This Open-Source Project Can Review Your Code Better Than Fable (Open Code Review)
Read the Portuguese translation →
Author: Better Stack · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: This Open-Source Project Can Review Your Code Better Than Fable (Open Code Review) · Channel: Better Stack · 01:34 · Watch on YouTube
Topics: AI agents, Software engineering with AI, AI developer tools, Open source and software ecosystems
Summary
Section titled “Summary”Open Code Review is an open-source code review project designed to address common LLM review problems, including missed files, inconsistent results, and inaccurately located issues. It uses a deterministic pipeline with precision file selection, related-file bundling, and fine-grained row matching. The project also uses code-review-focused prompt templates and a specialized set of tools, while relying on agents for dynamic decisions. Users can install its CLI to review code or scan a repository, and it can also run in CI/CD to help catch bugs before production.
Key points
Section titled “Key points”- The video says traditional LLM code reviews can miss files, vary between runs, or report issues in the wrong files.
- Open Code Review uses a deterministic pipeline intended to produce consistent results across runs and codebase sizes.
- Precision file selection identifies which files need review.
- Smart file bundling groups related files into a single review unit.
- Fine-grained row matching focuses the model’s attention and reduces information noise.
- The project uses prompt templates optimized for code review and specialized tools described as more predictable than generic tools.
- Users can install the CLI to review code or scan an entire repository.
- The presenter says the project can run in CI/CD to catch bugs before they reach production.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | Problems with LLM code review | The video describes missed files, inconsistent review quality, and issues reported in the wrong locations. |
| 00:31 | Open Code Review’s deterministic pipeline | The project uses file selection, bundling, and row matching to structure its code reviews. |
| 01:02 | Tools, CLI, and repository scans | The video introduces the review-focused prompts and tools, explains how to use the CLI, and shares the presenter’s experience. |
| 01:34 | Using code review in CI/CD | The project can run in CI/CD to help catch bugs before they reach production, followed by a channel subscription prompt. |
Quotes
Section titled “Quotes”incomplete coverage — 00:00
unstable quality — 00:00
smart file bundling — 00:31
just install the CLI — 01:02
catch bugs at their source before they make it into production — 01:34
Concepts and entities
Section titled “Concepts and entities”Open Code Review, Fable, LLM, Opus, CLI, CI/CD, Claude Code, Better Stack
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 This open-source model can review your code way better than even Phable. You see, most LLMs, even with very well-written prompts, run into a few problems like incomplete coverage where on large codebases, LLMs just tend to cut corners and forget files, unstable quality where you can run a review twice and get completely different results, or position drift where reported issues don’t actually exist in the same reported files. So, the project Open Code Review solves this by running your code through a deterministic pipeline that guarantees the same results no
00:31 matter how many times you run it and no matter how large your codebase. It runs the review through multiple deterministic steps like precision file selection, which determines exactly which files need review, so no important changes are missed. Second is smart file bundling by grouping related files into a single review unit. Third is fine-grained row matching, keeping the model’s attention sharply focused and eliminating information noise at the source. The agents are then used for their strengths of dynamic decision making. This is fine-tuned with prompt
01:02 templates deeply optimized for code review and a fine-tuned set of tools that are much more predictable and stable than the generic tools that you would get with your typical harness. To use this free tool, just install the CLI and then you can review your code. Or you can even scan your entire repository. The project has over 40,000 stars on GitHub after just one week. I’ve been using this project with Opus and going back through my previous projects using the scan tool, and honestly, it’s been really great at improving code quality and catching some bugs that have slipped through using typical setups like Claude Code. You can
01:34 even run this inside CI/CD to catch bugs at their source before they make it into production, meaning you can protect code changes for your entire team. If you want to stay up-to-date with tech and AI news, subscribe to Better Stack.