River Review docs (English)

日本語版はこちら—Japanese is the default and source of truth.

River Review documentation follows the Diátaxis documentation framework. English editions use the same filename with a .en.md suffix and may lag behind the Japanese originals.

We group docs into four types:

• Tutorials—step-by-step lessons for new users
• Guides—recipes for specific tasks
• Reference—technical specifications, APIs, and schemas
• Explanation—background, design decisions, and concepts

Docs live under pages/ and are served at /docs. Diátaxis is expressed via directories; language is expressed via filenames:

• tutorials/getting-started.md (ja) / tutorials/getting-started.en.md (en)
• guides/quickstart.md / guides/quickstart.en.md
• reference/skill-schema-reference.md / reference/skill-schema-reference.en.md
• explanation/riverbed-memory.md / explanation/riverbed-memory.en.md

Understand the concept

• Concept: turning review into an organizational judgment asset — the problems, the core model, the review targets, the responsibility boundary, and the non-goals.
• Welcome to River Review — a short introduction for first-time readers.
• What is River Review — a product overview covering features, usage, and the execution model.

Get started

• Getting started with River Review
• Quickstart

Advanced usage

• W-check (double review) — feed review results from other AI or human reviewers back in for re-verification, and check whether each finding is real.
• Repo-wide review — how to adopt and tune a review that reads repository context around the changed files, not just the PR diff.
• Cost estimation and optimization — estimating monthly cost with --estimate and --max-cost, then validating the estimate against measured usage.
• Agent workflow (--reviewers auto) — choosing between the entry points that call River Review from an AI agent (CLI, sub-agent, /review-local).
• Independent review synthesis — the synthesis pattern that merges several AI and human review results to support a merge decision.