feat(processed-audio-sources): phase 4 - runtime filter integration

Add AudioFilterPipeline for chained filter execution on AudioAnalysis.
Wire filter pipelines into AudioColorStripStream, AudioValueStream,
and WebSocket test endpoint. Add hot-update support via
ProcessorManager.refresh_audio_filter_pipelines(). Thread
AudioProcessingTemplateStore through dependency injection hierarchy.

server/src/wled_controller/api/routes/audio_processing_templates.py

@@ -6,6 +6,7 @@ from wled_controller.api.auth import AuthRequired
 from wled_controller.api.dependencies import (
     fire_entity_event,
     get_audio_processing_template_store,
+    get_processor_manager,
 )
 from wled_controller.api.schemas.audio_processing import (
     AudioProcessingTemplateCreate,
@@ -129,6 +130,12 @@ async def update_audio_processing_template(
             tags=data.tags,
         )
         fire_entity_event("audio_processing_template", "updated", template_id)
+        # Hot-update: rebuild filter pipelines for running streams using this template
+        try:
+            pm = get_processor_manager()
+            pm.refresh_audio_filter_pipelines(template_id)
+        except Exception:
+            pass  # Non-critical: streams will pick up changes on next restart
         return _apt_to_response(template)
     except EntityNotFoundError as e:
         raise HTTPException(status_code=404, detail=str(e))
@@ -154,6 +161,12 @@ async def delete_audio_processing_template(
         # TODO: Phase 3 will add reference checks against ProcessedAudioSource
         store.delete_template(template_id)
         fire_entity_event("audio_processing_template", "deleted", template_id)
+        # Hot-update: rebuild filter pipelines for running streams that used this template
+        try:
+            pm = get_processor_manager()
+            pm.refresh_audio_filter_pipelines(template_id)
+        except Exception:
+            pass  # Non-critical
     except HTTPException:
         raise
     except EntityNotFoundError as e:

server/src/wled_controller/api/routes/audio_sources.py

@@ -9,6 +9,7 @@ from starlette.websockets import WebSocket, WebSocketDisconnect
 from wled_controller.api.auth import AuthRequired
 from wled_controller.api.dependencies import (
     fire_entity_event,
+    get_audio_processing_template_store,
     get_audio_source_store,
     get_audio_template_store,
     get_color_strip_store,
@@ -210,11 +211,12 @@ async def test_audio_source_ws(
     ManagedAudioStream (ref-counted — shares with running targets), and streams
     AudioAnalysis snapshots as JSON at ~20 Hz.
 
-    NOTE: Audio processing filters from the template chain are NOT applied in
-    this WebSocket yet — that will be wired in Phase 4 when the stream runtime
-    integrates filter instances.
+    Audio processing filters from the template chain are applied to the
+    analysis before sending, so the WebSocket output matches what running
+    streams see.
     """
     from wled_controller.api.auth import verify_ws_token
+    from wled_controller.core.audio.filters.pipeline import build_pipeline_from_template_ids
 
     if not verify_ws_token(token):
         await websocket.close(code=4001, reason="Unauthorized")
@@ -223,6 +225,7 @@ async def test_audio_source_ws(
     # Resolve source → device info + processing template chain
     store = get_audio_source_store()
     template_store = get_audio_template_store()
+    apt_store = get_audio_processing_template_store()
     manager = get_processor_manager()
 
     try:
@@ -247,6 +250,15 @@ async def test_audio_source_ws(
             logger.debug("Audio template not found, falling back to best available engine: %s", e)
             pass  # Fall back to best available engine
 
+    # Build filter pipeline from processing template chain
+    pipeline = None
+    if resolved.audio_processing_template_ids and apt_store:
+        pipeline = build_pipeline_from_template_ids(
+            resolved.audio_processing_template_ids, apt_store
+        )
+        if pipeline.empty:
+            pipeline = None
+
     # Acquire shared audio stream
     audio_mgr = manager.audio_capture_manager
     try:
@@ -265,7 +277,10 @@ async def test_audio_source_ws(
             if analysis is not None and analysis.timestamp != last_ts:
                 last_ts = analysis.timestamp
 
-                # Send raw analysis — filter processing will be added in Phase 4
+                # Apply filter pipeline (channel extract, band extract, gain, etc.)
+                if pipeline is not None:
+                    analysis = pipeline.process(analysis)
+
                 await websocket.send_json(
                     {
                         "spectrum": analysis.spectrum.tolist(),
@@ -283,5 +298,7 @@ async def test_audio_source_ws(
     except Exception as e:
         logger.error(f"Audio test WebSocket error for {source_id}: {e}")
     finally:
+        if pipeline is not None:
+            pipeline.close()
         audio_mgr.release(device_index, is_loopback, engine_type)
         logger.info(f"Audio test WebSocket disconnected for source {source_id}")