Start to learn vocab from talkgroups to improve accuracy of STT

drb-c2-core/app/internal/transcription.py

@@ -28,6 +28,7 @@ async def transcribe_call(
     call_id: str,
     gcs_uri: str,
     talkgroup_name: Optional[str] = None,
+    system_id: Optional[str] = None,
 ) -> tuple[Optional[str], list[dict]]:
     """
     Transcribe audio at the given GCS URI and store the result in Firestore.
@@ -39,8 +40,17 @@ async def transcribe_call(
     if not gcs_uri or not gcs_uri.startswith("gs://"):
         return None, []
 
+    # Load vocabulary for this system (empty list if none yet)
+    vocabulary: list[str] = []
+    if system_id:
+        from app.internal.vocabulary_learner import get_vocabulary
+        vocab_data = await get_vocabulary(system_id)
+        vocabulary = vocab_data.get("vocabulary") or []
+
     try:
-        transcript, segments = await asyncio.to_thread(_sync_transcribe, gcs_uri, talkgroup_name)
+        transcript, segments = await asyncio.to_thread(
+            _sync_transcribe, gcs_uri, talkgroup_name, vocabulary
+        )
     except Exception as e:
         logger.warning(f"Transcription failed for call {call_id}: {e}")
         return None, []
@@ -61,7 +71,11 @@ async def transcribe_call(
     return transcript, segments
 
 
-def _sync_transcribe(gcs_uri: str, talkgroup_name: Optional[str] = None) -> tuple[Optional[str], list[dict]]:
+def _sync_transcribe(
+    gcs_uri: str,
+    talkgroup_name: Optional[str] = None,
+    vocabulary: Optional[list[str]] = None,
+) -> tuple[Optional[str], list[dict]]:
     """Download audio from GCS and transcribe with OpenAI Whisper."""
     from google.cloud import storage as gcs
     from google.oauth2 import service_account
@@ -94,7 +108,10 @@ def _sync_transcribe(gcs_uri: str, talkgroup_name: Optional[str] = None) -> tupl
     try:
         blob.download_to_filename(tmp_path)
 
-        prompt = (f"Talkgroup: {talkgroup_name}. " + _WHISPER_PROMPT) if talkgroup_name else _WHISPER_PROMPT
+        from app.internal.vocabulary_learner import build_whisper_vocab_prompt
+        vocab_prefix = build_whisper_vocab_prompt(vocabulary or [])
+        tg_prefix    = f"Talkgroup: {talkgroup_name}. " if talkgroup_name else ""
+        prompt       = tg_prefix + vocab_prefix + _WHISPER_PROMPT
 
         openai_client = OpenAI(api_key=settings.openai_api_key)
         with open(tmp_path, "rb") as f: