Add voice

backend/app/services.py

@@ -88,6 +88,15 @@ def build_llm_reply(db: Session, student_id: int, user_message: str) -> str:
     return response.output_text.strip()
 
 
+def transcribe_audio(filename: str, audio_bytes: bytes, content_type: str | None = None) -> str:
+    file_payload = (filename, audio_bytes, content_type or "application/octet-stream")
+    transcript = client.audio.transcriptions.create(
+        model="gpt-4o-mini-transcribe",
+        file=file_payload,
+    )
+    return (getattr(transcript, "text", "") or "").strip()
+
+
 def pick_next_skill(db: Session, student_id: int) -> models.Skill:
     weakest = (
         db.query(models.StudentSkillMastery)