feat: enable daily co in selfhosted + only schedule tasks when necessary (#883)

* feat: enable daily co in selfhosted + only schedule tasks when necessary

* feat: refactor aws storage to be platform agnostic + add local pad tracking with slfhosted support

server/.env.example

@@ -86,6 +86,18 @@ LLM_API_KEY=not-needed
 ## Context size for summary generation (tokens)
 LLM_CONTEXT_WINDOW=16000
 
+## =======================================================
+## Audio Padding
+##
+## backends: local (in-process PyAV), modal (HTTP API client)
+## Default is "local" — no external service needed.
+## Set to "modal" when using Modal.com or self-hosted gpu/self_hosted/ container.
+## =======================================================
+#PADDING_BACKEND=local
+#PADDING_BACKEND=modal
+#PADDING_URL=https://xxxxx--reflector-padding-web.modal.run
+#PADDING_MODAL_API_KEY=xxxxx
+
 ## =======================================================
 ## Diarization
 ##
@@ -137,6 +149,10 @@ TRANSCRIPT_STORAGE_AWS_REGION=us-east-1
 #DAILYCO_STORAGE_AWS_ROLE_ARN=...  # IAM role ARN for Daily.co S3 access
 #DAILYCO_STORAGE_AWS_BUCKET_NAME=reflector-dailyco
 #DAILYCO_STORAGE_AWS_REGION=us-west-2
+# Worker credentials for reading/deleting from Daily's recording bucket
+# Required when transcript storage is separate from Daily's bucket (e.g., selfhosted with Garage)
+#DAILYCO_STORAGE_AWS_ACCESS_KEY_ID=your-aws-access-key
+#DAILYCO_STORAGE_AWS_SECRET_ACCESS_KEY=your-aws-secret-key
 
 ## Whereby (optional separate bucket)
 #WHEREBY_STORAGE_AWS_BUCKET_NAME=reflector-whereby

server/.env.selfhosted.example

@@ -47,6 +47,9 @@ DIARIZATION_URL=http://transcription:8000
 TRANSLATION_BACKEND=modal
 TRANSLATE_URL=http://transcription:8000
 
+PADDING_BACKEND=modal
+PADDING_URL=http://transcription:8000
+
 # HuggingFace token — optional, for gated models (e.g. pyannote).
 # Falls back to public S3 model bundle if not set.
 # HF_TOKEN=hf_xxxxx
@@ -93,15 +96,42 @@ TRANSCRIPT_STORAGE_AWS_REGION=us-east-1
 # =======================================================
 # Daily.co Live Rooms (Optional)
 # Enable real-time meeting rooms with Daily.co integration.
-# Requires a Daily.co account: https://www.daily.co/
+# Configure these BEFORE running setup-selfhosted.sh and the
+# script will auto-detect and start Hatchet workflow services.
+#
+# Prerequisites:
+#   1. Daily.co account: https://www.daily.co/
+#   2. API key: Dashboard → Developers → API Keys
+#   3. S3 bucket for recordings: https://docs.daily.co/guides/products/live-streaming-recording/storing-recordings-in-a-custom-s3-bucket
+#   4. IAM role ARN for Daily.co to write recordings to your bucket
+#
+# After configuring, run: ./scripts/setup-selfhosted.sh <your-flags>
+# The script will detect DAILY_API_KEY and automatically:
+#   - Start Hatchet workflow engine + CPU/LLM workers
+#   - Generate a Hatchet API token
+#   - Enable FEATURE_ROOMS in the frontend
 # =======================================================
-# DEFAULT_VIDEO_PLATFORM=daily
 # DAILY_API_KEY=your-daily-api-key
 # DAILY_SUBDOMAIN=your-subdomain
-# DAILY_WEBHOOK_SECRET=your-daily-webhook-secret
+# DEFAULT_VIDEO_PLATFORM=daily
 # DAILYCO_STORAGE_AWS_BUCKET_NAME=reflector-dailyco
 # DAILYCO_STORAGE_AWS_REGION=us-east-1
 # DAILYCO_STORAGE_AWS_ROLE_ARN=arn:aws:iam::role/DailyCoAccess
+# Worker credentials for reading/deleting from Daily's recording bucket
+# Required when transcript storage is separate from Daily's bucket (e.g., selfhosted with Garage)
+# DAILYCO_STORAGE_AWS_ACCESS_KEY_ID=your-aws-access-key
+# DAILYCO_STORAGE_AWS_SECRET_ACCESS_KEY=your-aws-secret-key
+# DAILY_WEBHOOK_SECRET=your-daily-webhook-secret  # optional, for faster recording discovery
+
+# =======================================================
+# Hatchet Workflow Engine (Auto-configured for Daily.co)
+# Required for Daily.co multitrack recording processing.
+# The setup script generates HATCHET_CLIENT_TOKEN automatically.
+# Do not set these manually unless you know what you're doing.
+# =======================================================
+# HATCHET_CLIENT_TOKEN=<auto-generated-by-script>
+# HATCHET_CLIENT_SERVER_URL=http://hatchet:8888
+# HATCHET_CLIENT_HOST_PORT=hatchet:7077
 
 # =======================================================
 # Feature Flags

server/reflector/hatchet/workflows/daily_multitrack_pipeline.py

@@ -90,7 +90,6 @@ from reflector.processors.summary.summary_builder import SummaryBuilder
 from reflector.processors.types import TitleSummary, Word
 from reflector.processors.types import Transcript as TranscriptType
 from reflector.settings import settings
-from reflector.storage.storage_aws import AwsStorage
 from reflector.utils.audio_constants import (
     PRESIGNED_URL_EXPIRATION_SECONDS,
     WAVEFORM_SEGMENTS,
@@ -117,6 +116,7 @@ class PipelineInput(BaseModel):
     bucket_name: NonEmptyString
     transcript_id: NonEmptyString
     room_id: NonEmptyString | None = None
+    source_platform: str = "daily"
 
 
 hatchet = HatchetClientManager.get_client()
@@ -170,15 +170,10 @@ async def set_workflow_error_status(transcript_id: NonEmptyString) -> bool:
 
 
 def _spawn_storage():
-    """Create fresh storage instance."""
-    # TODO: replace direct AwsStorage construction with get_transcripts_storage() factory
-    return AwsStorage(
-        aws_bucket_name=settings.TRANSCRIPT_STORAGE_AWS_BUCKET_NAME,
-        aws_region=settings.TRANSCRIPT_STORAGE_AWS_REGION,
-        aws_access_key_id=settings.TRANSCRIPT_STORAGE_AWS_ACCESS_KEY_ID,
-        aws_secret_access_key=settings.TRANSCRIPT_STORAGE_AWS_SECRET_ACCESS_KEY,
-        aws_endpoint_url=settings.TRANSCRIPT_STORAGE_AWS_ENDPOINT_URL,
-    )
+    """Create fresh storage instance for writing to our transcript bucket."""
+    from reflector.storage import get_transcripts_storage  # noqa: PLC0415
+
+    return get_transcripts_storage()
 
 
 class Loggable(Protocol):
@@ -434,6 +429,7 @@ async def process_tracks(input: PipelineInput, ctx: Context) -> ProcessTracksRes
                 bucket_name=input.bucket_name,
                 transcript_id=input.transcript_id,
                 language=source_language,
+                source_platform=input.source_platform,
             )
         )
         for i, track in enumerate(input.tracks)
@@ -1195,7 +1191,10 @@ async def cleanup_consent(input: PipelineInput, ctx: Context) -> ConsentResult:
         )
         from reflector.db.recordings import recordings_controller  # noqa: PLC0415
         from reflector.db.transcripts import transcripts_controller  # noqa: PLC0415
-        from reflector.storage import get_transcripts_storage  # noqa: PLC0415
+        from reflector.storage import (  # noqa: PLC0415
+            get_source_storage,
+            get_transcripts_storage,
+        )
 
         transcript = await transcripts_controller.get_by_id(input.transcript_id)
         if not transcript:
@@ -1245,7 +1244,7 @@ async def cleanup_consent(input: PipelineInput, ctx: Context) -> ConsentResult:
         deletion_errors = []
 
         if input_track_keys and input.bucket_name:
-            master_storage = get_transcripts_storage()
+            master_storage = get_source_storage(input.source_platform)
             for key in input_track_keys:
                 try:
                     await master_storage.delete_file(key, bucket=input.bucket_name)

server/reflector/hatchet/workflows/padding_workflow.py

@@ -24,6 +24,7 @@ class PaddingInput(BaseModel):
     s3_key: str
     bucket_name: str
     transcript_id: str
+    source_platform: str = "daily"
 
 
 hatchet = HatchetClientManager.get_client()
@@ -45,20 +46,14 @@ async def pad_track(input: PaddingInput, ctx: Context) -> PadTrackResult:
     )
 
     try:
-        # Create fresh storage instance to avoid aioboto3 fork issues
-        from reflector.settings import settings  # noqa: PLC0415
-        from reflector.storage.storage_aws import AwsStorage  # noqa: PLC0415
-
-        # TODO: replace direct AwsStorage construction with get_transcripts_storage() factory
-        storage = AwsStorage(
-            aws_bucket_name=settings.TRANSCRIPT_STORAGE_AWS_BUCKET_NAME,
-            aws_region=settings.TRANSCRIPT_STORAGE_AWS_REGION,
-            aws_access_key_id=settings.TRANSCRIPT_STORAGE_AWS_ACCESS_KEY_ID,
-            aws_secret_access_key=settings.TRANSCRIPT_STORAGE_AWS_SECRET_ACCESS_KEY,
-            aws_endpoint_url=settings.TRANSCRIPT_STORAGE_AWS_ENDPOINT_URL,
+        from reflector.storage import (  # noqa: PLC0415
+            get_source_storage,
+            get_transcripts_storage,
         )
 
-        source_url = await storage.get_file_url(
+        # Source reads: use platform-specific credentials
+        source_storage = get_source_storage(input.source_platform)
+        source_url = await source_storage.get_file_url(
             input.s3_key,
             operation="get_object",
             expires_in=PRESIGNED_URL_EXPIRATION_SECONDS,
@@ -96,52 +91,28 @@ async def pad_track(input: PaddingInput, ctx: Context) -> PadTrackResult:
 
         storage_path = f"file_pipeline_hatchet/{input.transcript_id}/tracks/padded_{input.track_index}.webm"
 
-        # Presign PUT URL for output (Modal will upload directly)
-        output_url = await storage.get_file_url(
+        # Output writes: use transcript storage (our own bucket)
+        output_storage = get_transcripts_storage()
+        output_url = await output_storage.get_file_url(
             storage_path,
             operation="put_object",
             expires_in=PRESIGNED_URL_EXPIRATION_SECONDS,
         )
 
-        import httpx  # noqa: PLC0415
-
-        from reflector.processors.audio_padding_modal import (  # noqa: PLC0415
-            AudioPaddingModalProcessor,
+        from reflector.processors.audio_padding_auto import (  # noqa: PLC0415
+            AudioPaddingAutoProcessor,
         )
 
-        try:
-            processor = AudioPaddingModalProcessor()
-            result = await processor.pad_track(
-                track_url=source_url,
-                output_url=output_url,
-                start_time_seconds=start_time_seconds,
-                track_index=input.track_index,
-            )
-            file_size = result.size
+        processor = AudioPaddingAutoProcessor()
+        result = await processor.pad_track(
+            track_url=source_url,
+            output_url=output_url,
+            start_time_seconds=start_time_seconds,
+            track_index=input.track_index,
+        )
+        file_size = result.size
 
-            ctx.log(f"pad_track: Modal returned size={file_size}")
-        except httpx.HTTPStatusError as e:
-            error_detail = e.response.text if hasattr(e.response, "text") else str(e)
-            logger.error(
-                "[Hatchet] Modal padding HTTP error",
-                transcript_id=input.transcript_id,
-                track_index=input.track_index,
-                status_code=e.response.status_code if hasattr(e, "response") else None,
-                error=error_detail,
-                exc_info=True,
-            )
-            raise Exception(
-                f"Modal padding failed: HTTP {e.response.status_code}"
-            ) from e
-        except httpx.TimeoutException as e:
-            logger.error(
-                "[Hatchet] Modal padding timeout",
-                transcript_id=input.transcript_id,
-                track_index=input.track_index,
-                error=str(e),
-                exc_info=True,
-            )
-            raise Exception("Modal padding timeout") from e
+        ctx.log(f"pad_track: padding returned size={file_size}")
 
         logger.info(
             "[Hatchet] pad_track complete",

server/reflector/hatchet/workflows/track_processing.py

@@ -36,6 +36,7 @@ class TrackInput(BaseModel):
     bucket_name: str
     transcript_id: str
     language: str = "en"
+    source_platform: str = "daily"
 
 
 hatchet = HatchetClientManager.get_client()
@@ -59,20 +60,14 @@ async def pad_track(input: TrackInput, ctx: Context) -> PadTrackResult:
     )
 
     try:
-        # Create fresh storage instance to avoid aioboto3 fork issues
-        # TODO: replace direct AwsStorage construction with get_transcripts_storage() factory
-        from reflector.settings import settings  # noqa: PLC0415
-        from reflector.storage.storage_aws import AwsStorage  # noqa: PLC0415
-
-        storage = AwsStorage(
-            aws_bucket_name=settings.TRANSCRIPT_STORAGE_AWS_BUCKET_NAME,
-            aws_region=settings.TRANSCRIPT_STORAGE_AWS_REGION,
-            aws_access_key_id=settings.TRANSCRIPT_STORAGE_AWS_ACCESS_KEY_ID,
-            aws_secret_access_key=settings.TRANSCRIPT_STORAGE_AWS_SECRET_ACCESS_KEY,
-            aws_endpoint_url=settings.TRANSCRIPT_STORAGE_AWS_ENDPOINT_URL,
+        from reflector.storage import (  # noqa: PLC0415
+            get_source_storage,
+            get_transcripts_storage,
         )
 
-        source_url = await storage.get_file_url(
+        # Source reads: use platform-specific credentials
+        source_storage = get_source_storage(input.source_platform)
+        source_url = await source_storage.get_file_url(
             input.s3_key,
             operation="get_object",
             expires_in=PRESIGNED_URL_EXPIRATION_SECONDS,
@@ -99,18 +94,19 @@ async def pad_track(input: TrackInput, ctx: Context) -> PadTrackResult:
 
         storage_path = f"file_pipeline_hatchet/{input.transcript_id}/tracks/padded_{input.track_index}.webm"
 
-        # Presign PUT URL for output (Modal uploads directly)
-        output_url = await storage.get_file_url(
+        # Output writes: use transcript storage (our own bucket)
+        output_storage = get_transcripts_storage()
+        output_url = await output_storage.get_file_url(
             storage_path,
             operation="put_object",
             expires_in=PRESIGNED_URL_EXPIRATION_SECONDS,
         )
 
-        from reflector.processors.audio_padding_modal import (  # noqa: PLC0415
-            AudioPaddingModalProcessor,
+        from reflector.processors.audio_padding_auto import (  # noqa: PLC0415
+            AudioPaddingAutoProcessor,
         )
 
-        processor = AudioPaddingModalProcessor()
+        processor = AudioPaddingAutoProcessor()
         result = await processor.pad_track(
             track_url=source_url,
             output_url=output_url,
@@ -161,18 +157,18 @@ async def transcribe_track(input: TrackInput, ctx: Context) -> TranscribeTrackRe
             raise ValueError("Missing padded_key from pad_track")
 
         # Presign URL on demand (avoids stale URLs on workflow replay)
-        # TODO: replace direct AwsStorage construction with get_transcripts_storage() factory
-        from reflector.settings import settings  # noqa: PLC0415
-        from reflector.storage.storage_aws import AwsStorage  # noqa: PLC0415
-
-        storage = AwsStorage(
-            aws_bucket_name=settings.TRANSCRIPT_STORAGE_AWS_BUCKET_NAME,
-            aws_region=settings.TRANSCRIPT_STORAGE_AWS_REGION,
-            aws_access_key_id=settings.TRANSCRIPT_STORAGE_AWS_ACCESS_KEY_ID,
-            aws_secret_access_key=settings.TRANSCRIPT_STORAGE_AWS_SECRET_ACCESS_KEY,
-            aws_endpoint_url=settings.TRANSCRIPT_STORAGE_AWS_ENDPOINT_URL,
+        from reflector.storage import (  # noqa: PLC0415
+            get_source_storage,
+            get_transcripts_storage,
         )
 
+        # If bucket_name is set, file is still in the platform's source bucket (no padding applied).
+        # If bucket_name is None, padded file was written to our transcript storage.
+        if bucket_name:
+            storage = get_source_storage(input.source_platform)
+        else:
+            storage = get_transcripts_storage()
+
         audio_url = await storage.get_file_url(
             padded_key,
             operation="get_object",

server/reflector/processors/audio_padding_auto.py

@@ -0,0 +1,31 @@
+import importlib
+
+from reflector.settings import settings
+
+
+class AudioPaddingAutoProcessor:
+    _registry = {}
+
+    @classmethod
+    def register(cls, name, kclass):
+        cls._registry[name] = kclass
+
+    def __new__(cls, name: str | None = None, **kwargs):
+        if name is None:
+            name = settings.PADDING_BACKEND
+        if name not in cls._registry:
+            module_name = f"reflector.processors.audio_padding_{name}"
+            importlib.import_module(module_name)
+
+        # gather specific configuration for the processor
+        # search `PADDING_XXX_YYY`, push to constructor as `xxx_yyy`
+        config = {}
+        name_upper = name.upper()
+        settings_prefix = "PADDING_"
+        config_prefix = f"{settings_prefix}{name_upper}_"
+        for key, value in settings:
+            if key.startswith(config_prefix):
+                config_name = key[len(settings_prefix) :].lower()
+                config[config_name] = value
+
+        return cls._registry[name](**config | kwargs)

server/reflector/processors/audio_padding_local.py

@@ -0,0 +1,133 @@
+"""
+Local audio padding processor using PyAV.
+
+Pads audio tracks with silence directly in-process (no HTTP).
+Reuses the shared PyAV utilities from reflector.utils.audio_padding.
+"""
+
+import asyncio
+import os
+import tempfile
+
+import av
+
+from reflector.logger import logger
+from reflector.processors.audio_padding_auto import AudioPaddingAutoProcessor
+from reflector.processors.audio_padding_modal import PaddingResponse
+from reflector.utils.audio_padding import apply_audio_padding_to_file
+
+S3_TIMEOUT = 60
+
+
+class AudioPaddingLocalProcessor:
+    """Audio padding processor using local PyAV (no HTTP backend)."""
+
+    async def pad_track(
+        self,
+        track_url: str,
+        output_url: str,
+        start_time_seconds: float,
+        track_index: int,
+    ) -> PaddingResponse:
+        """Pad audio track with silence locally via PyAV.
+
+        Args:
+            track_url: Presigned GET URL for source audio track
+            output_url: Presigned PUT URL for output WebM
+            start_time_seconds: Amount of silence to prepend
+            track_index: Track index for logging
+        """
+        if not track_url:
+            raise ValueError("track_url cannot be empty")
+        if start_time_seconds <= 0:
+            raise ValueError(
+                f"start_time_seconds must be positive, got {start_time_seconds}"
+            )
+
+        log = logger.bind(track_index=track_index, padding_seconds=start_time_seconds)
+        log.info("Starting local PyAV padding")
+
+        loop = asyncio.get_event_loop()
+        return await loop.run_in_executor(
+            None,
+            self._pad_track_blocking,
+            track_url,
+            output_url,
+            start_time_seconds,
+            track_index,
+        )
+
+    def _pad_track_blocking(
+        self,
+        track_url: str,
+        output_url: str,
+        start_time_seconds: float,
+        track_index: int,
+    ) -> PaddingResponse:
+        """Blocking padding work: download, pad with PyAV, upload."""
+        import requests
+
+        log = logger.bind(track_index=track_index, padding_seconds=start_time_seconds)
+        temp_dir = tempfile.mkdtemp()
+        input_path = None
+        output_path = None
+
+        try:
+            # Download source audio
+            log.info("Downloading track for local padding")
+            response = requests.get(track_url, stream=True, timeout=S3_TIMEOUT)
+            response.raise_for_status()
+
+            input_path = os.path.join(temp_dir, "track.webm")
+            total_bytes = 0
+            with open(input_path, "wb") as f:
+                for chunk in response.iter_content(chunk_size=8192):
+                    if chunk:
+                        f.write(chunk)
+                        total_bytes += len(chunk)
+            log.info("Track downloaded", bytes=total_bytes)
+
+            # Apply padding using shared PyAV utility
+            output_path = os.path.join(temp_dir, "padded.webm")
+            with av.open(input_path) as in_container:
+                apply_audio_padding_to_file(
+                    in_container,
+                    output_path,
+                    start_time_seconds,
+                    track_index,
+                    logger=logger,
+                )
+
+            file_size = os.path.getsize(output_path)
+            log.info("Local padding complete", size=file_size)
+
+            # Upload padded track
+            log.info("Uploading padded track to S3")
+            with open(output_path, "rb") as f:
+                upload_response = requests.put(output_url, data=f, timeout=S3_TIMEOUT)
+            upload_response.raise_for_status()
+            log.info("Upload complete", size=file_size)
+
+            return PaddingResponse(size=file_size)
+
+        except Exception as e:
+            log.error("Local padding failed", error=str(e), exc_info=True)
+            raise
+        finally:
+            if input_path and os.path.exists(input_path):
+                try:
+                    os.unlink(input_path)
+                except Exception as e:
+                    log.warning("Failed to cleanup input file", error=str(e))
+            if output_path and os.path.exists(output_path):
+                try:
+                    os.unlink(output_path)
+                except Exception as e:
+                    log.warning("Failed to cleanup output file", error=str(e))
+            try:
+                os.rmdir(temp_dir)
+            except Exception as e:
+                log.warning("Failed to cleanup temp directory", error=str(e))
+
+
+AudioPaddingAutoProcessor.register("local", AudioPaddingLocalProcessor)

server/reflector/processors/audio_padding_modal.py

@@ -10,6 +10,7 @@ from pydantic import BaseModel
 
 from reflector.hatchet.constants import TIMEOUT_AUDIO
 from reflector.logger import logger
+from reflector.processors.audio_padding_auto import AudioPaddingAutoProcessor
 
 
 class PaddingResponse(BaseModel):
@@ -111,3 +112,6 @@ class AudioPaddingModalProcessor:
         except Exception as e:
             log.error("Modal padding unexpected error", error=str(e), exc_info=True)
             raise
+
+
+AudioPaddingAutoProcessor.register("modal", AudioPaddingModalProcessor)

server/reflector/services/transcript_process.py

@@ -40,6 +40,7 @@ class MultitrackProcessingConfig:
     track_keys: list[str]
     recording_id: NonEmptyString | None = None
     room_id: NonEmptyString | None = None
+    source_platform: str = "daily"
     mode: Literal["multitrack"] = "multitrack"
 
 
@@ -256,6 +257,7 @@ async def dispatch_transcript_processing(
                 "bucket_name": config.bucket_name,
                 "transcript_id": config.transcript_id,
                 "room_id": config.room_id,
+                "source_platform": config.source_platform,
             },
             additional_metadata={
                 "transcript_id": config.transcript_id,

server/reflector/settings.py

@@ -73,6 +73,9 @@ class Settings(BaseSettings):
     DAILYCO_STORAGE_AWS_BUCKET_NAME: str | None = None
     DAILYCO_STORAGE_AWS_REGION: str | None = None
     DAILYCO_STORAGE_AWS_ROLE_ARN: str | None = None
+    # Worker credentials for reading/deleting from Daily's recording bucket
+    DAILYCO_STORAGE_AWS_ACCESS_KEY_ID: str | None = None
+    DAILYCO_STORAGE_AWS_SECRET_ACCESS_KEY: str | None = None
 
     # Translate into the target language
     TRANSLATION_BACKEND: str = "passthrough"
@@ -106,7 +109,11 @@ class Settings(BaseSettings):
     # Diarization: modal backend
     DIARIZATION_MODAL_API_KEY: str | None = None
 
-    # Audio Padding (Modal.com backend)
+    # Audio Padding
+    # backends:
+    #   - local: in-process PyAV padding (no HTTP, runs in same process)
+    #   - modal: HTTP API client (works with Modal.com OR self-hosted gpu/self_hosted/)
+    PADDING_BACKEND: str = "local"
     PADDING_URL: str | None = None
     PADDING_MODAL_API_KEY: str | None = None
 

server/reflector/storage/__init__.py

@@ -17,6 +17,49 @@ def get_transcripts_storage() -> Storage:
     )
 
 
+def get_source_storage(platform: str) -> Storage:
+    """Get storage for reading/deleting source recording files from the platform's bucket.
+
+    Returns an AwsStorage configured with the platform's worker credentials
+    (access keys), or falls back to get_transcripts_storage() when platform-specific
+    credentials aren't configured (e.g., single-bucket setups).
+
+    Args:
+        platform: Recording platform name ("daily", "whereby", or other).
+    """
+    if platform == "daily":
+        if (
+            settings.DAILYCO_STORAGE_AWS_ACCESS_KEY_ID
+            and settings.DAILYCO_STORAGE_AWS_SECRET_ACCESS_KEY
+            and settings.DAILYCO_STORAGE_AWS_BUCKET_NAME
+        ):
+            from reflector.storage.storage_aws import AwsStorage
+
+            return AwsStorage(
+                aws_bucket_name=settings.DAILYCO_STORAGE_AWS_BUCKET_NAME,
+                aws_region=settings.DAILYCO_STORAGE_AWS_REGION or "us-east-1",
+                aws_access_key_id=settings.DAILYCO_STORAGE_AWS_ACCESS_KEY_ID,
+                aws_secret_access_key=settings.DAILYCO_STORAGE_AWS_SECRET_ACCESS_KEY,
+            )
+
+    elif platform == "whereby":
+        if (
+            settings.WHEREBY_STORAGE_AWS_ACCESS_KEY_ID
+            and settings.WHEREBY_STORAGE_AWS_SECRET_ACCESS_KEY
+            and settings.WHEREBY_STORAGE_AWS_BUCKET_NAME
+        ):
+            from reflector.storage.storage_aws import AwsStorage
+
+            return AwsStorage(
+                aws_bucket_name=settings.WHEREBY_STORAGE_AWS_BUCKET_NAME,
+                aws_region=settings.WHEREBY_STORAGE_AWS_REGION or "us-east-1",
+                aws_access_key_id=settings.WHEREBY_STORAGE_AWS_ACCESS_KEY_ID,
+                aws_secret_access_key=settings.WHEREBY_STORAGE_AWS_SECRET_ACCESS_KEY,
+            )
+
+    return get_transcripts_storage()
+
+
 def get_whereby_storage() -> Storage:
     """
     Get storage config for Whereby (for passing to Whereby API).

server/reflector/worker/app.py

@@ -24,6 +24,118 @@ RECONCILIATION_INTERVAL = _override or 30.0
 ICS_SYNC_INTERVAL = _override or 60.0
 UPCOMING_MEETINGS_INTERVAL = _override or 30.0
 
+
+def build_beat_schedule(
+    *,
+    whereby_api_key=None,
+    aws_process_recording_queue_url=None,
+    daily_api_key=None,
+    public_mode=False,
+    public_data_retention_days=None,
+    healthcheck_url=None,
+):
+    """Build the Celery beat schedule based on configured services.
+
+    Only registers tasks for services that are actually configured,
+    avoiding unnecessary worker wake-ups in selfhosted deployments.
+    """
+    beat_schedule = {}
+
+    _whereby_enabled = bool(whereby_api_key) or bool(aws_process_recording_queue_url)
+    if _whereby_enabled:
+        beat_schedule["process_messages"] = {
+            "task": "reflector.worker.process.process_messages",
+            "schedule": SQS_POLL_INTERVAL,
+        }
+        beat_schedule["reprocess_failed_recordings"] = {
+            "task": "reflector.worker.process.reprocess_failed_recordings",
+            "schedule": crontab(hour=5, minute=0),  # Midnight EST
+        }
+        logger.info(
+            "Whereby beat tasks enabled",
+            tasks=["process_messages", "reprocess_failed_recordings"],
+        )
+    else:
+        logger.info("Whereby beat tasks disabled (no WHEREBY_API_KEY or SQS URL)")
+
+    _daily_enabled = bool(daily_api_key)
+    if _daily_enabled:
+        beat_schedule["poll_daily_recordings"] = {
+            "task": "reflector.worker.process.poll_daily_recordings",
+            "schedule": POLL_DAILY_RECORDINGS_INTERVAL_SEC,
+        }
+        beat_schedule["trigger_daily_reconciliation"] = {
+            "task": "reflector.worker.process.trigger_daily_reconciliation",
+            "schedule": RECONCILIATION_INTERVAL,
+        }
+        beat_schedule["reprocess_failed_daily_recordings"] = {
+            "task": "reflector.worker.process.reprocess_failed_daily_recordings",
+            "schedule": crontab(hour=5, minute=0),  # Midnight EST
+        }
+        logger.info(
+            "Daily.co beat tasks enabled",
+            tasks=[
+                "poll_daily_recordings",
+                "trigger_daily_reconciliation",
+                "reprocess_failed_daily_recordings",
+            ],
+        )
+    else:
+        logger.info("Daily.co beat tasks disabled (no DAILY_API_KEY)")
+
+    _any_platform = _whereby_enabled or _daily_enabled
+    if _any_platform:
+        beat_schedule["process_meetings"] = {
+            "task": "reflector.worker.process.process_meetings",
+            "schedule": SQS_POLL_INTERVAL,
+        }
+        beat_schedule["sync_all_ics_calendars"] = {
+            "task": "reflector.worker.ics_sync.sync_all_ics_calendars",
+            "schedule": ICS_SYNC_INTERVAL,
+        }
+        beat_schedule["create_upcoming_meetings"] = {
+            "task": "reflector.worker.ics_sync.create_upcoming_meetings",
+            "schedule": UPCOMING_MEETINGS_INTERVAL,
+        }
+        logger.info(
+            "Platform tasks enabled",
+            tasks=[
+                "process_meetings",
+                "sync_all_ics_calendars",
+                "create_upcoming_meetings",
+            ],
+        )
+    else:
+        logger.info("Platform tasks disabled (no video platform configured)")
+
+    if public_mode:
+        beat_schedule["cleanup_old_public_data"] = {
+            "task": "reflector.worker.cleanup.cleanup_old_public_data_task",
+            "schedule": crontab(hour=3, minute=0),
+        }
+        logger.info(
+            "Public mode cleanup enabled",
+            retention_days=public_data_retention_days,
+        )
+
+    if healthcheck_url:
+        beat_schedule["healthcheck_ping"] = {
+            "task": "reflector.worker.healthcheck.healthcheck_ping",
+            "schedule": 60.0 * 10,
+        }
+        logger.info("Healthcheck enabled", url=healthcheck_url)
+    else:
+        logger.warning("Healthcheck disabled, no url configured")
+
+    logger.info(
+        "Beat schedule configured",
+        total_tasks=len(beat_schedule),
+        task_names=sorted(beat_schedule.keys()),
+    )
+
+    return beat_schedule
+
+
 if celery.current_app.main != "default":
     logger.info(f"Celery already configured ({celery.current_app})")
     app = celery.current_app
@@ -42,57 +154,11 @@ else:
         ]
     )
 
-    # crontab
-    app.conf.beat_schedule = {
-        "process_messages": {
-            "task": "reflector.worker.process.process_messages",
-            "schedule": SQS_POLL_INTERVAL,
-        },
-        "process_meetings": {
-            "task": "reflector.worker.process.process_meetings",
-            "schedule": SQS_POLL_INTERVAL,
-        },
-        "reprocess_failed_recordings": {
-            "task": "reflector.worker.process.reprocess_failed_recordings",
-            "schedule": crontab(hour=5, minute=0),  # Midnight EST
-        },
-        "reprocess_failed_daily_recordings": {
-            "task": "reflector.worker.process.reprocess_failed_daily_recordings",
-            "schedule": crontab(hour=5, minute=0),  # Midnight EST
-        },
-        "poll_daily_recordings": {
-            "task": "reflector.worker.process.poll_daily_recordings",
-            "schedule": POLL_DAILY_RECORDINGS_INTERVAL_SEC,
-        },
-        "trigger_daily_reconciliation": {
-            "task": "reflector.worker.process.trigger_daily_reconciliation",
-            "schedule": RECONCILIATION_INTERVAL,
-        },
-        "sync_all_ics_calendars": {
-            "task": "reflector.worker.ics_sync.sync_all_ics_calendars",
-            "schedule": ICS_SYNC_INTERVAL,
-        },
-        "create_upcoming_meetings": {
-            "task": "reflector.worker.ics_sync.create_upcoming_meetings",
-            "schedule": UPCOMING_MEETINGS_INTERVAL,
-        },
-    }
-
-    if settings.PUBLIC_MODE:
-        app.conf.beat_schedule["cleanup_old_public_data"] = {
-            "task": "reflector.worker.cleanup.cleanup_old_public_data_task",
-            "schedule": crontab(hour=3, minute=0),
-        }
-        logger.info(
-            "Public mode cleanup enabled",
-            retention_days=settings.PUBLIC_DATA_RETENTION_DAYS,
-        )
-
-    if settings.HEALTHCHECK_URL:
-        app.conf.beat_schedule["healthcheck_ping"] = {
-            "task": "reflector.worker.healthcheck.healthcheck_ping",
-            "schedule": 60.0 * 10,
-        }
-        logger.info("Healthcheck enabled", url=settings.HEALTHCHECK_URL)
-    else:
-        logger.warning("Healthcheck disabled, no url configured")
+    app.conf.beat_schedule = build_beat_schedule(
+        whereby_api_key=settings.WHEREBY_API_KEY,
+        aws_process_recording_queue_url=settings.AWS_PROCESS_RECORDING_QUEUE_URL,
+        daily_api_key=settings.DAILY_API_KEY,
+        public_mode=settings.PUBLIC_MODE,
+        public_data_retention_days=settings.PUBLIC_DATA_RETENTION_DAYS,
+        healthcheck_url=settings.HEALTHCHECK_URL,
+    )

server/reflector/worker/process.py

@@ -357,6 +357,7 @@ async def _process_multitrack_recording_inner(
             "bucket_name": bucket_name,
             "transcript_id": transcript.id,
             "room_id": room.id,
+            "source_platform": "daily",
         },
         additional_metadata={
             "transcript_id": transcript.id,
@@ -1068,6 +1069,7 @@ async def reprocess_failed_daily_recordings():
                     "bucket_name": bucket_name,
                     "transcript_id": transcript.id,
                     "room_id": room.id if room else None,
+                    "source_platform": "daily",
                 },
                 additional_metadata={
                     "transcript_id": transcript.id,

server/tests/test_beat_schedule.py

@@ -0,0 +1,247 @@
+"""Tests for conditional Celery beat schedule registration.
+
+Verifies that beat tasks are only registered when their corresponding
+services are configured (WHEREBY_API_KEY, DAILY_API_KEY, etc.).
+"""
+
+import pytest
+
+from reflector.worker.app import build_beat_schedule
+
+
+# Override autouse fixtures from conftest — these tests don't need database or websockets
+@pytest.fixture(autouse=True)
+def setup_database():
+    yield
+
+
+@pytest.fixture(autouse=True)
+def ws_manager_in_memory():
+    yield
+
+
+@pytest.fixture(autouse=True)
+def reset_hatchet_client():
+    yield
+
+
+# Task name sets for each group
+WHEREBY_TASKS = {"process_messages", "reprocess_failed_recordings"}
+DAILY_TASKS = {
+    "poll_daily_recordings",
+    "trigger_daily_reconciliation",
+    "reprocess_failed_daily_recordings",
+}
+PLATFORM_TASKS = {
+    "process_meetings",
+    "sync_all_ics_calendars",
+    "create_upcoming_meetings",
+}
+
+
+class TestNoPlatformConfigured:
+    """When no video platform is configured, no platform tasks should be registered."""
+
+    def test_no_platform_tasks(self):
+        schedule = build_beat_schedule()
+        task_names = set(schedule.keys())
+        assert not task_names & WHEREBY_TASKS
+        assert not task_names & DAILY_TASKS
+        assert not task_names & PLATFORM_TASKS
+
+    def test_only_healthcheck_disabled_warning(self):
+        """With no config at all, schedule should be empty (healthcheck needs URL)."""
+        schedule = build_beat_schedule()
+        assert len(schedule) == 0
+
+    def test_healthcheck_only(self):
+        schedule = build_beat_schedule(healthcheck_url="https://hc.example.com/ping")
+        assert set(schedule.keys()) == {"healthcheck_ping"}
+
+    def test_public_mode_only(self):
+        schedule = build_beat_schedule(public_mode=True)
+        assert set(schedule.keys()) == {"cleanup_old_public_data"}
+
+
+class TestWherebyOnly:
+    """When only Whereby is configured."""
+
+    def test_whereby_api_key(self):
+        schedule = build_beat_schedule(whereby_api_key="test-key")
+        task_names = set(schedule.keys())
+        assert WHEREBY_TASKS <= task_names
+        assert PLATFORM_TASKS <= task_names
+        assert not task_names & DAILY_TASKS
+
+    def test_whereby_sqs_url(self):
+        schedule = build_beat_schedule(
+            aws_process_recording_queue_url="https://sqs.us-east-1.amazonaws.com/123/queue"
+        )
+        task_names = set(schedule.keys())
+        assert WHEREBY_TASKS <= task_names
+        assert PLATFORM_TASKS <= task_names
+        assert not task_names & DAILY_TASKS
+
+    def test_whereby_task_count(self):
+        schedule = build_beat_schedule(whereby_api_key="test-key")
+        # Whereby (2) + Platform (3) = 5
+        assert len(schedule) == 5
+
+
+class TestDailyOnly:
+    """When only Daily.co is configured."""
+
+    def test_daily_api_key(self):
+        schedule = build_beat_schedule(daily_api_key="test-daily-key")
+        task_names = set(schedule.keys())
+        assert DAILY_TASKS <= task_names
+        assert PLATFORM_TASKS <= task_names
+        assert not task_names & WHEREBY_TASKS
+
+    def test_daily_task_count(self):
+        schedule = build_beat_schedule(daily_api_key="test-daily-key")
+        # Daily (3) + Platform (3) = 6
+        assert len(schedule) == 6
+
+
+class TestBothPlatforms:
+    """When both Whereby and Daily.co are configured."""
+
+    def test_all_tasks_registered(self):
+        schedule = build_beat_schedule(
+            whereby_api_key="test-key",
+            daily_api_key="test-daily-key",
+        )
+        task_names = set(schedule.keys())
+        assert WHEREBY_TASKS <= task_names
+        assert DAILY_TASKS <= task_names
+        assert PLATFORM_TASKS <= task_names
+
+    def test_combined_task_count(self):
+        schedule = build_beat_schedule(
+            whereby_api_key="test-key",
+            daily_api_key="test-daily-key",
+        )
+        # Whereby (2) + Daily (3) + Platform (3) = 8
+        assert len(schedule) == 8
+
+
+class TestConditionalFlags:
+    """Test PUBLIC_MODE and HEALTHCHECK_URL interact correctly with platform tasks."""
+
+    def test_all_flags_enabled(self):
+        schedule = build_beat_schedule(
+            whereby_api_key="test-key",
+            daily_api_key="test-daily-key",
+            public_mode=True,
+            healthcheck_url="https://hc.example.com/ping",
+        )
+        task_names = set(schedule.keys())
+        assert "cleanup_old_public_data" in task_names
+        assert "healthcheck_ping" in task_names
+        assert WHEREBY_TASKS <= task_names
+        assert DAILY_TASKS <= task_names
+        assert PLATFORM_TASKS <= task_names
+        # Whereby (2) + Daily (3) + Platform (3) + cleanup (1) + healthcheck (1) = 10
+        assert len(schedule) == 10
+
+    def test_public_mode_with_whereby(self):
+        schedule = build_beat_schedule(
+            whereby_api_key="test-key",
+            public_mode=True,
+        )
+        task_names = set(schedule.keys())
+        assert "cleanup_old_public_data" in task_names
+        assert WHEREBY_TASKS <= task_names
+        assert PLATFORM_TASKS <= task_names
+
+    def test_healthcheck_with_daily(self):
+        schedule = build_beat_schedule(
+            daily_api_key="test-daily-key",
+            healthcheck_url="https://hc.example.com/ping",
+        )
+        task_names = set(schedule.keys())
+        assert "healthcheck_ping" in task_names
+        assert DAILY_TASKS <= task_names
+        assert PLATFORM_TASKS <= task_names
+
+
+class TestTaskDefinitions:
+    """Verify task definitions have correct structure."""
+
+    def test_whereby_task_paths(self):
+        schedule = build_beat_schedule(whereby_api_key="test-key")
+        assert (
+            schedule["process_messages"]["task"]
+            == "reflector.worker.process.process_messages"
+        )
+        assert (
+            schedule["reprocess_failed_recordings"]["task"]
+            == "reflector.worker.process.reprocess_failed_recordings"
+        )
+
+    def test_daily_task_paths(self):
+        schedule = build_beat_schedule(daily_api_key="test-daily-key")
+        assert (
+            schedule["poll_daily_recordings"]["task"]
+            == "reflector.worker.process.poll_daily_recordings"
+        )
+        assert (
+            schedule["trigger_daily_reconciliation"]["task"]
+            == "reflector.worker.process.trigger_daily_reconciliation"
+        )
+        assert (
+            schedule["reprocess_failed_daily_recordings"]["task"]
+            == "reflector.worker.process.reprocess_failed_daily_recordings"
+        )
+
+    def test_platform_task_paths(self):
+        schedule = build_beat_schedule(daily_api_key="test-daily-key")
+        assert (
+            schedule["process_meetings"]["task"]
+            == "reflector.worker.process.process_meetings"
+        )
+        assert (
+            schedule["sync_all_ics_calendars"]["task"]
+            == "reflector.worker.ics_sync.sync_all_ics_calendars"
+        )
+        assert (
+            schedule["create_upcoming_meetings"]["task"]
+            == "reflector.worker.ics_sync.create_upcoming_meetings"
+        )
+
+    def test_all_tasks_have_schedule(self):
+        """Every registered task must have a 'schedule' key."""
+        schedule = build_beat_schedule(
+            whereby_api_key="test-key",
+            daily_api_key="test-daily-key",
+            public_mode=True,
+            healthcheck_url="https://hc.example.com/ping",
+        )
+        for name, config in schedule.items():
+            assert "schedule" in config, f"Task '{name}' missing 'schedule' key"
+            assert "task" in config, f"Task '{name}' missing 'task' key"
+
+
+class TestEmptyStringValues:
+    """Empty strings should be treated as not configured (falsy)."""
+
+    def test_empty_whereby_key(self):
+        schedule = build_beat_schedule(whereby_api_key="")
+        assert not set(schedule.keys()) & WHEREBY_TASKS
+
+    def test_empty_daily_key(self):
+        schedule = build_beat_schedule(daily_api_key="")
+        assert not set(schedule.keys()) & DAILY_TASKS
+
+    def test_empty_sqs_url(self):
+        schedule = build_beat_schedule(aws_process_recording_queue_url="")
+        assert not set(schedule.keys()) & WHEREBY_TASKS
+
+    def test_none_values(self):
+        schedule = build_beat_schedule(
+            whereby_api_key=None,
+            daily_api_key=None,
+            aws_process_recording_queue_url=None,
+        )
+        assert len(schedule) == 0

server/tests/test_storage.py

@@ -367,3 +367,126 @@ async def test_aws_storage_none_endpoint_url():
     assert storage.base_url == "https://reflector-bucket.s3.amazonaws.com/"
     # No s3 addressing_style override — boto_config should only have retries
     assert not hasattr(storage.boto_config, "s3") or storage.boto_config.s3 is None
+
+
+# --- Tests for get_source_storage() ---
+
+
+def test_get_source_storage_daily_with_credentials():
+    """Daily platform with access keys returns AwsStorage with Daily credentials."""
+    with patch("reflector.storage.settings") as mock_settings:
+        mock_settings.DAILYCO_STORAGE_AWS_ACCESS_KEY_ID = "daily-key"
+        mock_settings.DAILYCO_STORAGE_AWS_SECRET_ACCESS_KEY = "daily-secret"
+        mock_settings.DAILYCO_STORAGE_AWS_BUCKET_NAME = "daily-bucket"
+        mock_settings.DAILYCO_STORAGE_AWS_REGION = "us-west-2"
+
+        from reflector.storage import get_source_storage
+
+        storage = get_source_storage("daily")
+
+        assert isinstance(storage, AwsStorage)
+        assert storage._bucket_name == "daily-bucket"
+        assert storage._region == "us-west-2"
+        assert storage._access_key_id == "daily-key"
+        assert storage._secret_access_key == "daily-secret"
+        assert storage._endpoint_url is None
+
+
+def test_get_source_storage_daily_falls_back_without_credentials():
+    """Daily platform without access keys falls back to transcript storage."""
+    with patch("reflector.storage.settings") as mock_settings:
+        mock_settings.DAILYCO_STORAGE_AWS_ACCESS_KEY_ID = None
+        mock_settings.DAILYCO_STORAGE_AWS_SECRET_ACCESS_KEY = None
+        mock_settings.DAILYCO_STORAGE_AWS_BUCKET_NAME = "daily-bucket"
+        mock_settings.TRANSCRIPT_STORAGE_BACKEND = "aws"
+        mock_settings.TRANSCRIPT_STORAGE_AWS_BUCKET_NAME = "transcript-bucket"
+        mock_settings.TRANSCRIPT_STORAGE_AWS_REGION = "us-east-1"
+        mock_settings.TRANSCRIPT_STORAGE_AWS_ACCESS_KEY_ID = "transcript-key"
+        mock_settings.TRANSCRIPT_STORAGE_AWS_SECRET_ACCESS_KEY = "transcript-secret"
+        mock_settings.TRANSCRIPT_STORAGE_AWS_ENDPOINT_URL = None
+
+        from reflector.storage import get_source_storage
+
+        with patch("reflector.storage.get_transcripts_storage") as mock_get_transcripts:
+            fallback = AwsStorage(
+                aws_bucket_name="transcript-bucket",
+                aws_region="us-east-1",
+                aws_access_key_id="transcript-key",
+                aws_secret_access_key="transcript-secret",
+            )
+            mock_get_transcripts.return_value = fallback
+
+            storage = get_source_storage("daily")
+
+            mock_get_transcripts.assert_called_once()
+            assert storage is fallback
+
+
+def test_get_source_storage_whereby_with_credentials():
+    """Whereby platform with access keys returns AwsStorage with Whereby credentials."""
+    with patch("reflector.storage.settings") as mock_settings:
+        mock_settings.WHEREBY_STORAGE_AWS_ACCESS_KEY_ID = "whereby-key"
+        mock_settings.WHEREBY_STORAGE_AWS_SECRET_ACCESS_KEY = "whereby-secret"
+        mock_settings.WHEREBY_STORAGE_AWS_BUCKET_NAME = "whereby-bucket"
+        mock_settings.WHEREBY_STORAGE_AWS_REGION = "eu-west-1"
+
+        from reflector.storage import get_source_storage
+
+        storage = get_source_storage("whereby")
+
+        assert isinstance(storage, AwsStorage)
+        assert storage._bucket_name == "whereby-bucket"
+        assert storage._region == "eu-west-1"
+        assert storage._access_key_id == "whereby-key"
+        assert storage._secret_access_key == "whereby-secret"
+
+
+def test_get_source_storage_unknown_platform_falls_back():
+    """Unknown platform falls back to transcript storage."""
+    with patch("reflector.storage.settings"):
+        from reflector.storage import get_source_storage
+
+        with patch("reflector.storage.get_transcripts_storage") as mock_get_transcripts:
+            fallback = MagicMock()
+            mock_get_transcripts.return_value = fallback
+
+            storage = get_source_storage("unknown-platform")
+
+            mock_get_transcripts.assert_called_once()
+            assert storage is fallback
+
+
+@pytest.mark.asyncio
+async def test_source_storage_presigns_for_correct_bucket():
+    """Source storage presigns URLs using the platform's credentials and the override bucket."""
+    with patch("reflector.storage.settings") as mock_settings:
+        mock_settings.DAILYCO_STORAGE_AWS_ACCESS_KEY_ID = "daily-key"
+        mock_settings.DAILYCO_STORAGE_AWS_SECRET_ACCESS_KEY = "daily-secret"
+        mock_settings.DAILYCO_STORAGE_AWS_BUCKET_NAME = "daily-bucket"
+        mock_settings.DAILYCO_STORAGE_AWS_REGION = "us-west-2"
+
+        from reflector.storage import get_source_storage
+
+        storage = get_source_storage("daily")
+
+        mock_client = AsyncMock()
+        mock_client.generate_presigned_url = AsyncMock(
+            return_value="https://daily-bucket.s3.amazonaws.com/track.webm?signed"
+        )
+        mock_client.__aenter__ = AsyncMock(return_value=mock_client)
+        mock_client.__aexit__ = AsyncMock(return_value=None)
+
+        with patch.object(storage.session, "client", return_value=mock_client):
+            url = await storage.get_file_url(
+                "track.webm",
+                operation="get_object",
+                expires_in=3600,
+                bucket="override-bucket",
+            )
+
+            assert "track.webm" in url
+            mock_client.generate_presigned_url.assert_called_once()
+            call_kwargs = mock_client.generate_presigned_url.call_args
+            params = call_kwargs[1].get("Params") or call_kwargs[0][1]
+            assert params["Bucket"] == "override-bucket"
+            assert params["Key"] == "track.webm"