chore(deps): protoagent 0.108.0 base + pr-reviewer v0.16.0#14
Merged
Conversation
The base bump is the point of this one: protoAgent#2168 (recipe-declared
fan-out width + per-step timings) landed AFTER the v0.107.0 tag, so
`max_concurrency: 5` in the review recipe has been inert since v0.12.0 — the
five finders kept running as 4+1 under the host's default cap of 4, paying the
slowest finder's cost twice.
Measured over 60 reviews before the fix: the five-finder recipe's p50 was 458s
against 322s for the otherwise-identical four-finder one, which solves to ~136s
per finder and ~186s for the sequential synthesize -> verify -> report tail.
Expected saving ~136s per structural review, about 30%.
It also starts `step_s` / `slowest_step` flowing. Every review so far has
recorded them as null, so the split above is an estimate inferred from two
recipe populations; after this it is measured. The sequential tail is the larger
half and is the next thing worth attacking — but not before the numbers exist.
Plugin v0.14.1 -> v0.16.0 brings:
v0.14.2 grounding no longer mistakes prose for a code quote (it downgraded a
TRUE finding on projectBoard-plugin#94 within an hour of deploying)
v0.15.0 `@vera review` — on-demand panel, admin-gated, server-side permission
v0.16.0 every guard reports its DECISION, not only when it fires; and a
finding with `line: 0` no longer silently blocks convergence forever
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The base bump is the point of this one.
The fan-out fix finally takes effect
protoAgent#2168 — recipe-declared fan-out width + per-step timings — landed after the
v0.107.0tag. Somax_concurrency: 5in the review recipe has been inert since v0.12.0: the five finders kept running as 4+1 under the host's default cap of 4, paying the slowest finder's cost twice while the second wave ran a single step alone and saved nothing.Measured over 60 reviews before the fix:
Expected saving: ~136s per structural review, about 30%.
And
step_sstarts flowingEvery review so far recorded
step_s: nullandslowest_step: null, because the timings come from the engine change in the same PR. The split above is an estimate inferred from two recipe populations; after this it is measured.That matters for what comes next: the sequential tail (
synthesize → verify → report, ~186s) is now the larger half, andsynthesizeis close to a no-op when there's one finding. But optimising it before the numbers exist is how you make something slower while believing otherwise.Plugin v0.14.1 → v0.16.0
@vera review— on-demand panel, admin-gated, permission resolved server-side (#28 slice 1)line: 0no longer silently blocks convergence foreverThe observability one is worth calling out: four rules now sit between the panel's findings and the merge gate, and three of them were only visible when they acted — so "ran and correctly declined" was indistinguishable from "never ran". Twice that meant verifying a guard by hand-fetching blobs from GitHub.
Verification
Base and
verified_againstmoved together (they must stay in step). Upstreammaingreen at 260 tests;v0.16.0tagged.Merging deploys — GHCR rebuild then watchtower rolls Vera in ~60s.
🤖 Generated with Claude Code