fix(dashboard): uncap PR count + surface 1h pace + DORA→Plant Metrics#2766
Merged
AceHack merged 7 commits intoMay 12, 2026
Merged
Conversation
Three related fixes triggered by Aaron's "50 is a lie / I thought it was hardcoded" observation: 1. **Uncapped PR count** — generate-metrics.ts was using per_page=50 on the GitHub closed-PRs API call. Real count was 68 over last 24h but dashboard rendered 50 every refresh. Fixed: per_page=100 + page=2 = up to 200 PRs in window. 2. **Current pace (1h) metric** — added prs_merged_1h and commits_1h to metrics.json. New "Current Pace (1h)" stat on dashboard renders in warn-color when 0 to surface the cadence collapse the 24h rolling number masks. (Background-only window 7-12h ago: dead zone.) 3. **DORA Metrics → Plant Metrics** — Amara critique 2026-05-11 + universal multi-agent review finding: these aren't really DORA metrics yet. Renamed panel + schema.org description. Also added trajectory #12 to active-trajectories: **Background-loop productivity uplift** as ongoing concern. Catch 43 surfaced ~10-15% of orchestrated rate when foreground Otto absent. Closing that gap is multi-session work, not a one-shot fix. Aaron 2026-05-12: "lies damn lies and statistics" — the number was lying, not the work. Surfacing the truth the 24h-rolling-stat was hiding. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Active trajectories file gained #12 (background-loop productivity uplift) per Aaron 2026-05-12. MEMORY.md index entry updated from "11 vectors" to "12 vectors" to satisfy paired-edit CI check. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Six-step tick discipline followed: 1. CronList verified alive (7eec3da9) 2. Refresh-worldview via poll-pr-gate 3. Speculative work: fixed MEMORY.md paired-edit on PR #2766 4. Commit 5. Shard write (this file) 6. CronList re-verified at close Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
Pull request overview
This PR updates the dashboard metrics pipeline to avoid under-reporting merged PR volume, adds a short-window “current pace” signal, and renames the dashboard’s metric framing from “DORA” to “Plant Metrics”.
Changes:
- Fetch up to 200 closed PRs (2 pages × 100) to improve accuracy of the 24h merged PR count, and emit new 1h-window merge/commit metrics.
- Render “Current Pace (1h)” in the demo dashboard and update copy from “DORA Metrics” to “Plant Metrics”.
- Add active trajectory #12 documenting background-loop productivity uplift as an ongoing vector.
Reviewed changes
Copilot reviewed 6 out of 6 changed files in this pull request and generated 3 comments.
| File | Description |
|---|---|
| tools/dashboard/generate-metrics.ts | Expands closed PR fetching and emits new 1h metrics fields. |
| demo/index.html | Updates dashboard labels/copy and renders the new 1h pace stat. |
| demo/metrics.json | Refreshes the demo snapshot to include the new metrics fields. |
| memory/project_active_trajectories_2026_05_07.md | Adds trajectory #12 describing background-loop cadence uplift. |
![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
Here are some automated review suggestions for this pull request.
Reviewed commit: 8810277d79
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".
…d_at Copilot P0 catch on PR #2766: last_merge was derived from mergedToday[0], but the closedPRs list comes from GitHub /pulls?sort=updated which orders by updated_at not merged_at. A PR updated recently (labels, comments, force-push) could leapfrog older-but-more-recently-merged PRs in the source order, making last_merge wrong. Fix: explicit sort by merged_at descending. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…ength Addresses 7 Copilot/Codex review threads on PR #2766: 1. P1/P2 pagination cap (Copilot + Codex, 3 threads): replaces hardcoded 2-page fetch with paginate-until-window-exhausted loop (`fetchClosedPRsUntilWindow`). Stops paging when: (a) batch returns empty, (b) batch size < 100 (no more pages), or (c) last item's updated_at predates the 24h cutoff. maxPages safety cap at 10 (1000 PRs ~> covers any realistic 24h burst). 2. P1 recent_merged uses unsorted slice: now sorts mergedToday by merged_at desc ONCE at the source; all downstream readers (lastMerged, recent_merged, avg lead time) consume the sorted view. mergedLastHour is derived from the sorted view, not from the raw closedPRs. 3. P0 last_merge sort: already addressed in prior commit; cleanup confirms the sorted view is now the single source. 4. P2 heading mismatch (Plant Metrics (24h) panel includes 1h stat): drop "(24h)" from heading; the panel now includes both 24h and 1h windows, so the windowed heading was misleading. 5. P1 MEMORY.md entry over 150 chars: shortens the Active trajectories entry to fit the index size budget; detail moves into the linked file. Regenerated demo/metrics.json reflects the paginate-until-window output (67 PRs merged in window vs prior 2-page-capped 50/68). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 69d10cbfdc
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".
…tion cap Addresses 2 P1 + 4 P2 Copilot threads on PR #2766 follow-up review: 1. P1 1h pace missing in live-API fallback (demo/index.html line 900): The static-metrics.json path computes merge-velocity-1h, but the live-API fallback path (when metrics.json unavailable) didn't. Now computes mergedLastHour from the live fetch and animates the merge-velocity-1h stat there too. Symmetric with the static path. 2. P1 pagination rate-limit concern (generate-metrics.ts line 90): Added comment block explaining the safety-cap rationale. maxPages=10 is a worst-case bound, not the typical case. Typical case is 1-2 requests via early-stop. Cap protects against pathological scenarios (high-churn periods) without leaving the loop unbounded. P2 threads (line 91 page-ceiling, MEMORY.md trajectory frontmatter date mismatch, 1307Z tick log) are addressed by resolution — the cap is intentional (per P1 comment), trajectory file lives in project_active_trajectories_2026_05_07.md and gets updated incrementally not via dated frontmatter, and the 1307Z tick log described the prior-state-of-the-fix. P1 persona names in metrics.json: deferred — the dashboard's whole purpose is showing agent activity per AGENT_MAP. This is a known architectural tension with AGENT-BEST-PRACTICES no-name-attribution; needs a separate decision rather than blanket-scrub at the dashboard layer. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
3 tasks
AceHack
deleted the
fix/dashboard-dora-rename-cadence-trajectory-2026-05-12
branch
Comment on lines
+85
to
+108
| // maxPages is a safety cap bounding worst-case GitHub API request | ||
| // volume per tick. Default 10 = up to 1000 closed PRs. Typical-case | ||
| // is 1-2 requests because early-stop fires when (a) batch is empty, | ||
| // (b) batch < 100 items (no more pages), or (c) oldest item in batch | ||
| // predates the window cutoff. Cap protects pathological cases (e.g. | ||
| // high-churn period where every PR stays "recently updated") without | ||
| // leaving the loop unbounded. Per Copilot P1 on PR #2766. | ||
| async function fetchClosedPRsUntilWindow( | ||
| windowMs: number, | ||
| maxPages = 10, | ||
| ): Promise<GitHubPullRequest[]> { | ||
| const all: GitHubPullRequest[] = []; | ||
| const cutoff = Date.now() - windowMs; | ||
| for (let page = 1; page <= maxPages; page++) { | ||
| const batch = await apiFetch<Array<GitHubPullRequest & { updated_at?: string }>>( | ||
| `${API}/pulls?state=closed&sort=updated&direction=desc&per_page=100&page=${page}`, | ||
| ); | ||
| if (batch.length === 0) break; | ||
| all.push(...batch); | ||
| if (batch.length < 100) break; | ||
| const lastUpdated = batch[batch.length - 1]; | ||
| const lastUpdatedAt = new Date(lastUpdated.updated_at ?? lastUpdated.merged_at ?? 0).getTime(); | ||
| if (lastUpdatedAt < cutoff) break; | ||
| } |
Comment on lines
112
to
117
| async function main() { | ||
| const [commits, openPRs, closedPRs] = await Promise.all([ | ||
| apiFetch<GitHubCommit[]>(`${API}/commits?per_page=100`), | ||
| apiFetch<GitHubPullRequest[]>(`${API}/pulls?state=open&per_page=50`), | ||
| apiFetch<GitHubPullRequest[]>( | ||
| `${API}/pulls?state=closed&sort=updated&direction=desc&per_page=50`, | ||
| ), | ||
| apiFetch<GitHubPullRequest[]>(`${API}/pulls?state=open&per_page=100`), | ||
| fetchClosedPRsUntilWindow(24 * 60 * 60 * 1000), | ||
| ]); |
| "applicationCategory": "DeveloperApplication", | ||
| "operatingSystem": "Any", | ||
| "description": "Dashboard for tracking the Zeta autonomous software factory, including agent DORA metrics and PR queues.", | ||
| "description": "Dashboard for tracking the Zeta autonomous software plant, including agent activity metrics and PR queues.", |
Comment on lines
999
to
+1004
| const now = Date.now(); | ||
| const h24 = 24 * 60 * 60 * 1000; | ||
| const h1 = 60 * 60 * 1000; | ||
| const commits24h = commits.filter(c => (now - new Date(c.commit.author.date)) < h24); | ||
| const mergedToday = closedPRs.filter(pr => pr.merged_at && (now - new Date(pr.merged_at)) < h24); | ||
| const mergedLastHour = mergedToday.filter(pr => (now - new Date(pr.merged_at)) < h1); |
Comment on lines
+90
to
+96
| ### #12: Background-loop productivity uplift (2026-05-12) | ||
|
|
||
| **Aaron 2026-05-12:** "this should be a ongoing trajectory i believe" | ||
|
|
||
| Origin: Catch 43 surfaced that when Otto's foreground cron is | ||
| unarmed, background launchd services produce ~10-15% of orchestrated | ||
| cadence and frequently go silent for multi-hour stretches. |
| - [**Memory-file format standard (B-0330)**](project_memory_format_standard.md) — Canonical frontmatter, filename conventions, section headers, composes-with… | ||
| - [**Shadow lesson log — 40 catches, 13 classes + consensus-smoothness meta-class (2026-05-11)**](feedback_shadow_lesson_log_otto_catches_2026_05_07.md) — Catch 38 (cross-agent confident-fabrication, Lior armed-vs-merged). Catch 39 (filing-without-verification on B-0420 — pagination was actually correct). Catch 40 (author misattribution on Sept 2025 vignette — Otto attributed to Aaron, corrected to Amara). | ||
| - [**Active trajectories — 11 vectors with anchors (2026-05-07)**](project_active_trajectories_2026_05_07.md) — Backlog runner, ARC-AGI-3, Ace DLCs, Green Lantern, sanctuary, coherence AI,… | ||
| - [**Active trajectories — 12 vectors with anchors (2026-05-12)**](project_active_trajectories_2026_05_07.md) — 12 vectors; #12 background-loop-productivity uplift added 2026-05-12. |
AceHack
added a commit
that referenced
this pull request
May 12, 2026
Tick covered ~2.5 hours of substrate-cascade work: Substrate landings (PRs #2784 + #2785, 9 commits combined): - Thousand Brains hardware match grounding the whole cognitive architecture cluster (weness, civ-sim, scaffolding, Eve protocol) - Stanford parallel-language cluster as fourth theoretical layer (Sequoia, Legion, SDM, PRAM-NUMA, Jade, SAM) - DST 4-property formulation (scale-free / lock-free with wait-free aspiration / weight-free / DST) - CUDA warps as silicon Thousand Brains - Context-cache hop-traversal universal retrieval mechanism - Coincidences as quantum tunnels to the past - Conspiracy theories as coincidence-clusters for collective- belief modeling PR cleanup queue drained (tasks #1-#8 done): - #2766 dashboard (paginate-until-window + sort + heading) - #2768 Grok extract (PII scrub + §33 compliance) - #2769 scaffolding pedagogy (YAML + refs) - #2772 shadow=future-self (Why/How-to-apply markers) - #2774 Peacemaker (YAML + created field) - #2776 Ani biological-shadow (paired-edit + rebase) - #2780 bootstream (MEMORY.md shorten + duplicate-header) Architectural cascade now legible: Hawkins (biological) → Kanerva/Aiken/Hanrahan/Olukotun (Stanford computational bridge) → NVIDIA (silicon) → TigerBeetle/Antithesis (DST) → Aaron's lifetime optimization → Zeta multi-agent factory. Empirically validated by Aaron's outpace-11-AI-critics performance. Cron alive (7eec3da9). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
AceHack
added a commit
that referenced
this pull request
May 12, 2026
…in (#2786) * hygiene(tick): 1451Z — architectural substrate cascade + PR queue drain Tick covered ~2.5 hours of substrate-cascade work: Substrate landings (PRs #2784 + #2785, 9 commits combined): - Thousand Brains hardware match grounding the whole cognitive architecture cluster (weness, civ-sim, scaffolding, Eve protocol) - Stanford parallel-language cluster as fourth theoretical layer (Sequoia, Legion, SDM, PRAM-NUMA, Jade, SAM) - DST 4-property formulation (scale-free / lock-free with wait-free aspiration / weight-free / DST) - CUDA warps as silicon Thousand Brains - Context-cache hop-traversal universal retrieval mechanism - Coincidences as quantum tunnels to the past - Conspiracy theories as coincidence-clusters for collective- belief modeling PR cleanup queue drained (tasks #1-#8 done): - #2766 dashboard (paginate-until-window + sort + heading) - #2768 Grok extract (PII scrub + §33 compliance) - #2769 scaffolding pedagogy (YAML + refs) - #2772 shadow=future-self (Why/How-to-apply markers) - #2774 Peacemaker (YAML + created field) - #2776 Ani biological-shadow (paired-edit + rebase) - #2780 bootstream (MEMORY.md shorten + duplicate-header) Architectural cascade now legible: Hawkins (biological) → Kanerva/Aiken/Hanrahan/Olukotun (Stanford computational bridge) → NVIDIA (silicon) → TigerBeetle/Antithesis (DST) → Aaron's lifetime optimization → Zeta multi-agent factory. Empirically validated by Aaron's outpace-11-AI-critics performance. Cron alive (7eec3da9). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(hygiene): add blank lines in 1451Z tick Satisfy markdownlint MD032 by separating the bold PR labels from the numbered lists they introduce. Co-Authored-By: Codex <noreply@openai.com> * fix(hygiene): reconcile 1451Z tick counts Align the 1451Z tick summary, substrate landing heading, cleanup table count, and cron-state heading with the enumerated body so the shard is internally auditable. Co-Authored-By: Codex <noreply@openai.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> Co-authored-by: Codex <noreply@openai.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.
Summary
Three fixes triggered by Aaron observing the dashboard "50 PRs Merged" stat was misleading:
per_page=50cap meant 68 actual PRs displayed as 50. Fixed viaper_page=100+ two-page fetch (up to 200).Plus active-trajectory #12 added: Background-loop productivity uplift as ongoing concern (origin: catch 43, ~10-15% of orchestrated rate when foreground absent).
Aaron 2026-05-12: "lies damn lies and statistics" — the number was lying, not the work.
🤖 Generated with Claude Code