docs: open-source AI code review tools research

[Svtter]

Summary

• Surveyed 10+ open-source AI code review projects to evaluate whether any can replace opencode-actions' multi-agent parallel review
• Conclusion: no open-source project matches the multi-agent parallel review + coordinator synthesis architecture
• Key competitors evaluated: Kodus AI, PR-Agent, Tabby, CodeRabbit (SaaS), villesau/ai-codereviewer, and others

[github-actions]

✅ 可合并 / CAN MERGE

此 PR 仅新增了一份开源 AI 代码审查工具的研究文档，未修改任何代码逻辑。所有 reviewer 均无阻塞项，无冲突，多数 reviewer 无任何问题。

🟢 建议项 / Suggestions (2)

• 文件名拼写错误 — 文件名 alternative-review-tools-research.md 拼写有误，应为 alternative（缺少字母 'n'）。建议修正为 alternative-review-tools-research.md。（来源：quality）
• 转化为 ADR — 文档内容属于调研记录，建议后续将调研结论转化为架构决策记录（ADR）放在 docs/adr/ 目录下，以长期保留架构选型的决策依据。（来源：architecture）

💰 Review Cost — $0.0000

Role	Cost (USD)	Input	Output	Reasoning	Cache Read	Cache Write
quality	$0.0000	9,527	123	570	0	0
security	$0.0000	9,536	78	376	0	0
performance	$0.0000	9,539	77	125	0	0
architecture	$0.0000	9,540	128	188	0	0
regression-test	$0.0000	9,977	65	170	0	0
test-value	$0.0000	9,667	42	97	0	0
coordinator	$0.0000	8,946	194	372	0	0
Total	$0.0000	66,732	707	1,898	0	0

📋 各 Reviewer 详细审查结果

quality

可合并 / CAN MERGE

该 PR 添加了一份关于开源 AI 代码审查工具的研究文档，内容详实、结构清晰，格式规范。没有代码质量问题、逻辑错误或风格不一致的问题。

阻塞项：无

警告项：无

建议项：

• 文件名 alternative-review-tools-research.md 拼写有误，应为 alternative（缺少字母 'n'）。虽然不会影响功能，但建议修正为 alternative-review-tools-research.md。

security

安全无虞 / SECURE

此 PR 仅添加了一份纯研究文档 docs/alternative-review-tools-research.md，内容为开源 AI 代码审查工具的对比分析。不涉及任何代码逻辑、配置文件变更或 CI/CD 流程修改，不存在安全风险。

阻塞项：无

警告项：无

建议项：无

performance

性能良好 / GOOD

本 PR 仅新增了一份文档文件 docs/alternative-review-tools-research.md，内容为开源替代工具的市场调研报告。该文件不包含任何可执行代码、脚本、配置变更或业务逻辑修改，因此不存在任何性能相关的问题。

阻塞项：无

警告项：无

建议项：无

architecture

架构合理 / SOUND

本次 PR 仅新增了一个调研文档 docs/alternative-review-tools-research.md，未修改任何代码文件。该文档是对外部开源项目的对比分析，结论是维护现有架构（多智能体并行评审）而非替换。从架构角度看：

阻塞项：无

警告项：无

建议项：

1. 文档内容本身属于调研记录，无架构问题。建议后续将调研结论转化为 ADR（架构决策记录）放在 docs/adr/ 目录下，以长期保留架构选型的决策依据。

regression-test

无需回归测试

PR 类型：CHORE（纯文档变更）

此 PR 仅新增了一份调研文档 docs/alternative-review-tools-research.md，内容为开源代码审查工具的对比分析，不涉及任何代码逻辑修改、行为变更或 bug 修复。因此无需添加回归测试。

test-value

ALL TESTS HAVE VALUE

This PR contains no test files — the diff only adds a research document (docs/alternative-review-tools-research.md). 没有可审查的测试代码。

[Svitter]

review:p1 — Well-structured research document. The core conclusion (no open-source tool matches the multi-agent parallel review architecture) is well-supported and the comparison tables are useful. Two convention issues need attention before merge:

1. Missing CHANGELOG entry — docs/ additions still need a line under [Unreleased].
2. Language inconsistency — existing docs/plan/ files are in Chinese; this is in English. Needs a deliberate choice either way.

Minor nits on star-count dating and Tabby's categorization. The factual claims about competitor features are reasonable given the available data. Thanks for the thorough survey — this is useful context for the project's positioning.

Rank: P1 (mergeable after changelog + language call).

[Svitter]

should-fix — Language inconsistency. The other docs/plan/ documents (e.g. preload-review-context.md) are written in Chinese. This document is entirely in English. If the project's doc convention is Chinese, this should be rewritten. If English is intentional for an international-facing research artifact, that's fine — but it should be a deliberate choice, not an oversight. Which is it?

[Svitter]

nit — Star counts (1.2k, ~11k, 33k, ~1k, 366, 57, 39) are stale-sensitive. Consider adding a note like "as of 2026-06-24" to the table header, or converting stars to a less volatile metric. Not a blocker, but readers will want to know the snapshot date.

[Svitter]

nit — Tabby (33k stars) is primarily a code completion / IDE assistant, not a code review tool. Listing it under "Self-Hosted AI Review" is misleading. It's listed as ❌ Not review-first which is correct, but it probably belongs in a separate "adjacent tools" note rather than the Tier 1 table. A reader skimming the table might assume Tabby is a direct competitor.