Choose the destination clip before transitioning, play the matching
member->member morph (154 clips, all adjacent video pairs, both
directions), chain through each crossed altitude, and lock the clip
until the altitude is changed (no secondary swap, no re-roll).
Session 0024 (writing-plans). Stops at the review gate.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The 'no sound' debugging is resolved (stale-server resilience). Move the audio
status readout + native test player into the Dev Mode panel — kept for future
'can't hear' debugging, out of the installation UI. 7 E2E pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The operator's video wouldn't turn on: their stale uvicorn (old Python) requires a
7-way 'content' and 422s the {visual,audio} payload, and returns {content:{video}}
not {render:{video:{shown}}}. So the alteration POST failed and video never updated
(audio works because it's client-driven). Fix: controls() also sends a back-compat
'content' field (current server ignores it; old one ignores visual/audio), and
update() reads video-shown from either response shape. Verified end-to-end against a
simulated fully-stale server: video shows + audio plays, zero errors.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Root cause of 'no sound', found via the live readout: the operator's /api/ring
returned the ring WITHOUT the per-scale audio field (a server process predating the
audio code) → soundtrackUrl() was null → the element never got a src (readyState 0,
no error). The native test player proved the file itself plays. Fix: client falls
back to a scale-id→soundtrack map (mirrors build_pool_manifest.SCALE_AUDIO) so audio
works without a server restart. Verified: stale-ring sim now plays; audio-first-then-
video leaves both on. +2 E2E regression tests (7 total). Chromium + WebKit clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The operator's 'PAUSED readyState 0' (no error) means soundtrackUrl() returned null
— the running app's /api/ring has no audio for the scale, the signature of a STALE
uvicorn server (Python predates the audio code; static reload doesn't restart it).
Status now shows url=<…|NONE>, ring=server|SYNTH, scale id so it's unambiguous.
Fresh server verified: ▶ playing rs=4.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Three blind fixes failed because headless browsers relax autoplay (can't reproduce)
and play() errors were swallowed. Instrument instead:
- #audio-status: live readout (playing/paused/blocked/media-error) under the toggle
- surface the play() rejection reason + element error (was .catch(()=>{}))
- 'Audio check' fieldset with a native <audio controls> on the soundtrack file, to
test the file directly with zero app code in the path
Diagnostic only — no behavior change. 37 tests pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Removes the tap-to-start wall (Loading Universe splash instead), fixes audio still
silent on real Safari (single <audio> played synchronously in the toggle gesture),
both toggles default off, first Video-on couples Audio. Verified Chromium + WebKit.
290 tests + 5 Playwright E2E pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
# Altitude Clip-Pair Morph + Lock Implementation Plan
> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
**Goal:** When the viewer changes altitude, choose the destination clip *first*, play the morph footage that goes from the currently-shown clip to that chosen clip, then lock the clip in place so it never changes again until the altitude changes.
**Architecture:** Today the ring carries **5** per-edge morphs (each `scale_a → scale_b`, primary→primary), the server picks a random pool member only for the *landed* scale, and `advance()` plays the edge morph (which ends on the scale's *primary*) then does a *second*, jarring `fadeThroughBlack` swap to the randomly-chosen member. We replace the per-edge model with a **per-clip-pair** model: every pair of clips in adjacent altitudes gets its own morph, in both directions (**154** clips total). The server picks the destination member(s) *before* selecting the morph, threads the currently-shown clip in as the source, and chains morphs through each crossed altitude on a multi-detent jump. The renderer plays those exact morphs and locks on the final clip — no secondary swap, no re-roll while parked.
**Tech Stack:** Python 3.13 / FastAPI (`simulator/app.py`), pure ring logic (`player/ring.py`), pure manifest/schema (`simulator/clips.py`), media baking via ffmpeg `xfade=zoomin` (`simulator/build_pool_manifest.py`), vanilla JS renderer (`simulator/static/app.js`), pytest, and a new Playwright E2E tier.
## Global Constraints
- **Morph file naming:** `transitions/<src_clip_id>__<dst_clip_id>.mp4` (double-underscore separator — clip ids themselves use single `_`, e.g. `cosmos_miri`). The zoom-out companion is `<src>__<dst>.rev.mp4` and represents the **reverse** direction (`dst → src`).
- **Direction convention:** a morph is *baked* descending (higher altitude → lower altitude = zoom IN). The `.rev.mp4` companion is the ascending (zoom OUT) playback. The manifest lists **both** directions as explicit, separately-keyed entries so the renderer needs no reverse logic at play time.
- **Ring order (descending, wraps):** `["cosmos", "orbit", "coast", "reef", "abyss"]`; `abyss → cosmos` is the wrap edge.
- **Lock invariant:** the displayed base clip changes **only** inside `advance()` (a deliberate altitude change) and the one initial `landScale()`. Nothing else may reload base media. The Dev-Mode re-roll button stays as an explicit dev affordance.
- **Randomness stays injected:** `player/ring.py` is a pure module; the impure `random.random()` draw happens once at the API boundary (`simulator/app.py`). Pure functions take injected `float` picks in `[0,1)`.
- **Back-compat:** a manifest with no `pool` (pool of one) and the synthesized single-clip fallback ring must still work. A pool-of-one scale needs no morphs to itself.
Re-key the manifest builder from 5 per-edge entries to 154 per-clip-pair entries (both directions), each carrying `from`/`to` clip ids. Pure dict assembly — no ffmpeg.
Generalize the ffmpeg morph recipe from scale-primary→scale-primary to any member→member, and have `generate_media()` bake every adjacent member pair forward + reverse. Tested with a stubbed ffmpeg runner so the test is fast and deterministic (the real bake runs as a media step at execution time).
Change the `Transition` dataclass and ring loader to carry `from_clip`/`to_clip`, build a `(from, to) → file` lookup on the ring, and update the structural invariant. Touches both `player/ring.py` (dataclass + ring) and `simulator/clips.py` (parse + serialize).
- Produces: `ScaleRing.morph_for(from_clip: str, to_clip: str) -> str | None` — the morph file for that directed pair, or `None` if absent.
-`ScaleRing.__post_init__`: drop the "N edges == N transitions" rule (no longer per-edge); keep "≥1 scale, single-scale ring has no transitions". Build the lookup dict in `__post_init__` (store on a private attr via `object.__setattr__`, since the dataclass is frozen).
-`ring_to_dict` transitions become `[{"from","to","file","model"}, ...]`.
Expected: PASS — fix any existing test that constructed `Transition(file=..., model=...)` positionally to the new 3-arg form (`Transition(from_clip, to_clip, file)`), and any test asserting the old `N == N` invariant.
### Task 4: Resolve chained member→member morphs (pick before transition)
Add a pure function that, given a structural `RingMove`, the currently-shown clip, and one injected pick per step, walks the steps — choosing a destination member for each crossed altitude and resolving the morph from the *previous* clip to it — and reports the final locked clip. This is the heart of "choose the clip first, then play the matching morph," chained through each altitude.
-`resolve_move(ring: ScaleRing, move: RingMove, from_clip_id: str, picks: tuple[float, ...]) -> ResolvedMove` — `len(picks)` must equal `len(move.steps)`; each `picks[k]` chooses the member of the scale that step `k` lands on. `from_clip` of step 0 is `from_clip_id`; each subsequent step's `from_clip` is the prior step's `to_clip`. `target_clip_id` is the last step's `to_clip` (or `from_clip_id` for an empty move).
### Task 5: Fast spin plays the whole chain quickly (no collapse)
The operator chose "chain through each altitude," with fast spins playing that chain *quickly* rather than collapsing to a single arrival pass. Change `advance_ring`'s fast-spin branch to keep every step but mark each `blended=True` (so the renderer plays each morph at `FAST_BLEND_RATE`) and still set `RingMove.fast=True`.
**Files:**
- Modify: `player/ring.py:219-241` (fast-spin branch in `advance_ring`)
- Test: `tests/test_player_ring.py`
**Interfaces:**
- Unchanged signature. Behavior change: a fast spin returns **all** steps (each `blended=True`), not one collapsed step.
- [ ]**Step 1: Write the failing test (and update any existing fast-spin assertions)**
- [ ]**Step 4: Run the full ring suite (catch the old collapse test)**
Run: `pytest tests/test_player_ring.py -v`
Expected: PASS — update/replace any prior test that asserted fast spin yields exactly one step.
- [ ]**Step 5: Commit**
```bash
git add player/ring.py tests/test_player_ring.py
git commit -m "feat(ring): fast spin plays the full morph chain quickly instead of collapsing"
```
---
### Task 6: API picks before transitioning and serializes the resolved chain
`/api/ring/advance` accepts the currently-shown clip (`from_clip_id`), draws one random pick per step, resolves the chained morphs, and returns the resolved steps + locked `target_clip_id`. Also extend `media-versions` to cover every transition file.
### Task 7: Renderer plays the chosen morph and locks the clip (no secondary swap)
Rewire the frontend: send the currently-shown clip, play each resolved morph straight (the manifest already encodes direction, so drop the `reversed`/`reverseFile` logic), land + lock on `target_clip_id` with a plain (re)load, and **delete the secondary `fadeThroughBlack` swap**. Build a `(from,to)→file` lookup from the ring and preload only the morphs reachable from the current locked clip, refreshed on each landing (154 clips is too many to preload eagerly).
- New module state: `morphByPair = {}` — `"from→to"` → file string, built from `ring.transitions`.
- [ ] **Step 1: Build the morph lookup when the ring loads**
```javascript
// simulator/static/app.js — in loadRing(), after `ring = ...` is set:
morphByPair = {};
if (ring && ring.transitions)
for (const t of ring.transitions) morphByPair[`${t.from}→${t.to}`] = t.file;
```
Add `let morphByPair = {};` near the other module state (around line 38). The synthesized single-clip fallback ring has no transitions → empty lookup → advance is a no-op (pool of one), which is correct.
- [ ] **Step 2: Simplify `playTransition` (no reverse logic)**
activeClipId = move.target_clip_id; // the locked clip — the morph ended on it
currentClipId = null; // force ensureClipMedia to (re)load + loop it
renderScaleReadout();
} finally {
busy = false;
update(); // ensureClipMedia loads the locked clip; nothing re-rolls
applyAudio();
refreshReachablePreload();
}
}
```
Then **delete** `fadeThroughBlack` (now unused) and `reverseFile` (the `.rev` companion is an explicit manifest entry, no longer derived). Grep to confirm no other callers before deleting.
- [ ] **Step 4: Preload only the reachable morphs, refreshed per landing**
```javascript
// simulator/static/app.js — replace the eager transition preload in preloadOrder()/preloadList()
// Reachable morphs from the CURRENT locked clip: to/from every member of the
// neighboring altitudes (the only clips one detent can reach). 154 total is too
// many to preload up front; we focus on what the next move can need.
function reachableMorphFiles() {
const files = new Set();
if (!ring || ring.scales.length < 2 || !activeClipId) return files;
const n = ring.scales.length;
for (const sign of [1, -1]) {
const nb = ring.scales[((ringIndex + sign) % n + n) % n];
// Kick a background cache fill of the reachable morphs (bounded concurrency,
// same cacheClip helper used for bases). Idempotent — cached files are skipped.
for (const f of reachableMorphFiles()) cacheClip(f);
}
```
In `preloadOrder()` replace the `if (ring.transitions) for (const t ...) { add(t.file); add(reverseFile(t.file)); }` block with `for (const f of reachableMorphFiles()) add(f);`. Keep base-clip preloading (current + neighbor pools) as-is. (`cacheClip` = factor out the per-file fetch→blob step already inside the preload loop so `refreshReachablePreload` can reuse it.)
- [ ] **Step 5: Manual smoke + commit**
Run: `cd simulator && python -m uvicorn app:app --port 8011` then open `http://localhost:8011/`, turn the Altitude dial one detent each way and several detents; confirm: a zoom morph plays, it lands on a clip, and that clip **does not change** while parked (watch the Dev-Mode clip readout). No black-flash second swap.
```bash
git add simulator/static/app.js
git commit -m "feat(sim): play chosen morph then lock clip per altitude; drop secondary swap"
Stand up a minimal Playwright tier (none exists) and add one spec that proves the two behaviors the operator asked for: after zooming, the displayed clip is stable (locked) until the next altitude change, and the morph that plays matches `from→to`. §9 requires an E2E browser tier for this UI change.
- Create: `simulator/e2e/README.md` (how to run; how it boots uvicorn)
**Interfaces:**
- The spec drives the real app (uvicorn on a fixed port via Playwright `webServer`), reads the Dev-Mode clip readout (enable Dev Mode via localStorage before load), and intercepts `/media/transitions/*` requests to assert the morph file played.
- [ ] **Step 1: Scaffold the tier (failing because no test yet)**
### Task 9: Regenerate sample media + manifest, full suite green
Bake the 154 morphs into `simulator/sample_media`, rebuild the manifest, and run the entire pytest suite + the E2E spec so the shipped sample media matches the new schema.
- "Choose which clip plays in the next altitude *before* transitioning" → Task 4 (`resolve_move` picks dst then morph) + Task 6 (server draws picks before resolving).
- "Play the appropriate morph footage" → Task 7 (renderer plays the resolved `step.file`).
- "Morph footage for all transitions between all videos in adjacent altitudes" → Tasks 1–2 (manifest + bake 154, both directions) + Task 9 (regenerate).
- "Video must not change after we've zoomed to a new altitude" / "lock per altitude until the altitude is changed" → Task 7 (delete secondary swap; lock `activeClipId`; reload only on altitude change) + Task 8 (E2E asserts stability while parked).
- "Good transition for every video combination, in and out" → both directions baked (Task 1/2), reverse via explicit `.rev` entries (Task 3/4), E2E covers in + out (extend Task 8 spec with a zoom-out case if desired).
- Multi-altitude jump chains through each altitude → Task 4 (`resolve_move` threads src→dst per step) + Task 5 (fast spin plays the chain quickly).
**Placeholder scan:** No TBD/TODO/"handle edge cases" — every code step shows real code; tests show real assertions.
**Type consistency:** `Transition(from_clip, to_clip, file, model)` (T3) used by `morph_for` (T3), `load_ring` (T3), `resolve_move` (T4), `resolved_move_to_dict` (T6). `MorphStep`/`ResolvedMove` fields (T4) match the serializer (T6) and the renderer's `step.{from_clip,to_clip,file,to_index,blended}` + `move.target_clip_id` (T7). `morphByPair` key `"from→to"` consistent in build (T7 S1) and lookup (T7 S4).
**Open items (resolved at the review gate by the operator):**
1. **Storage/repo weight:** ✅ Commit the 154 short mp4s into `sample_media` (extends the existing precedent of committing the 5 + bases).
2. **`build_pool_manifest.py` CLI:** ✅ Approach approved; confirm the actual entrypoint flag during execution (Task 9 Step 1).
- Video + Audio are simple on/off toggles (Dev-Mode style). Audio on = the current
altitude's soundtrack. The two real-browser bugs should now be fixed — but please
confirm on a real device (especially iPad/Safari).
- Deferred: white-noise (easy to re-add as a 3rd Audio position), the per-altitude
music layer, the networked-control-surface build, Pi renderer.
## Next /goal
```
/goal by-ear/by-eye review the audio + Video/Audio toggles on a REAL device (esp. iPad/Safari — headless can't validate Safari autoplay or GPU compositing), then source/compose the deferred per-altitude music layer per docs/audio-candidate-pool.md (or re-enable white-noise as a 3rd Audio position if wanted)
- If sound is STILL silent on the real device, that points to something beyond autoplay
(codec/MIME, file, or element error) — next session should capture `aud.error` and the
exact browser.
## Next /goal
```
/goal confirm on a REAL device (esp. iPad/Safari) that audio is now audible when Video/Audio is toggled on and the "Loading Universe…" → app flow feels right; if still silent, capture aud.error + the exact browser. Then the deferred per-altitude music layer per docs/audio-candidate-pool.md
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.