🤖 feat: add hyper transcript density#3400
Conversation
Adds a global Transcript Density setting and Hyper mode rendering that condenses operational transcript rows into expandable bundles while preserving the underlying message history. Validation: bun test src/browser/features/Messages/OperationalBundleMessage.test.tsx src/browser/utils/messages/transcriptRenderProjection.test.ts src/browser/features/Tools/Shared/HookOutputDisplay.test.ts; make static-check. --- _Generated with `mux` • Model: `openai:gpt-5.5` • Thinking: `xhigh` • Cost: `$114.65`_ <!-- mux-attribution: model=openai:gpt-5.5 thinking=xhigh costs=114.65 -->
Adds Hyper transcript density work bundles that summarize completed assistant turns as worked-duration rows while preserving nested operational bundle collapse inside expanded work. Validation: make static-check. --- _Generated with `mux` • Model: `openai:gpt-5.5` • Thinking: `xhigh` • Cost: `$133.12`_ <!-- mux-attribution: model=openai:gpt-5.5 thinking=xhigh costs=133.12 -->
|
/coder-agents-review |
|
@codex review |
|
Chat: Spend limit reached | View chat Review history
deep-review v0.5.0 | Round 8 | Last posted: Round 8, 27 findings (1 P1, 7 P2, 13 P3, 5 Nit, 1 Note), COMMENT. Review Finding inventoryFindings
Contested and acknowledgedCRF-2 (P2, transcriptRenderProjection.ts:282) - isOperationalBundleMessage bundles ALL tools
Round logRound 1Netero-only. 1 P1, 1 P2, 2 P3, 2 Nit, 1 Note. Reviewed against 5d79d9b..391e632. Round 2BLOCKED. All findings silent. No review. Round 3Panel round. CRF-1,3,4,5,6,7 author fixed. CRF-2 contested, panel closed (4/4 accept). 5 P2, 8 P3, 2 Nit new. Reviewed against 5d79d9b..4ad2ba3. Round 4CRF-8-22 all author fixed (c15c843). All fixes verified by panel. 2 P3, 1 Nit new. Reviewed against 5d79d9b..c15c843. Round 5CRF-23,24,25 all author fixed (e680f6f). 0 open findings. Reviewed against 5d79d9b..e680f6f. Round 6No new code (same SHA as R5). No open findings. No review needed. Round 7Panel round (3 reviewers). CRF-20 regressed (play function deleted in adb13f8). CRF-17 fix unreachable (withFailureIndicator dead code). 1 P2, 1 P3 new. Reviewed against 5d79d9b..a1f3023. Round 8BLOCKED. CRF-26 and CRF-27 both silent. No review. About deep-reviewCRF = Coder Review Finding (P0-P4, Nit, Note)
|
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Line 1 in 391e632
Move context docs under the docs tree
This adds a root-level Markdown document, but /workspace/mux/AGENTS.md says “No free-floating Markdown” and requires user docs to live in docs/ with docs.json navigation. Leaving this at the repo root bypasses the documented docs layout and navigation/link-check workflow; please move it under docs/ (or make it an inline developer note if it is not user-facing).
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
![coder-agents-review[bot]](https://avatars.githubusercontent.com/in/3454270?s=60&v=4){kind=small}
There was a problem hiding this comment.
Hyper Transcript Density is a well-scoped presentation-only feature with clean separation between projection logic and rendering, good use of stable bundle keys, proper ARIA on bundle headers, and solid Storybook coverage. The re-export refactor for hook output helpers and the tool coalesce kind export are both nice deduplication.
1 P1, 1 P2, 2 P3, 2 Nit, 1 Note.
This is a first-pass review only: these are mechanical findings from Netero. The full review panel has not yet reviewed this PR. The panel will review after these findings are addressed.
The rendering order explanation is still relevant for understanding the interplay between operational bundles and tool coalescing, but no replacement comment was added.
Netero, looking at the intersection of three grouping systems
🤖 This review was automatically generated with Coder Agents.
Removes the flaky interaction play test from the transcript density Storybook story while retaining smoke coverage and component-level expansion tests. Validation: bun x test-storybook --url http://127.0.0.1:6006 --testTimeout 30000 src/browser/features/Messages/TranscriptDensity.stories.tsx; make static-check. --- _Generated with `mux` • Model: `openai:gpt-5.5` • Thinking: `xhigh` • Cost: `$133.12`_ <!-- mux-attribution: model=openai:gpt-5.5 thinking=xhigh costs=133.12 -->
|
/coder-agents-review |
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5d6b5e43c3
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
Round 2 is blocked. All 7 findings from round 1 are silent: no code changes address them, no author responses.
Unaddressed findings:
- CRF-1 (P1)
transcriptRenderProjection.ts:163Double-counts reasoning incomputeOperationalBundleInfoswhen reasoning not followed by assistant. - CRF-2 (P2)
transcriptRenderProjection.ts:282isOperationalBundleMessagebundles ALL tools with no allowlist per plan/CONTEXT.md. - CRF-3 (P3)
transcriptRenderProjection.ts:75computeWorkBundleInfosusesComputeOperationalBundleInfosOptionstype name. - CRF-4 (P3)
storage.ts:395isTranscriptDensityexported but no external callers. - CRF-5 (Nit)
ChatPane.tsx:1157Deleted explanatory comments about tool-coalesce rendering order without replacement. - CRF-6 (Nit)
CONTEXT.md:15310 new em-dash (U+2014) instances in added dialogue lines. - CRF-7 (Note)
transcriptRenderProjection.ts:4OperationalBundleMessagetype alias shares name with React component.
Further review is blocked until the author responds to or pushes fixes for these findings. The full review panel has not yet reviewed this PR.
🤖 This review was automatically generated with Coder Agents.
Fixes operational bundle reasoning duplication, aligns transcript density docs with the accepted hyper behavior, cleans minor naming/export issues, and stabilizes Storybook smoke coverage. Validation: bun test src/browser/utils/messages/transcriptRenderProjection.test.ts; bun test src/browser/features/Messages/OperationalBundleMessage.test.tsx src/browser/features/Messages/WorkBundleMessage.test.tsx; bun test src/browser/features/Tools/Shared/HookOutputDisplay.test.ts; bun x test-storybook --url http://127.0.0.1:6006 --testTimeout 30000 src/browser/features/Messages/TranscriptDensity.stories.tsx; make static-check. --- _Generated with `mux` • Model: `openai:gpt-5.5` • Thinking: `xhigh` • Cost: `$133.12`_ <!-- mux-attribution: model=openai:gpt-5.5 thinking=xhigh costs=133.12 -->
|
/coder-agents-review |
|
@codex review |
Scopes work bundle active checks to the candidate assistant turn so completed prior work remains bundled while a later response streams. Validation: bun test src/browser/utils/messages/transcriptRenderProjection.test.ts; make static-check. --- _Generated with `mux` • Model: `openai:gpt-5.5` • Thinking: `xhigh` • Cost: `$133.12`_ <!-- mux-attribution: model=openai:gpt-5.5 thinking=xhigh costs=133.12 -->
|
/coder-agents-review |
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 4ad2ba3f05
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Excludes reasoning-only assistant turns from operational bundling so truncated reasoning remains visible in Hyper density. Validation: bun test src/browser/utils/messages/transcriptRenderProjection.test.ts; make static-check. --- _Generated with `mux` • Model: `openai:gpt-5.5` • Thinking: `xhigh` • Cost: `$133.12`_ <!-- mux-attribution: model=openai:gpt-5.5 thinking=xhigh costs=133.12 -->
|
/coder-agents-review |
|
@codex review |
This comment has been minimized.
This comment has been minimized.
|
Addressed the Codex findings in 9f11dd0:
@codex review |
There was a problem hiding this comment.
The projection layer is well-designed: pure functions, stable keys, compile-time type narrowing via OperationalBundleMemberMessage, and correct React Compiler interaction (ChatPane skipped by compiler due to ref access; manual useMemo is appropriate). The hookOutput extraction refactor is clean, accessibility basics are correct (semantic buttons, aria-expanded, focus outlines), and the storage normalization is solid.
5 P2, 8 P3, 2 Nit. CRF-2 (bundle-everything design) closed by panel vote: 4/4 accept the opt-in Hyper behavior. Adjacent concerns captured as separate findings.
The dominant structural issue is C1+C2: expanded work bundle entries render through a separate code path that drops bashOutputGroup, barriers, and tool coalesce props. Nine reviewers flagged it independently. The root cause is the missing shared render helper the plan called for (Phase 5). A single renderMessageForIndex helper resolving all per-message props would fix the prop divergence, the barrier loss, and the three duplicated toggle handlers in one refactor.
CRF-1 fix verified: leading-reasoning entries correctly cleared on scan reset. CRF-3,4,5,6,7 fixes all verified clean.
"The implementation intentionally remains hyper-aggressive."
The author-agent, updating the spec to match the code rather than the code to match the spec
🤖 This review was automatically generated with Coder Agents.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 9f11dd0259
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
Addressed the latest Codex finding in 78c4f41: failed |
|
@codex review |
|
I could not reproduce this Codex finding on the current head (fa0c167).
Resolving this as stale; please re-review the latest head. |
|
@codex review |
|
Codex Review: Didn't find any major issues. Nice work! ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
There was a problem hiding this comment.
All 15 findings from round 3 verified fixed by the panel. The shared renderMessageAtIndex helper (CRF-10/CRF-11) is the standout fix: it eliminates the prop divergence, barrier loss, and toggle duplication in one refactor. Reasoning order preserved on expansion (CRF-12), failure indicators in collapsed headers (CRF-17), explicit plural labels (CRF-8), and the updated settings copy (CRF-13) and PR description (CRF-14) all land cleanly.
2 P3, 1 Nit remaining. None blocking.
"Two-layer projection composition is structurally sound."
Meruem, after tracing both projection functions through the shared render helper
🤖 This review was automatically generated with Coder Agents.
|
Addressed coder-agents-review follow-ups in e680f6f:
|
|
/coder-agents-review @codex review |
There was a problem hiding this comment.
All 25 findings across 5 rounds are resolved. CRF-23 (undefined duration test), CRF-24 (Object.freeze on entries), and CRF-25 (headIndex doc comment) all addressed cleanly in e680f6f. Codex also reports clean on the latest head.
25 findings total: 1 P1, 6 P2, 12 P3, 4 Nit, 1 Note, 1 contested (closed by panel vote). All fixed or closed. No open findings.
🤖 This review was automatically generated with Coder Agents.
|
Codex Review: Didn't find any major issues. You're on a roll. ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
|
/coder-agents-review |
There was a problem hiding this comment.
No new code since round 5 (same SHA e680f6f). All 25 findings remain resolved. R4 approval stands. Nothing to review.
🤖 This review was automatically generated with Coder Agents.
|
/coder-agents-review |
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a1f30232ef
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
Addressed the Codex documentation-placement finding in 487d62a by removing the root |
|
@codex review |
|
/coder-agents-review |
|
@claude review |
|
Codex Review: Didn't find any major issues. What shall we delve into next? ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
There was a problem hiding this comment.
Round 7: two new commits since R5 (storybook test stabilization, review parser update). Netero clean. Panel (3 reviewers) found two issues:
1 P2 (CRF-20 regression), 1 P3 (CRF-17 fix unreachable).
CRF-26 is the higher-priority item: the "stabilize" commit deleted the play function that was the CRF-20 fix, removing the only interaction-level test for the composed nesting path (work bundle -> operational bundle -> messages). This was verified fixed in R4 and regressed in a subsequent commit.
CRF-27 is a convergence finding (2 reviewers independently traced the same root cause): withFailureIndicator scans for status === "failed" tools, but isBundleableToolMessage only admits completed/pending/executing tools into bundles. Failed tools never enter bundles, so the failure indicator never fires. The test passes because it calls summarizeOperationalBundle directly with fabricated failed tools, bypassing the bundling guard. The user-facing behavior is correct (failed tools stay visible as first-class rows), but the dead code and misleading test should be cleaned up.
🤖 This review was automatically generated with Coder Agents.
There was a problem hiding this comment.
Round 8 is blocked. Both open findings from round 7 are silent: no code changes address them, no author responses.
- CRF-26 (P2)
TranscriptDensity.stories.tsx:91CRF-20 fix reverted: play function deleted in adb13f8; no interaction coverage for nesting path. - CRF-27 (P3)
transcriptRenderProjection.ts:317withFailureIndicatoris dead code;isBundleableToolMessageexcludes failed tools before they enter bundles.
The only new commit (487d62a, "remove root transcript glossary changes") addresses a Codex documentation-placement finding, not our findings. CI is green.
Further review is blocked until the author responds to or pushes fixes for CRF-26 and CRF-27.
🤖 This review was automatically generated with Coder Agents.
|
Addressed coder-agents-review follow-ups in a77708c:
|
|
/coder-agents-review |
|
@codex review |
|
Codex Review: Didn't find any major issues. Chef's kiss. ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
|
/coder-agents-review |
1 participant
Summary
Adds Hyper Transcript Density, a presentation-only transcript mode that compresses dense agent work into expandable bundles while preserving underlying history, provider context, backend schemas, and Normal transcript rendering.
Background
Long execution turns can push the useful final answer and conversational pacing off screen. Hyper density keeps the final response visible, summarizes completed turn work with a duration row, and keeps detailed reasoning/tool activity available on demand.
Implementation
Worked for …row; expanding it uses the same message render path as normal transcript rows so bash output grouping, barriers, task links, review notes, and file-tool coalescing stay consistent.Validation
bun test src/browser/utils/messages/transcriptRenderProjection.test.tsbun test src/browser/features/Messages/OperationalBundleMessage.test.tsx src/browser/features/Messages/WorkBundleMessage.test.tsxbun test src/browser/features/Tools/Shared/HookOutputDisplay.test.tsbun x test-storybook --url http://127.0.0.1:6006 --testTimeout 30000 src/browser/features/Messages/TranscriptDensity.stories.tsxmake typecheckmake static-checkRisks
Medium UI regression risk in transcript rendering because the ChatPane render loop now coordinates both work-level and operation-level grouping. The feature is gated to Hyper density, leaving Normal mode on the existing rendering path.
Pains
The implementation direction changed after the original conservative plan: Hyper now intentionally uses turn-level Work Bundles and nested Operational Bundles rather than only grouping a narrow allowlist of adjacent tool rows. The PR body now describes the shipped behavior directly instead of retaining the obsolete plan text.
Generated with
mux• Model:openai:gpt-5.5• Thinking:xhigh• Cost:$165.33