Add support for transcript only scribes

care_scribe/migrations/0012_scribe_transcript_only.py

@@ -0,0 +1,98 @@
+from django.db import migrations, models
+
+
+def rename_processing_meta_keys_forward(apps, schema_editor):
+    """Rename old processing meta keys to the new keys.
+
+    Old -> New:
+        provider     -> chat_provider
+        audio_model  -> transcribe_model
+
+    Also adds `transcribe_provider` mirroring `chat_provider` so historical
+    entries match the new shape.
+    """
+    Scribe = apps.get_model("care_scribe", "Scribe")
+    to_update = []
+    for scribe in Scribe.objects.exclude(meta={}).iterator():
+        meta = scribe.meta or {}
+        processings = meta.get("processings")
+        if not processings:
+            continue
+        changed = False
+        for processing in processings:
+            if not isinstance(processing, dict):
+                continue
+            if "provider" in processing and "chat_provider" not in processing:
+                processing["chat_provider"] = processing.pop("provider")
+                changed = True
+            if "audio_model" in processing and "transcribe_model" not in processing:
+                processing["transcribe_model"] = processing.pop("audio_model")
+                changed = True
+            if (
+                "chat_provider" in processing
+                and "transcribe_provider" not in processing
+            ):
+                processing["transcribe_provider"] = processing["chat_provider"]
+                changed = True
+        if changed:
+            scribe.meta = meta
+            to_update.append(scribe)
+            if len(to_update) >= 500:
+                Scribe.objects.bulk_update(to_update, ["meta"])
+                to_update = []
+    if to_update:
+        Scribe.objects.bulk_update(to_update, ["meta"])
+
+
+def rename_processing_meta_keys_reverse(apps, schema_editor):
+    """Revert the rename: new keys -> old keys."""
+    Scribe = apps.get_model("care_scribe", "Scribe")
+    to_update = []
+    for scribe in Scribe.objects.exclude(meta={}).iterator():
+        meta = scribe.meta or {}
+        processings = meta.get("processings")
+        if not processings:
+            continue
+        changed = False
+        for processing in processings:
+            if not isinstance(processing, dict):
+                continue
+            if "chat_provider" in processing and "provider" not in processing:
+                processing["provider"] = processing.pop("chat_provider")
+                changed = True
+            if "transcribe_model" in processing and "audio_model" not in processing:
+                processing["audio_model"] = processing.pop("transcribe_model")
+                changed = True
+            if "transcribe_provider" in processing:
+                processing.pop("transcribe_provider")
+                changed = True
+        if changed:
+            scribe.meta = meta
+            to_update.append(scribe)
+            if len(to_update) >= 500:
+                Scribe.objects.bulk_update(to_update, ["meta"])
+                to_update = []
+    if to_update:
+        Scribe.objects.bulk_update(to_update, ["meta"])
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('care_scribe', '0011_scribefile_mime_type'),
+    ]
+
+    operations = [
+        migrations.AddField(
+            model_name='scribe',
+            name='transcript_only',
+            field=models.BooleanField(
+                default=False,
+                help_text='If True, only transcribe the audio without running any AI form-fill processing.',
+            ),
+        ),
+        migrations.RunPython(
+            rename_processing_meta_keys_forward,
+            rename_processing_meta_keys_reverse,
+        ),
+    ]

care_scribe/models/scribe.py

@@ -124,6 +124,10 @@ class Status(models.TextChoices):
     chat_model = models.CharField(max_length=100, null=True, blank=True)
     audio_model = models.CharField(max_length=100, null=True, blank=True)
     chat_model_temperature = models.FloatField(null=True, blank=True)
+    transcript_only = models.BooleanField(
+        default=False,
+        help_text="If True, only transcribe the audio without running any AI form-fill processing.",
+    )
 
     is_feedback_positive = models.BooleanField(null=True, blank=True, help_text="Whether the user has given positive feedback on the AI response")
     feedback_comments = models.TextField(null=True, blank=True, help_text="Details of the feedback provided by the user")

care_scribe/serializers/scribe.py

@@ -77,6 +77,7 @@ class Meta:
             "chat_model",
             "audio_model",
             "chat_model_temperature",
+            "transcript_only",
             "is_feedback_positive",
             "feedback_comments",
         ]

care_scribe/settings.py

@@ -86,29 +86,40 @@ def validate(self) -> None:
                     f'Please set the "{setting}" in the environment or the {PLUGIN_NAME} plugin config.'
                 )
 
-        if getattr(self, "SCRIBE_API_PROVIDER") not in ("openai", "azure", "google"):
+        valid_providers = ("openai", "azure", "google")
+        providers_in_use = set()
+
+        for setting_name in ("SCRIBE_CHAT_MODEL_NAME", "SCRIBE_TRANSCRIBE_MODEL_NAME"):
+            value = getattr(self, setting_name)
+            if "/" not in value:
+                raise ImproperlyConfigured(
+                    f'Invalid value for "{setting_name}". '
+                    f'Expected format "provider/model-name" '
+                    f'(provider must be one of {valid_providers}).'
+                )
+            provider = value.split("/", 1)[0]
+            if provider not in valid_providers:
+                raise ImproperlyConfigured(
+                    f'Invalid provider "{provider}" in "{setting_name}". '
+                    f'Provider must be one of {valid_providers}.'
+                )
+            providers_in_use.add(provider)
+
+        if "openai" in providers_in_use and not getattr(self, "SCRIBE_OPENAI_API_KEY"):
             raise ImproperlyConfigured(
-                'Invalid value for "SCRIBE_API_PROVIDER". '
-                'Please set the "SCRIBE_API_PROVIDER" to "openai", "google" or "azure".'
+                'The "SCRIBE_OPENAI_API_KEY" setting is required when using OpenAI API. '
+                f'Please set it in the environment or the {PLUGIN_NAME} plugin config.'
             )
 
-        if getattr(self, "SCRIBE_API_PROVIDER") == "openai":
-            for setting in ("SCRIBE_OPENAI_API_KEY",):
-                if not getattr(self, setting):
-                    raise ImproperlyConfigured(
-                        f'The "{setting}" setting is required when using OpenAI API. '
-                        f'Please set the "{setting}" in the environment or the {PLUGIN_NAME} plugin config.'
-                    )
-
-        if getattr(self, "SCRIBE_API_PROVIDER") == "azure":
+        if "azure" in providers_in_use:
             for setting in ("SCRIBE_AZURE_API_VERSION", "SCRIBE_AZURE_ENDPOINT", "SCRIBE_AZURE_API_KEY"):
                 if not getattr(self, setting):
                     raise ImproperlyConfigured(
                         f'The "{setting}" setting is required when using Azure API. '
                         f'Please set the "{setting}" in the environment or the {PLUGIN_NAME} plugin config.'
                     )
 
-        if getattr(self, "SCRIBE_API_PROVIDER") == "google":
+        if "google" in providers_in_use:
             for setting in ("SCRIBE_GOOGLE_PROJECT_ID", "SCRIBE_GOOGLE_LOCATION"):
                 if not getattr(self, setting):
                     raise ImproperlyConfigured(
@@ -129,19 +140,19 @@ def reload(self) -> None:
 
 REQUIRED_SETTINGS = {
     "SCRIBE_CHAT_MODEL_NAME",
-    "SCRIBE_API_PROVIDER",
+    "SCRIBE_TRANSCRIBE_MODEL_NAME",
 }
 
 DEFAULTS = {
     "SCRIBE_OPENAI_API_KEY": "",
     "SCRIBE_AZURE_API_KEY": "",
-    "SCRIBE_AUDIO_MODEL_NAME": "whisper-1",
-    "SCRIBE_CHAT_MODEL_NAME": "gpt-4o",
-    "SCRIBE_API_PROVIDER": "openai",
+    "SCRIBE_TRANSCRIBE_MODEL_NAME": "openai/whisper-1",
+    "SCRIBE_CHAT_MODEL_NAME": "openai/gpt-4o",
     "SCRIBE_AZURE_API_VERSION": "",
     "SCRIBE_AZURE_ENDPOINT": "",
     "SCRIBE_GOOGLE_PROJECT_ID" : "",
     "SCRIBE_GOOGLE_LOCATION" : "",
+    "SCRIBE_TRANSCRIBE_LANGUAGE": "", # only works for google. OpenAI can return source language or only translate to English.
     "SCRIBE_TNC": "<Please add your terms and conditions here>",
 }