fix(code-reviews): show model and tokens in review summary for v2 reviews#978
Merged
alex-alecu merged 2 commits intomainfrom Mar 10, 2026
Merged
fix(code-reviews): show model and tokens in review summary for v2 reviews#978alex-alecu merged 2 commits intomainfrom
alex-alecu merged 2 commits intomainfrom
Conversation
…eviews PR #407 added model/token tracking to the review summary, but it only works for v1 (SSE) reviews. All reviews now use v2 (cloud-agent-next), which skips SSE stream processing entirely, so usage is never reported. Fix: when the code_reviews record has no usage data, query the billing tables (microdollar_usage + metadata) by cli_session_id to get the model, tokens, and cost. Back-fill the code_reviews record so future reads skip the aggregation.
![kilo-code-bot[bot]](https://avatars.githubusercontent.com/in/2193792?s=60&v=4){kind=small}
Contributor
Code Review SummaryStatus: 2 Issues Found | Recommendation: Address before merge Overview
Issue Details (click to expand)WARNING
Fix these issues in Kilo Cloud Other Observations (not in diff)None. Files Reviewed (2 files)
|
The billing query grouped by model and picked one row, so tokens from other models in the same session were lost. Split into two queries: one for session-wide totals, one for the dominant model name.
jeanduplessis
approved these changes
Mar 10, 2026
| let review = await getCodeReviewById(reviewId); | ||
|
|
||
| // Short poll: usage may arrive from the orchestrator just before the callback | ||
| for (let attempt = 0; attempt < MAX_RETRIES && review && !review.model; attempt++) { |
Contributor
There was a problem hiding this comment.
WARNING: This adds a fixed delay to every v2 completion callback
The comment above says cloud-agent-next reviews never receive orchestrator usage, but this loop still waits through the full exponential backoff whenever review.model is empty. For v2 reviews that means every completion path pays ~1.4s before we even try the billing fallback, which delays the reaction/comment update for every successful review.
| cliSessionId: string | ||
| ): Promise<SessionUsageSummary | null> { | ||
| try { | ||
| const sessionFilter = eq(microdollar_usage_metadata.session_id, cliSessionId); |
Contributor
There was a problem hiding this comment.
WARNING: This fallback filters on an unindexed column
microdollar_usage_metadata only has a created_at index in the schema today, so both aggregation queries will end up scanning the metadata table by session_id on every completed v2 review. Because this runs on the completion callback path, larger billing tables will make review completion and summary updates noticeably slower.
5 tasks
alex-alecu
added a commit
that referenced
this pull request
Mar 10, 2026
…on v2 reviews (#979) ## Summary Follow-up to [PR #978](#978). The billing fallback query that fetches token/model data for v2 reviews was timing out in production, so the usage footer ("Reviewed by model · X tokens") was never shown. **Root cause:** the query filters `microdollar_usage_metadata` by `session_id`, but that column has no index. The table has ~469M rows, so every query did a full table scan and timed out. The `catch` block silently returned `null`, and the footer was skipped. **Fix:** - Add a `created_at >= reviewCreatedAt` lower bound to the billing query. This lets Postgres use the existing `created_at` index (query cost drops from full-scan to ~288). Billing rows can't exist before the review was created, so the bound is exact. - Skip the v1 poll loop for v2 reviews (saves ~1.4s of wasted retries). - Remove the `session_id` index migration — with the time bound, it's not needed. - Clean up the admin dashboard: remove agent version filter and performance chart that are no longer useful now that all reviews are v2. ## Verification - [x] `pnpm typecheck` — no new errors (only pre-existing kiloclaw errors) - [x] `pnpm test usage-footer` — 10/10 pass - [x] `pnpm test schema` — 15/15 pass (no unmigrated schema changes) - [x] Checked `EXPLAIN` plan on prod DB — query uses `idx_microdollar_usage_metadata_created_at` with cost ~288 - [x] Confirmed billing data exists for test session `ses_3282e02f5ffe2vPRBSqdpc0e40` (PR #981 review) — 8 rows returned in <1s with time-bounded query ## Visual Changes N/A ## Reviewer Notes - Every completed v2 review in prod has `model = NULL` — the billing fallback has never worked. This fix unblocks all future v2 reviews. - The back-fill write (fire-and-forget) still runs after fetching billing data, so repeat reads skip the aggregation.
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
2 participants
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.
Summary
PR #407 added model + token info to the PR review summary comment (the "Reviewed by claude-sonnet-4.6 · 12,345 tokens" footer). But it only works for v1 (SSE-based) reviews. All reviews now run on v2 (cloud-agent-next), which is callback-based — it never streams SSE events, so usage data is never collected and the
model,total_tokens_in,total_tokens_outcolumns stay null.The fix: when the
code_reviewsrecord has no usage data, we query the billing tables (microdollar_usage+microdollar_usage_metadata) bycli_session_id. The billing system already tracks every LLM call with model, tokens, and cost — we just aggregate it per session. We also back-fill thecode_reviewsrecord so future reads don't repeat the aggregation.Verification
pnpm typecheck— no new errors (pre-existing errors in kiloclaw only)pnpm test usage-footer— 10/10 passmicrodollar_usage_metadata.session_idmatchescode_reviews.cli_session_idVisual Changes
N/A
Reviewer Notes
code_reviewsis fire-and-forget (.catch()) — if it fails, the footer still shows correctly; we just won't cache the result.ExecutionCallbackPayload, but that's a bigger change. This fix works today with no changes outside the Next.js app.