Update backend/main.py

backend/main.py

@@ -24,6 +24,8 @@ from mongodb.operations.calls import *
 from mongodb.models.calls import UserCall, UpdateCall
 # from mongodb.endpoints.calls import *
 
+from utils.text_rank import extract_terms
+
 from transformers import AutoProcessor, SeamlessM4Tv2Model
 
 # from seamless_communication.inference import Translator
@@ -129,7 +131,6 @@ static_files = {
     },
 }
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
-# processor = AutoProcessor.from_pretrained("facebook/seamless-m4t-v2-large", force_download=True)
 processor = AutoProcessor.from_pretrained("facebook/seamless-m4t-v2-large")
 
 # PM - hardcoding temporarily as my GPU doesnt have enough vram
@@ -152,34 +153,11 @@ def get_collection_calls():
     return app.database["call_test"]
 
 
-@app.get("/test/", response_description="Welcome User")
+@app.get("/home/", response_description="Welcome User")
 def test():
-
     return {"message": "Welcome to InterpreTalk!"}
 
 
-@app.post("/test_post/", response_description="List more test call records")
-def test_post():
-    request_data = {
-        "call_id": "TESTID000001"
-    }
-
-    result = create_calls(get_collection_calls(), request_data)
-
-    # return {"message": "Welcome to InterpreTalk!"}
-    return result
-
-@app.put("/test_put/", response_description="List test call records")
-def test_put():
-
-    # result = list_calls(get_collection_calls(), 100)
-    # result = send_captions("TEST", "TEST", "TEST", "oUjUxTYTQFVVjEarIcZ0")
-    result = send_captions("TEST", "TEST", "TEST", "TESTID000001")
-
-    print(result)
-    return result
-
-
 async def send_translated_text(client_id, original_text, translated_text, room_id):
     print('SEND_TRANSLATED_TEXT IS WOKRING IN FASTAPI BACKEND...')
     print(rooms) # Debugging
@@ -207,10 +185,33 @@ async def connect(sid, environ):
     gunicorn_logger.warning(clients)
 
 @sio.on("disconnect")
-async def disconnect(sid): # BO - also pass call id as parameter for updating MongoDB
+async def disconnect(sid):
     gunicorn_logger.debug(f"📤 [event: disconnected] sid={sid}")
     clients.pop(sid, None)
-    # BO -> Update Call record with call duration, key terms
+
+@sio.on("term_extraction")
+async def term_extraction(sid, call_id):
+    gunicorn_logger.debug(f"📤 [event: term_extraction] sid={sid}, call={call_id}")
+
+    # call_id = "0FIdAosKy9ysQDkp14T2"
+
+    # Get combined caption field for call record based on call_id
+    combined_text = get_caption_text(get_collection_calls(), call_id)
+
+    if combined_text: # > min_caption_length: -> poor term extraction on short 
+        print("THE COMBINED TEXT IS:", combined_text)
+
+        # Extract Key Terms from Concatenated Caption Field
+        key_terms = extract_terms(combined_text, len(combined_text))
+
+        # BO -> Update Call record with call duration, key terms
+        print("THE KEY TERMS ARE:", key_terms)
+
+        request_data = {
+            "key_terms": key_terms
+        }
+
+        update_calls(get_collection_calls(), call_id, request_data)
 
 @sio.on("target_language")
 async def target_language(sid, target_lang):
@@ -232,16 +233,16 @@ async def call_user(sid, call_id):
     # BO - Get call id from dictionary created during socketio connection
     client_id = clients[sid].client_id
     
-    # gunicorn_logger.warning(f"NOW TRYING TO CREATE DB RECORD FOR Caller with ID: {client_id} for call: {call_id}")
-    # # BO -> Create Call Record with Caller and call_id field (None for callee, duration, terms..)
-    # request_data = {
-    #     "call_id": str(call_id),
-    #     "caller_id": str(client_id),
-    #     "creation_date": str(datetime.now())
-    # }
+    gunicorn_logger.warning(f"NOW TRYING TO CREATE DB RECORD FOR Caller with ID: {client_id} for call: {call_id}")
+    # BO -> Create Call Record with Caller and call_id field (None for callee, duration, terms..)
+    request_data = {
+        "call_id": str(call_id),
+        "caller_id": str(client_id),
+        "creation_date": str(datetime.now())
+    }
 
-    # response =  create_calls(get_collection_calls(), request_data)
-    # print(response) # BO - print created db call record
+    response =  create_calls(get_collection_calls(), request_data)
+    print(response) # BO - print created db call record
 
 @sio.on("audio_config")
 async def audio_config(sid, sample_rate):
@@ -265,27 +266,27 @@ async def answer_call(sid, call_id):
     # BO - Get call id from dictionary created during socketio connection
     client_id = clients[sid].client_id
     
-    # # BO -> Update Call Record with Callee field based on call_id
-    # gunicorn_logger.warning(f"NOW UPDATING MongoDB RECORD FOR Caller with ID: {client_id} for call: {call_id}")
-    # # BO -> Create Call Record with callee_id field (None for callee, duration, terms..)
-    # request_data = {
-    #     "callee_id": client_id
-    # }
-
-    # response =  update_calls(get_collection_calls(), call_id, request_data)
-    # print(response) # BO - print created db call record
+    # BO -> Update Call Record with Callee field based on call_id
+    gunicorn_logger.warning(f"NOW UPDATING MongoDB RECORD FOR Caller with ID: {client_id} for call: {call_id}")
+    # BO -> Create Call Record with callee_id field (None for callee, duration, terms..)
+    request_data = {
+        "callee_id": client_id
+    }
+
+    response =  update_calls(get_collection_calls(), call_id, request_data)
+    print(response) # BO - print created db call record
     
 
 @sio.on("incoming_audio")
 async def incoming_audio(sid, data, call_id):
-    gunicorn_logger.info("RUNNNING INCOMING AUDIO FUNCTION")
     try:
         clients[sid].add_bytes(data)
 
         if clients[sid].get_length() >= MAX_BYTES_BUFFER:
             gunicorn_logger.info('Buffer full, now outputting...')
             output_path = clients[sid].output_path
-            vad_result, resampled_audio = clients[sid].resample_and_write_to_file()
+            resampled_audio = clients[sid].resample_and_clear()
+            vad_result = clients[sid].vad_analyse(resampled_audio)
             # source lang is speakers tgt language 😃
             src_lang = clients[sid].target_language
             if vad_result:
@@ -304,21 +305,18 @@ async def incoming_audio(sid, data, call_id):
                 translated_data = model.generate(**t2t_tokens, tgt_lang=tgt_lang, generate_speech=False)[0].tolist()[0]
                 translated_text = processor.decode(translated_data, skip_special_tokens=True)
                 print(f"TRANSLATED TEXT = {translated_text}")
-
-                # BO -> send translated_text to mongodb as caption record update based on call_id
-                # send_captions(clients[sid].client_id, asr_text, translated_text, call_id)
                 
                 # TRANSLATED TEXT
                 # PM - text_output is a list with 1 string
                 await send_translated_text(clients[sid].client_id, asr_text, translated_text, call_id)
 
                 # BO -> send translated_text to mongodb as caption record update based on call_id
-                # send_captions(clients[sid].client_id, asr_text, translated_text, call_id)
+                await send_captions(clients[sid].client_id, asr_text, translated_text, call_id)
 
     except Exception as e:
         gunicorn_logger.error(f"Error in incoming_audio: {e.with_traceback()}")
     
-def send_captions(client_id, original_text, translated_text, call_id):
+async def send_captions(client_id, original_text, translated_text, call_id):
     # BO -> Update Call Record with Callee field based on call_id    
     print(f"Now updating Caption field in call record for Caller with ID: {client_id} for call: {call_id}")