedding-3-smallAZURE_OPENAI_EMBEDDING_DIMENSIONS — defaults to 512AZURE_OPENAI_EMBEDDING_DEPLOYMENT — optional explicit deployment nameAZURE_OPENAI_DEPLOYMENT_NAME_MAP — optional model=deployment mapping when deployment names differ from model idsIf your Azure resource does not yet have an embedding deployment, you can temporarily set OPENAI_API_KEY as an embedding-only fallback while quick reverse and title generation stay on Azure.
GITHUB_TOKEN — optional; increases GitHub API rate limits.SUPABASE_URL + SUPABASE_PUBLISHABLE_KEY — optional; enables server-side caching of quick prompts in prompt_cache and exposes the /library page.VIEWS_IP_SALT — required in production. Generate one with openssl rand -hex 32. The app will refuse to start in production without a non-default value.| Path | Description |
|---|---|
/ |
Home — quick reverse (codebase or website) |
/library |
Browse cached quick prompts (requires Supabase) |
/history |
Your prompt history (profile menu when signed in; syncs to Supabase) |
/[owner]/[repo] |
Shareable quick-reverse link |
/[owner]/[repo]/tree/... |
Redirects to /[owner]/[repo] |
pnpm install
pnpm devOpen http://localhost:3000.
pnpm build
pnpm start
pnpm lintNo open issues yet, or sync has not completed.