Skip to content

Speakr is a personal, self-hosted web application designed for transcribing audio recordings

Python

Fix ASR_DIARIZE=false being ignored by connector pipeline

Murtaza Nasir2f23f42a8404
diff --git a/src/config/app_config.py b/src/config/app_config.py
index 2f6de865..cae52cd1 100644
--- a/src/config/app_config.py
+++ b/src/config/app_config.py
@@ -126,10 +126,13 @@ def initialize_config(app):
app.logger.info(f"Connector capabilities: {capabilities}")
# Log diarization support prominently
- if connector.supports_diarization:
- app.logger.info("Speaker diarization: ENABLED")
- else:
+ diarize_default = getattr(connector, 'default_diarize', connector.supports_diarization)
+ if not connector.supports_diarization:
app.logger.info("Speaker diarization: NOT AVAILABLE (connector does not support it)")
+ elif not diarize_default:
+ app.logger.info("Speaker diarization: DISABLED (ASR_DIARIZE=false)")
+ else:
+ app.logger.info("Speaker diarization: ENABLED")
except Exception as e:
app.logger.error(f"Failed to initialize transcription connector: {e}")
diff --git a/src/services/transcription/connectors/asr_endpoint.py b/src/services/transcription/connectors/asr_endpoint.py
index c39c98b7..0d231503 100644
--- a/src/services/transcription/connectors/asr_endpoint.py
+++ b/src/services/transcription/connectors/asr_endpoint.py
@@ -143,15 +143,14 @@ class ASREndpointConnector(BaseTranscriptionConnector):
# Determine if we should diarize
should_diarize = request.diarize if request.diarize is not None else self.default_diarize
- if should_diarize:
- # Send both parameter names for compatibility:
- # - 'diarize' is used by whisper-asr-webservice
- # - 'enable_diarization' is used by WhisperX
- params['diarize'] = True
- params['enable_diarization'] = True
-
- if self.return_embeddings:
- params['return_speaker_embeddings'] = True
+ # Send both parameter names for compatibility:
+ # - 'diarize' is used by whisper-asr-webservice
+ # - 'enable_diarization' is used by WhisperX
+ params['diarize'] = should_diarize
+ params['enable_diarization'] = should_diarize
+
+ if should_diarize and self.return_embeddings:
+ params['return_speaker_embeddings'] = True
if request.min_speakers:
params['min_speakers'] = request.min_speakers
diff --git a/src/tasks/processing.py b/src/tasks/processing.py
index 0024d2e4..a3b691a4 100644
--- a/src/tasks/processing.py
+++ b/src/tasks/processing.py
@@ -1544,8 +1544,8 @@ def transcribe_with_connector(app_context, recording_id, filepath, original_file
# Determine if we should diarize
if diarize is None:
- # Use connector's default diarization setting
- should_diarize = connector.supports_diarization
+ # Use connector's configured default (respects ASR_DIARIZE env var)
+ should_diarize = getattr(connector, 'default_diarize', connector.supports_diarization)
else:
should_diarize = diarize and connector.supports_diarization
@@ -1960,8 +1960,8 @@ def transcribe_incognito(filepath, original_filename, language=None, min_speaker
except Exception as e:
current_app.logger.warning(f"[Incognito] Could not get audio duration: {e}")
- # Determine diarization settings
- should_diarize = connector.supports_diarization
+ # Determine diarization settings (respects ASR_DIARIZE env var)
+ should_diarize = getattr(connector, 'default_diarize', connector.supports_diarization)
# Use user's language preference if not explicitly provided
if language is None and user: