feat(web): multi-phase review agent with per-file parallel LLM calls - #1164
Open
fatmcgav wants to merge 3 commits into
Open
feat(web): multi-phase review agent with per-file parallel LLM calls#1164fatmcgav wants to merge 3 commits into
fatmcgav wants to merge 3 commits into
Conversation
Copy link
Copy Markdown
Contributor
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
|
Too much diff to scan? Review this PR in Change Stack to start with the highest-impact changes. No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: 4266d122-e4a5-40cb-b534-fc72d5945969 📥 CommitsReviewing files that changed from the base of the PR and between 08481cd and 6bc5a12. 📒 Files selected for processing (2)
WalkthroughUpdates docs and changelog to describe the AI Code Review Agent’s new two-phase flow: a single MR-summary LLM pass for cross-file concerns, then parallel per-file LLM reviews (up to 5 concurrent) that post inline comments; also changes the manual trigger from /review to review. ChangesDocumentation & Changelog
Sequence Diagram(s)sequenceDiagram
participant Client
participant ReviewAgent as ReviewAgent.generatePrReview
participant MRLLM as LLM (MR Summary)
participant FileLLM as LLM (Per-file Review)
Client->>ReviewAgent: pr_payload
ReviewAgent->>MRLLM: generateMrSummary(pr_payload)
MRLLM-->>ReviewAgent: pr_summary context
ReviewAgent->>FileLLM: per-file prompt + pr_summary + file diffs (parallel, max 5)
FileLLM-->>ReviewAgent: file_diff_review (inline comments)
ReviewAgent-->>Client: aggregated file_diff_review[]
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
Suggested reviewers
❌ Failed checks (1 warning)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches 🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. ❤️ ShareComment @coderabbitai help to get the list of available commands and usage tips. |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)docs/docs/features/agents/review-agent.mdx (1)🤖 Prompt for all review comments with AI agents6-23: ⚡ Quick win
Rewrite this new section in second person.
The new intro and How it works steps switch to third person (the agent runs, the agent posts, it returns). Please recast this as you/your to match the docs voice.
As per coding guidelines, docs/**/*.mdx should “write in second person present tense”.
🤖 Prompt for AI AgentsVerify each finding against the current code and only fix it if needed. In `@docs/docs/features/agents/review-agent.mdx` around lines 6 - 23, Rewrite the introduction and "How it works" steps to use second-person present-tense phrasing (you/your) instead of third-person; change sentences like "This agent provides", "When a review is triggered, the agent runs", "the agent posts", and "it returns nothing" to address the reader directly and use present tense, and update step headings and descriptions ("MR summary pass", "Per-file reviews", "Inline comments") so they read like instructions to the user (e.g., "You run a MR summary pass", "You run per-file reviews", "You post inline comments" or similar) while preserving the original meaning and sequence.
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@packages/web/src/features/agents/review-agent/nodes/generateMrSummary.ts`:
- Around line 31-32: The code incorrectly uses the undefined variable modelName
in the models.find call inside the REVIEW_AGENT_MODEL branch; update the
predicate to compare m.displayName === env.REVIEW_AGENT_MODEL so it matches the
condition and the later usage (replace the modelName reference with
env.REVIEW_AGENT_MODEL in the models.find call that resolves the model).
---
Nitpick comments:
In `@docs/docs/features/agents/review-agent.mdx`:
- Around line 6-23: Rewrite the introduction and "How it works" steps to use
second-person present-tense phrasing (you/your) instead of third-person; change
sentences like "This agent provides", "When a review is triggered, the agent
runs", "the agent posts", and "it returns nothing" to address the reader
directly and use present tense, and update step headings and descriptions ("MR
summary pass", "Per-file reviews", "Inline comments") so they read like
instructions to the user (e.g., "You run a MR summary pass", "You run per-file
reviews", "You post inline comments" or similar) while preserving the original
meaning and sequence.
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info ⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: ab12e7a3-a174-49f2-b51f-16607592ad85
📥 CommitsReviewing files that changed from the base of the PR and between 29dfe17 and f643f86.
📒 Files selected for processing (6)- docs/docs/features/agents/review-agent.mdx
- packages/web/src/features/agents/review-agent/nodes/generateDiffReviewPrompt.ts
- packages/web/src/features/agents/review-agent/nodes/generateMrSummary.ts
- packages/web/src/features/agents/review-agent/nodes/generatePrReview.ts
- packages/web/src/features/agents/review-agent/nodes/gitlabMrParser.ts
- packages/web/src/features/agents/review-agent/nodes/invokeDiffReviewLlm.ts
Sorry, something went wrong.
Comment thread
packages/web/src/features/agents/review-agent/nodes/generateMrSummary.ts
Outdated
Replace the per-chunk (N LLM calls) architecture with a two-phase approach: 1. A single MR summary pass over all changed files to detect cross-file semantic changes (renames, signature changes, removed exports, etc.) that individual file reviewers should be aware of. 2. Per-file LLM reviews that batch all hunks for a file into a single call, parallelised across files via a concurrency-capped pool (MAX_CONCURRENT_FILE_REVIEWS = 5). This reduces LLM calls from one-per-hunk to one-per-file (plus one summary call), while giving each file review the full picture via the MR summary context. Additional changes: - Export `validateLogPath` from `invokeDiffReviewLlm` for reuse in the summary node - Add "How it works" section to the review agent docs Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
fatmcgav
force-pushed
the
feat-improve-agent-llm-efficiency
branch
from
08481cd to
6bc5a12
Compare
June 5, 2026 19:41
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Reviewers
At least 1 approving review is required to merge this pull request.
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Replace the per-chunk (N LLM calls) architecture with a two-phase approach:
semantic changes (renames, signature changes, removed exports, etc.)
that individual file reviewers should be aware of.
parallelised across files via a concurrency-capped pool
(MAX_CONCURRENT_FILE_REVIEWS = 5).
This reduces LLM calls from one-per-hunk to one-per-file (plus one summary
call), while giving each file review the full picture via the MR summary
context.
Additional changes:
summary node
Co-Authored-By: Claude Sonnet 4.6 noreply@anthropic.com
Summary by CodeRabbit
New Features
Improvements
Documentation