Question 1

Is my code or data sent to any third party?

Accepted Answer

No. All requests are processed on dedicated Apple Silicon hardware we own and operate in Austin, TX — there's no routing through OpenAI, Anthropic, or any other third-party inference provider, and no telemetry collection. Every response includes two headers you can verify directly:

- x-ro2-data-residency: on-prem-austin-tx
- x-ro2-third-party-routing: none

If you're sending proprietary code for review, it never leaves hardware we control.

Question 2

What's the difference between /v1/review and /v1/chat/completions?

Accepted Answer

`/v1/review` is the primary endpoint — send a unified diff (or raw files) and get back **structured, machine-readable findings** (`file`, `line`, `severity`, `issue`, `suggestion`) instead of prose, built for CI pipelines and PR bots that need to parse output programmatically. `/v1/chat/completions` is a general-purpose, OpenAI-compatible inference endpoint for anything beyond code review — same request/response shape as GPT-4o, drop-in with existing OpenAI SDKs.

Model access scales with plan:

| Plan | Model |
|---|---|
| Free | `qwen2.5-coder:32b` (base) |
| Basic | `crank-reviewer v5` (tuned for code review) |
| Pro | `crank-reviewer v5` (tuned, priority throughput) |
| Mega | `crank-reviewer v5` (tuned, dedicated B2B) |

Code Review API - AI Pull Request and Diff Reviewer

Frequently Asked Questions