d1ff — State-of-the-art AI code review. Open source.

The only code review tool that saves money for you, not for itself.

d1ff uses smart model routing: frontier models for complex logic, lightweight models for simple changes. You pay only for the tokens used — no per-seat fees, no markup. The result: review quality on par with the best commercial tools at a fraction of the cost.

Demo GIF coming soon — see releases for preview screenshots

---

Why d1ff?

State-of-the-art review quality

d1ff doesn't cut corners. Every PR goes through a 3-pass pipeline — Architecture, Logic, Style — using the same frontier models (GPT-5.4, Claude Opus 4.6, Gemini 3.1 Pro) that power the most expensive commercial tools.

Smart Model Routing

Not every diff needs a $0.10 frontier-model call. d1ff analyzes each change and automatically picks the right model:

Change complexity	Model tier	Cost per review
Simple (typos, formatting, renames)	gpt-5.4-nano, gemini-3.1-flash-lite	~$0.002
Medium (new functions, refactors)	gpt-5.4-mini, claude-sonnet-4-6	~$0.01
Complex (architecture, security)	gpt-5.4, claude-opus-4-6	~$0.05

Average cost across a typical team: ~$0.02 per review.

Open source. Full control.

• MIT licensed — fork it, modify it, deploy it commercially
• Self-hosted — your code never leaves your infrastructure
• BYOK (Bring Your Own Key) — use your own API keys, no middleman
• No vendor lock-in — swap providers anytime, no data held hostage

We save money for you, not for ourselves

Traditional SaaS code review tools charge per seat — they profit when your team grows, regardless of how much you actually use the tool. d1ff flips the model: you pay only for the LLM tokens consumed. We have zero incentive to waste your tokens.

Team Size	CodeRabbit Pro (6 mo)	d1ff BYOK (6 mo, ~200 PRs/mo)	6-Month Savings
5 devs	$900	~$24 in tokens	~$876
10 devs	$1,800	~$48 in tokens	~$1,752
20 devs	$3,600	~$96 in tokens	~$3,504

CodeRabbit Pro: $30/dev/mo. d1ff estimate based on ~$0.02/review with smart model routing. Actual cost depends on PR size and complexity.

---

Quickstart (2 minutes)

Hosted (GitHub App)

1. Sign up at app.d1ff.dev with your GitHub account
2. Add your LLM API key in the d1ff dashboard
3. Open a pull request — d1ff will post a review automatically

Self-Hosted (Docker)

cp .env.example .env
# Fill in your values in .env
docker run -d \
  --env-file .env \
  -p 8000:8000 \
  ghcr.io/d1ff-dev/d1ff:latest

See docs/self-hosting.md for full configuration options.

---

Features

• 3-Pass Review — Architecture → Logic → Style, each pass focused and noise-free
• Smart Model Routing — Frontier models for hard problems, fast models for simple ones
• Suggestion Blocks — Inline code suggestions you can apply with one click
• Slash Commands — /review, /explain, /suggest directly in PR comments
• BYOK — Use your own OpenAI, Anthropic, Google, or DeepSeek API key
• Self-Hosted — Full control: your data stays in your infrastructure

---

Supported LLM Providers

Provider	Example Models
OpenAI	gpt-5.4, gpt-5.4-mini, gpt-5.4-nano
Anthropic	claude-opus-4-6, claude-sonnet-4-6
Google	gemini-3.1-pro, gemini-3.1-flash-lite
DeepSeek	deepseek-v3, deepseek-r1
Custom	Any OpenAI-compatible endpoint via LLM_API_BASE

---

Self-Hosting

Full self-hosting guide: docs/self-hosting.md

---

Contributing

Contributions are welcome! See CONTRIBUTING.md for guidelines.

---

License

MIT — free to use, fork, and deploy commercially.