feat: move analysis cache to recordings/analyses/, prune orphans on startup

- Cache files now live in recordings/analyses/<filename>.analysis.json
  (mirroring the relative path for files in subdirectories) rather than
  alongside each audio file.
- _api_delete now removes the corresponding cache file after deleting audio.
- prune_orphan_analyses() runs at startup and removes any .analysis.json
  whose audio file no longer exists (handles files deleted outside the UI).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

README.md

@@ -159,7 +159,7 @@ Shows recordings grouped by day with collapsible sections. Features:
 - **Day groups** — recordings are grouped under a collapsible day heading showing date, file count, total duration, and total size. The most recent day is expanded by default; older days start collapsed. Expanded state is preserved across filter changes.
 - **Day highlights** — click **★ Highlights** on any day heading to run loudness analysis across all WAV/FLAC files in that day and display a combined activity timeline SVG. Orange segments show when loud sections occurred relative to the day's time span; blue shows the file extents. Labels show the start, midpoint, and end times.
 - **Inline playback** — collapsible `▶ Play` button per row; audio loads lazily via a seekable `/stream/` endpoint with HTTP Range support. Metadata is fetched immediately so the duration is visible without pressing play.
-- **Waveform analysis** — on demand per file; computes RMS per 100 ms window and highlights loud sections. Supported for WAV and FLAC (FLAC requires `numpy` + `soundfile`). Pure-Python fallback for WAV when numpy is absent.
+- **Waveform analysis** — on demand per file; computes RMS per 100 ms window and highlights loud sections. Supported for WAV and FLAC (FLAC requires `numpy` + `soundfile`). Pure-Python fallback for WAV when numpy is absent. Results are cached in `recordings/analyses/<filename>.analysis.json`; subsequent requests at the same threshold and min-gap settings return instantly without re-reading the audio. The cache file is deleted automatically when the audio file is deleted. Orphaned cache files (audio deleted outside the UI) are pruned on startup.
 - **Grace period** — configurable in the controls bar (default 2 s). Loud sections separated by less than this gap are merged into one. Raise this (e.g. to 15–30 s) when a single event generates many timestamps due to brief quiet gaps within it.
 - **Timestamp jump** — after analysis, click any loud-section chip to seek the player to that position and pre-fill the cut panel. Use **J** / **K** keyboard shortcuts to jump to the previous / next section while audio is playing.
 - **Cut & download** — `✂ Cut` button opens the player row and reveals a cut panel. Enter start and end times in `m:ss` or `h:mm:ss` format and click **↓ Download cut** to receive an ffmpeg-trimmed copy without re-encoding. Requires ffmpeg (included in the Docker image).

ROADMAP.md

@@ -1,154 +0,0 @@
-# ISR Roadmap
-
-## notify.py — NTFY Loudness Notifications
-
-### Context
-
-Street ambience recorder. Goal: detect notable audio events (speech, thunder,
-sustained unusual sounds) in hourly recording files and push a notification via
-a self-hosted NTFY server. Generic short events (car horn, passing vehicle)
-should be filtered out by a minimum section duration.
-
-### Design decisions
-
-| Topic | Decision |
-|---|---|
-| Detection | RMS + minimum section duration filter (KISS — no FFT for now) |
-| Timing | Configurable: `immediate` / `daily` / `both` |
-| Config | `[notify]` section in existing `config.ini` |
-| Code structure | `notify.py` imports `analyze_wav` / `analyze_flac` from `web.py` (DRY) |
-| Source name | Included in notification body; configurable display name per source |
-
----
-
-### Config additions (`config.example.ini`)
-
-Add a `[notify]` section to `config.ini`:
-
-```ini
-[notify]
-enabled              = true
-ntfy_url             = https://ntfy.example.com/mytopic   ; full URL incl. topic
-mode                 = immediate   ; immediate | daily | both
-daily_time           = 08:00       ; HH:MM — used in daily and both modes
-debounce_minutes     = 60          ; immediate mode: suppress repeat notifications within this window
-min_section_duration = 2.0         ; seconds — sections shorter than this are ignored (filters car horns etc.)
-min_sections         = 1           ; number of qualifying sections required to trigger a notification
-loudness_threshold   = 0.05        ; RMS 0–1, same scale as web.py analysis threshold
-```
-
-Per recording source, add an optional `display_name`:
-
-```ini
-[radio1]
-type         = stream
-url          = http://icecast.example.com:8000/live
-display_name = Street mic north   ; shown in notification; defaults to section name [radio1]
-```
-
----
-
-### Notification format
-
-```
-Title:  ISR — Notable audio · Street mic north
-Body:   radio1_20260427_0300.wav
-        3 notable sections (≥ 2.0 s each)
-        → 00:12 – 00:18
-        → 01:45 – 01:52
-        → 47:03 – 47:11
-        Peak RMS: 0.312
-```
-
-Daily digest example:
-
-```
-Title:  ISR Daily Digest · 2026-04-27
-Body:   Street mic north — 4 files with notable events
-        03:00 file · 3 sections (peak 0.312)
-        07:00 file · 1 section  (peak 0.091)
-        14:00 file · 2 sections (peak 0.204)
-        21:00 file · 1 section  (peak 0.178)
-```
-
----
-
-### Implementation plan
-
-#### Phase 1 — Core
-
-1. **`config.example.ini`** — add `[notify]` section and `display_name` key to
-   source section examples (as shown above).
-
-2. **`notify.py` — file watcher**
-   - Polls `recordings/status.json` every 30 s.
-   - Tracks which files were in `active` on the previous poll.
-   - When a file disappears from `active` it was just closed → queues it for
-     analysis.
-   - Skips files with extensions that cannot be analysed (anything other than
-     `.wav` / `.flac`).
-
-3. **`notify.py` — analysis + filter**
-   - Imports `analyze_wav` / `analyze_flac` from `web.py`.
-   - Applies `loudness_threshold` from `[notify]` config.
-   - Filters resulting sections to those with duration ≥ `min_section_duration`.
-   - Counts filtered sections against `min_sections` threshold.
-
-4. **`notify.py` — NTFY HTTP POST**
-   - Plain `urllib` POST to `ntfy_url` (no extra dependencies).
-   - Sets `Title` and message body as described above.
-   - Logs success / failure to stdout.
-
-#### Phase 2 — Cadence modes
-
-5. **Immediate mode with debounce**
-   - Fires right after the file closes and analysis passes.
-   - Persists last-notification timestamp per source to a small
-     `notify_state.json` in the recordings directory.
-   - Suppresses sending if last notification for that source was within
-     `debounce_minutes`.
-
-6. **Daily digest mode**
-   - Appends qualifying events to `notify_log.jsonl` in the recordings
-     directory (one JSON line per event: timestamp, source, filename, sections,
-     peak RMS).
-   - On each poll checks whether `daily_time` has passed today and no digest
-     has been sent yet (tracked in `notify_state.json`).
-   - Reads all undigested entries from `notify_log.jsonl`, groups by
-     `display_name`, sends one notification per source with notable activity.
-   - Marks entries as digested.
-
-7. **Both mode**
-   - Immediate path: only fires when peak RMS exceeds a second, higher
-     threshold (`alarm_threshold`, default `0.3`; add to `[notify]` config).
-   - Daily digest path: fires for everything that passes `min_sections`.
-
-#### Phase 3 — Integration
-
-8. **Docker** — optional `notify` service in `docker-compose.yml`:
-   ```yaml
-   notify:
-     build: .
-     command: python notify.py
-     volumes:
-       - ./recordings:/app/recordings
-       - ./config.ini:/app/config.ini:ro
-     restart: unless-stopped
-   ```
-
-9. **README** — new section documenting `notify.py` usage, config keys, and
-   Docker setup.
-
----
-
-### Open questions (decide before implementing)
-
-- **Log rotation**: `notify_log.jsonl` grows indefinitely. Options: cap at N
-  days (configurable), cap at N MB, or leave cleanup to the user. No decision
-  made yet.
-- **Multiple NTFY topics per source**: current design uses one global topic.
-  If per-source topics are ever needed, `ntfy_url` could be moved to the source
-  section and override the global one.
-- **FFT / frequency analysis** (future): distinguishing thunder (low rumble,
-  50–200 Hz) from speech (300–3000 Hz) from vehicles would reduce false
-  positives further. Deferred — requires `numpy` and adds meaningful complexity.

web.py

@@ -203,6 +203,33 @@ def analyze_flac(path: Path, window_samples: int = WINDOW_SAMPLES,
     return _package_result(rms_values, framerate, n_frames, window_samples, threshold, min_gap)
 
 
+# ---------------------------------------------------------------------------
+# Analysis cache helpers
+# ---------------------------------------------------------------------------
+
+def _analysis_cache_path(base: Path, audio_path: Path) -> Path:
+    rel = audio_path.relative_to(base)
+    return base / 'analyses' / rel.parent / (rel.name + '.analysis.json')
+
+
+def prune_orphan_analyses(base: Path):
+    analyses_dir = base / 'analyses'
+    if not analyses_dir.exists():
+        return
+    removed = 0
+    for cache in analyses_dir.rglob('*.analysis.json'):
+        rel = cache.relative_to(analyses_dir)
+        audio_path = base / rel.parent / rel.name[:-len('.analysis.json')]
+        if not audio_path.exists():
+            try:
+                cache.unlink()
+                removed += 1
+            except Exception:
+                pass
+    if removed:
+        print(f'Pruned {removed} orphaned analysis cache file(s)')
+
+
 # ---------------------------------------------------------------------------
 # File listing
 # ---------------------------------------------------------------------------
@@ -331,6 +358,16 @@ class _Handler(BaseHTTPRequestHandler):
         except Exception:
             pass
 
+        base = Path(self.recordings_dir).resolve()
+        cache_path = _analysis_cache_path(base, path)
+        try:
+            cached = json.loads(cache_path.read_text('utf-8'))
+            if cached.get('threshold') == threshold and cached.get('min_gap') == min_gap:
+                self._send(200, json.dumps(cached['result']).encode('utf-8'), 'application/json')
+                return
+        except Exception:
+            pass
+
         ext = path.suffix.lower()
         if ext == '.wav':
             result = analyze_wav(path, threshold=threshold, min_gap=min_gap)
@@ -343,6 +380,14 @@ class _Handler(BaseHTTPRequestHandler):
             self._json_err(400, f'Loudness analysis is not available for {ext} files')
             return
 
+        try:
+            cache_path.parent.mkdir(parents=True, exist_ok=True)
+            tmp = cache_path.with_suffix('.tmp')
+            tmp.write_text(json.dumps({'threshold': threshold, 'min_gap': min_gap, 'result': result}), 'utf-8')
+            os.replace(tmp, cache_path)
+        except Exception:
+            pass
+
         self._send(200, json.dumps(result).encode('utf-8'), 'application/json')
 
     def _api_status(self):
@@ -465,6 +510,11 @@ class _Handler(BaseHTTPRequestHandler):
             self._json_err(500, f'Failed to delete: {e}')
             return
 
+        try:
+            _analysis_cache_path(Path(self.recordings_dir).resolve(), path).unlink()
+        except Exception:
+            pass
+
         self._send(200, json.dumps({'deleted': filename}).encode(), 'application/json')
 
     def _api_cut(self, qs):
@@ -876,7 +926,7 @@ function seekToSection(idx, filename, startSec, endSec, sectionIdx) {
   activePlayerIdx = idx;
   const audio  = document.getElementById('aud-'+idx);
   const seekTo = Math.max(0, startSec - getPreroll());
-  const doSeek = () => { audio.currentTime = seekTo; };
+  const doSeek = () => { audio.currentTime = seekTo; audio.play().catch(() => {}); };
   if (audio.readyState >= 1) doSeek();
   else audio.addEventListener('loadedmetadata', doSeek, {once: true});
   setCutFields(idx, startSec, endSec);
@@ -1449,7 +1499,7 @@ function jumpToDaySection(si) {
   const audio = document.getElementById('aud-' + fileIdx);
   if (!audio) return;
   const seekTo = Math.max(0, start - getPreroll());
-  const doSeek = () => { audio.currentTime = seekTo; };
+  const doSeek = () => { audio.currentTime = seekTo; audio.play().catch(() => {}); };
   if (audio.readyState >= 1) doSeek();
   else audio.addEventListener('loadedmetadata', doSeek, { once: true });
 
@@ -1551,6 +1601,8 @@ def main():
     if not rec_dir.exists():
         print(f"Warning: recordings directory '{args.dir}' does not exist yet.")
 
+    prune_orphan_analyses(rec_dir.resolve())
+
     class Handler(_Handler):
         recordings_dir = str(rec_dir.resolve())
         threshold      = args.threshold