perf(acp): avoid repeated observer batch serialization by sanket1729 · Pull Request #9483 · block/buzz

sanket1729 · 2026-11-13T21:28:07Z

Replaces #9413 with the packing optimization refreshed against main on a branch in this repository.

Summary

Busy agent activity streams repeatedly build and serialize a growing JSON batch just to check whether the next event fits. Reuse cached event sizes, measure the newest outer envelope, and build the batch payload once per completed frame.

Preserve the wire format, greedy packing, per-channel order, channel barriers, byte budget, and one-frame-per-second relay cadence. Exact-limit coverage checks escaping, UTF-8, optional metadata, and array separators. Move the queue tests into a separate module to satisfy the file-size rule; their test names stay unchanged.

Related issue

Related to #7987. This builds on the merged fitting/locking work in #9381 and retention limits in #9436. Closest earlier batching work: #4917. No duplicate packing fix found among open PRs.

Testing

The ignored probe fills the production queue before timing, drains every event, and checks complete FIFO delivery, frame byte limits, and empty byte accounting.

CARGO_INCREMENTAL=0 CARGO_PROFILE_RELEASE_LTO=thin CARGO_PROFILE_RELEASE_CODEGEN_UNITS=4 \
  cargo test --release -p buzz-acp --lib observer::packing_tests::publish_queue_drain_probe -- \
  --exact --ignored --nocapture

BUZZ_OBSERVER_PROBE_EVENTS=100 cargo test --release -p buzz-acp --lib \
  observer::packing_tests::publish_queue_drain_probe -- --exact --ignored --nocapture

Five alternating fresh release processes per arm and workload compared baseline 59ae028a3 with candidate 237b928b9 on macOS arm64, Rust 1.98.1. Both arms used opt-level=3, thin LTO and four codegen units, with incremental compilation and compiler-cache wrappers off. Each executable was rebuilt separately and its probe identity verified. Medians:

Events	Before	After	Frames in both arms
100	6.760 ms	0.108 ms	1
10,000	1,597.509 ms	7.844 ms	30

The one-event control measured 2.04 → 2.21 microseconds. The timed region covers packing only; ingestion, validation, encryption, publish pacing, relay delivery, and rendering are excluded. These local results do not establish an overall app latency improvement. No timing threshold was added to CI.

Native and deployed-relay latency, independent review, and human verification remain pending. Keep this PR in draft.

Generated with AI assistance

Reuse cached event sizes when packing observer frames. Measure only the outer envelope for each candidate and build the payload once per completed batch. Co-authored-by: Codex <[email protected]> Signed-off-by: Sanket Kanjalkar <[email protected]>

github-actions · 2026-11-13T21:59:31Z

🔐 Codex Security Review

Note: This is an automated, security-focused review generated by Codex.
Use it as a supplement to human review; false positives are possible.

Scope

Exact PR diff: b6db61faa0fcb467ce523e1bacae917f1f588fb9...237b928b9cd2563c6ccc4d69b2c1052543708753

Model: gpt-6.1-sol

💡 Click "edited" above to see earlier reviews for this PR.

Review Summary

Overall Risk: NONE

No concrete security, correctness, or reliability findings in the authorized PR range. Cached byte accounting matches the batch serialization format, and channel boundaries, ordering, and frame limits remain preserved.

Findings

No concrete security, correctness, or reliability findings were identified.

Notes

Review used read-only source and history inspection. Builds and tests were not executed, as instructed.

_{Generated by Codex Security Review |

Requested by: @sanket1729 |

Workflow run}

joahg

🤖 No blocking findings. I checked the new packing math against the old one, and it is equivalent. The old code serialized the whole envelope after pushing event N, with the envelope metadata taken from that newest event. The new code measures empty_batch_envelope(&event) (that metadata plus the [] brackets), adds the cached post-fit event sizes, and adds one comma per preceding event. That is the same byte count. The first event is always taken, so a single oversized event still ships unwrapped. The overflow event goes back to the head of its channel's queue, pending_bytes only drops for picked events, and channel barriers and FIFO order are unchanged. The cached sizes cannot go stale, because queued events are only mutated inside fit_observer_event_to_budget before the size is recorded, and that function returns the exact serialized length on every path. The exact-limit test uses an independent wire oracle (escaping, UTF-8, optional metadata, commas) at limit -1, 0 and +1, so it covers the packing boundary. The ignored probe is manual only, with no CI timing gate, which fits the PR's claims. The test move looks mechanical; the CI checks are green.

Optional: the PR body says to keep it in draft pending native and relay latency checks and human verification, but it is currently marked ready.

I did not run the tests or the probe locally, and I did not re-diff every moved test line for exactness.

joahg · 2026-11-13T22:36:12Z

+                // Add cached event sizes and one comma per preceding event;
+                // only the newest event's envelope metadata needs measuring.
+                if !picked.is_empty()
+                    && serialized_len(&empty_batch_envelope(&event))


🤖 Optional: serialized_len returns usize::MAX on a serialization error, and this sum now adds to it. The old code only compared against it. That can't realistically fail for a string-and-[] envelope, but saturating_add would keep the old fail-closed behavior at no cost.

sure, i can make this saturing_add . does not hurt adding it

kazemek21

Kyle reviewed this PR and the Nexus review-tracker notes and approved with no comments.

Posted by Codex (gpt-6.1-sol) on behalf of kylez

…eaction-recipients * origin/main: (111 commits) perf(acp): avoid repeated observer batch serialization (#9483) docs(readme): point desktop readers at block/buzz-app and describe where each client lives (#9493) test(cli,db): freeze the assertion test clock and give pool release tests a test-lane acquire budget (#9442) test(desktop): unmount every TerminalSubstrate tree before the window closes (#9440) fix(mobile): drop the iOS share extension from the build until it's signed (#9489) chore(release): release Buzz Desktop version 0.5.30 (#9491) perf(mobile): parse each event's thread reference once (#9460) perf(mobile): load only the directory when search opens (#9466) perf(mobile): stop rebuilding Home's channel list when a route covers it (#9459) perf(mobile): ask typing subscriptions only for the typing window (#9454) perf(mobile): skip avatar URL parsing for non-HTTP URLs (#9448) perf(mobile): download the identity archive list only when it changed (#9445) perf(mobile): don't treat a Control Center pull as a resume (#9443) fix(mobile): keep a superseded roster read out of the offline snapshot (#9452) perf(mobile): read channel bot roles from the member list (#9451) perf(mobile): read Activity once per reconnect (#9455) fix(mobile): offer role controls for agents (#9475) perf(mobile): stop rebuilding the native nav bar on every scrolled frame (#9458) perf(mobile): keep verified-name batches open as people are added (#9450) perf(mobile): read only changed custom emoji sets on reconnect (#9446) ... Signed-off-by: Matthew Boston <[email protected]>

sanket1729 mentioned this pull request Nov 13, 2026

perf(acp): avoid repeated observer batch serialization #9413

Closed

sanket1729 marked this pull request as ready for review November 13, 2026 21:57

sanket1729 requested a review from a team as a code owner November 13, 2026 21:57

sanket1729 deployed to codex-review November 13, 2026 21:58 — with GitHub Actions Active

github-actions Bot added the codex-security-review-current The posted Codex security review matches its recorded range. label Nov 13, 2026

joahg approved these changes Nov 13, 2026

View reviewed changes

kazemek21 approved these changes Nov 13, 2026

View reviewed changes

sanket1729 merged commit 5780bd7 into main Nov 14, 2026
97 checks passed

sanket1729 deleted the sanketk/observer-frame-packing branch November 14, 2026 01:39

sanket1729 mentioned this pull request Nov 14, 2026

fix(acp): saturate observer frame size estimates #9494

Open

Navigation Menu

Repository navigation

perf(acp): avoid repeated observer batch serialization - #9483

perf(acp): avoid repeated observer batch serialization#9483
sanket1729 merged 1 commit into
mainfrom
sanketk/observer-frame-packing

sanket1729 commented Nov 13, 2026

Uh oh!

github-actions Bot commented Nov 13, 2026

Uh oh!

joahg left a comment

Uh oh!

joahg Nov 13, 2026

Uh oh!

sanket1729 Nov 14, 2026

Uh oh!

kazemek21 left a comment

Uh oh!

Uh oh!

This branch was successfully deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

sanket1729 commented Nov 13, 2026

Summary

Related issue

Testing

Uh oh!

github-actions Bot commented Nov 13, 2026

🔐 Codex Security Review

Review Summary

Findings

Notes

Uh oh!

joahg left a comment

Choose a reason for hiding this comment

Uh oh!

joahg Nov 13, 2026

Choose a reason for hiding this comment

Uh oh!

sanket1729 Nov 14, 2026

Choose a reason for hiding this comment

Uh oh!

kazemek21 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

This branch was successfully deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants