feat: PyAutoEyes dashboard content, critique route, Brain board chip, conductor --instance (PyAutoEyes 2)

[Jammy2211]

Overview

Phase 2 of the PyAutoEyes birth epic (pyautoeyes-birth). Phase 1b (#448) gave the organ its skeleton: registry.yaml, the manifest reader, a dashboard with thumbnails and /eyes review lines. This phase fills the dashboard with the survey and critique content, gives every figure a route for suggesting an improvement, puts an Eyes chip on the Brain board family footer, and teaches the Brain Eyes conductor to resolve instances by name through the organ's registry.

Plan

• Dashboard content: per instance, the figure count, the conductor survey (stale renders, gaps, orphans) when a local checkout exists, freshness against PyPI, and the open critiques (Mind drafts that mention the instance). Also a badge.json.
• Critique route per figure: the copyable /eyes review <instance> <figure> line, plus a pre-filled GitHub "new issue" link on the project repo (title figure: <domain>/<file>, body with the raw PNG link and a "Suggested improvement:" stub, label eyes-critique). The dashboard files nothing itself.
• Brain board chip: boards.eyes: PyAutoEyes in canonical organ order, an ORGANS["eyes"] palette and MARKS["eyes"] glyph, and a collect_eyes() strip that reads PyAutoEyes' dashboard.md counts.
• Eyes conductor: --instance <name> resolved through PyAutoEyes registry.yaml. Given the organ root, survey/review cover every registered instance. The conductor stays repo-name-free. The recursive-producer bug is left to its own draft.
• Hermetic tests for every new piece, ruff clean, and prose updated (PyAutoEyes README/AGENTS/REFERENCE, Brain eyes AGENTS.md and skills/eyes/eyes.md).

Feature Agent note: the Agent suggested split-into-phases. The epic's declared phasing wins, since this prompt is already phase 2 of 5, so it ships as one task.

Detailed implementation plan

Work Classification

Library (organ code: PyAutoEyes + PyAutoBrain). No workspace follow-up.

Affected Repositories

• PyAutoEyes (primary)
• PyAutoBrain

Branch Survey

Repository	Current Branch	Dirty?
organs/PyAutoEyes	main	3 untracked leftover dirs (dataset/, output/, scripts/), not from this task and left untouched
organs/PyAutoBrain	main	clean

worktree_check_conflict shows no conflict. The #448 claim has already been removed by its close-out.

Suggested branch: feature/eyes-board-conductor-registry
Worktree root: ~/Code/PyAutoLabs-wt/eyes-board-conductor-registry/

Implementation Steps

1. PyAutoEyes eyes/board.py
   • InstanceView gains survey (parsed pyauto-brain eyes --json survey of the local checkout, or None with a note) and critiques (titles and paths of Mind drafts under draft/*/ that mention the instance repo or name).
   • The summary table gains Figures, Stale/Gaps/Orphans and Critiques columns.
   • Each figure row or card gets the review line plus an "Suggest improvement" link to https://github.com/<github>/issues/new?title=…&body=…&labels=eyes-critique. The body carries the raw PNG URL, the manifest generated date and rendered version, and a Suggested improvement: stub.
   • badge.json is written beside the dashboard. The render stays deterministic: the survey is summarised as counts and names only, never mtimes.
2. PyAutoEyes eyes/cli.py
   • board runs the survey (--no-survey opts out) and the critique grep (the Mind checkout comes from $PYAUTO_MIND or the workspace root).
   • pages_dashboard.yml also publishes badge.json.
3. Brain config/policy.yaml: boards: eyes: PyAutoEyes goes between memory and heart.
4. Brain board/_theme.py: ORGANS["eyes"], an amber palette with contrast of at least 4.5:1 and distinct from the Hands orange, and MARKS["eyes"], an eye glyph in line art inside a ring. Update the ordered-family assertions in tests/test_board_theme.py.
5. Brain board/_board.py: collect_eyes() reads the PyAutoEyes dashboard.md summary table (instances, figures, behind), and the md and html renders gain an "Eyes" line next to Cortex. Add eyes to SURFACE_KEYS in the tests.
6. Brain agents/conductors/eyes/_eyes.py
   • --instance <name> (repeatable) and --instances-from <organ root>. The PyAutoEyes checkout resolves via $PYAUTO_EYES, then the sibling, then _repo_paths.repo_path(root, "PyAutoEyes").
   • A stdlib parser reads registry.yaml. The instance path resolves from the workspace root (grouped path, then flat repo).
   • Given the organ root as the workspace argument, the conductor covers every registered instance.
   • The gallery check also looks at the registry's manifest path.
   • New hermetic tests use a fabricated organ and registry.
7. Prose: PyAutoEyes README/AGENTS/REFERENCE, plus Brain agents/conductors/eyes/AGENTS.md and skills/eyes/eyes.md, cover --instance and the critique route.
8. Label: create eyes-critique on PyAutoLabs/autolens_visualization.

Witness

• PyAutoEyes pytest tests, a live bin/pyauto-eyes check, and bin/pyauto-eyes board (show the dashboard.md head).
• pyauto-brain eyes survey --instance lens shows imaging 22 and interferometer 18 with no gaps.
• The full Brain suite outside the bundle env reports 0 failed.
• The Heart and Hands pytest tests -k footer runs pass against the Brain branch.

Key Files

• PyAutoEyes/eyes/board.py, eyes/cli.py, tests/, .github/workflows/pages_dashboard.yml
• PyAutoBrain/config/policy.yaml, board/_theme.py, board/_board.py, agents/conductors/eyes/_eyes.py, tests/test_board_theme.py, tests/test_board.py, tests/test_eyes_conductor.py

Original Prompt

Click to expand starting prompt

PyAutoEyes phase 2 — cross-project dashboard, critique route, Brain chip, conductor registry

Type: feature
Target: PyAutoEyes
Repos:

• PyAutoEyes
• PyAutoBrain
  Themes:
• visualization
• infrastructure
  Difficulty: large
  Autonomy: supervised
  Priority: high
  Lane: local-dev
  Status: draft
  Consequence: judge
  Witness: the PyAutoEyes Pages dashboard publishes with a thumbnail per lens figure linked to its raw PNG in autolens_visualization, a freshness line, and a working "suggest an improvement" affordance per figure; a render.yml run in autolens_visualization triggers dashboard_refresh.yml; the Brain board footer shows the Eyes chip; tests/test_board_theme.py green; pyauto-brain eyes survey --instance lens resolves through registry.yaml
  Review-minutes: 15
  Epic: pyautoeyes-birth
  Phase: 2
  Filed: 2026-09-25

Blocked on: none — phase 1b COMPLETE 2026-09-28 (PyAutoMind#448; record complete/2026/09/eyes-organ-skeleton.md).

Task

• Dashboard content (eyes/board.py → dashboard.md/.html, the epic
  ledger): per registered instance, a thumbnail grid of every figure in its
  tracked manifest, each thumbnail linked to the raw PNG in the project
  repo (the organ never copies figures), click-through to full size, grouped
  by domain/producer; per-instance freshness (rendered stack version vs the
  latest library release on PyPI); survey gaps and orphans (from
  pyauto-eyes survey); open critiques (Mind drafts carrying the instance's
  target); badge.json.
• Critique route. Each figure carries a "suggest an improvement"
  affordance that pre-fills a critique routed through intake: a copyable
  /eyes review <instance> <figure> command and/or a pre-filled GitHub issue
  link on the project repo (title + figure path + manifest version). The
  dashboard never files anything itself.
• Brain board chip: config/policy.yaml boards.eyes: PyAutoEyes;
  board/_theme.py ORGANS["eyes"] palette + MARKS["eyes"] glyph
  (tests/test_board_theme.py enforces); optional collect_eyes() strip in
  board/_board.py reading the dashboard's counts.
• Eyes conductor: --instance <name> resolving through the organ's
  registry.yaml (survey/review default to every registered instance
  when given the organ root); the conductor stays repo-name-free — the
  registry is data, not code. The non-recursive producer scan stays its own
  bug draft (draft/bug/pyautobrain/eyes_survey_recursive_producers.md).

[Jammy2211]

Library PRs Created

• feat: dashboard survey + critiques, per-figure critique route, badge.json (PyAutoEyes 2) PyAutoEyes#3: dashboard survey and critiques, the per-figure critique route (/eyes review line plus a pre-filled eyes-critique issue link), a head counts table, and badge.json. Merge first.
• feat: Eyes board chip + strip, eyes conductor --instance via the PyAutoEyes registry (PyAutoEyes 2) PyAutoBrain#429: the Eyes board chip (policy, palette, mark), the collect_eyes() strip, and the eyes conductor --instance through the PyAutoEyes registry.yaml.

Feature Agent note. The Agent returned split-into-phases. The epic's declared phasing wins, because this prompt is already phase 2 of 5, so it ships as one task.

Witness

• PyAutoEyes: pytest tests 65 passed and ruff is clean. A live bin/pyauto-eyes check returns check: OK (40 figures resolve). bin/pyauto-eyes board reports lens: survey 40 png · 0 gaps · 0 orphans · 0 stale; critiques 5.
• pyauto-brain eyes survey --instance lens shows imaging 22 png and interferometer 18 png, with no gaps, and the tracked manifest is present.
• The Brain full suite, run outside the bundle env, gives 1085 passed. The Heart and Hands -k footer runs against the Brain branch give 3 + 3 passed.
• The eyes-critique label was created on PyAutoLabs/autolens_visualization.

Heart at ship. YELLOW, and every reason is in the human-acknowledged set. There is also an extra STALE marker, "release validation incomplete: no rehearsal for current source", which concerns release validation only.

Next: human /prm, merging PyAutoEyes#3 first and then PyAutoBrain#429. No workspace impact.

[Jammy2211]

Shipped

PyAutoEyes phase 2 (epic pyautoeyes-birth) is merged:

• feat: dashboard survey + critiques, per-figure critique route, badge.json (PyAutoEyes 2) PyAutoEyes#3 (merge a2e1ee7): the dashboard now shows the Brain conductor's survey (PNGs, gaps, orphans, stale renders) and the open Mind critiques for each instance. Every figure gets a "Suggest an improvement" link that opens a pre-filled eyes-critique issue on the project repo. A counts table heads dashboard.md, badge.json is generated beside the dashboard, and readings carry forward so CI renders do not flap.
• feat: Eyes board chip + strip, eyes conductor --instance via the PyAutoEyes registry (PyAutoEyes 2) PyAutoBrain#429 (merge 79a8652): an Eyes board chip and palette/mark in every family footer, a collect_eyes() strip that reads the Eyes counts table, and pyauto-brain eyes survey|review --instance <name> resolved through the PyAutoEyes registry.yaml, with fan-out from the organ root.

Tests at ship: Eyes 65 passed, Brain 1085 passed. Live pyauto-eyes check OK (40 figures) and eyes survey --instance lens OK.

Not verified: the PyAutoEyes Pages Dashboard workflow failed on the merge commit (and on #2). configure-pages reports "Create Pages site failed: Resource not accessible by integration" because GitHub Pages is not enabled on PyAutoEyes. A human needs to enable Pages (Source: GitHub Actions), then re-run it. The post-merge dashboard_refresh.yml dispatch has not been run.

Pending release: PyAutoEyes#3, PyAutoBrain#429.