Show/save transcipt

discourse.py

@@ -11,12 +11,16 @@ speech_config = speechsdk.SpeechConfig(subscription=os.environ['AZURE_SPEECH_KEY
 speech_config.speech_synthesis_voice_name = "nl-NL-ColetteNeural"
 speech_synthesizer = speechsdk.SpeechSynthesizer(speech_config=speech_config, audio_config=None)
 
+context = [{"role": "system", "content": 'Je bent een slimme en behulpzame gesprekspartner. \
+                                          Antwoord beknopt en ter zake.\
+                                          Vermeld niet dat je een AI of een soort service bent.'}]
+
 def transcribe(model: str, audio: str):
     audio_file = open(audio, "rb")
     transcript = openai.Audio.transcribe(model, audio_file, language=LANGUAGE)
     return transcript
 
-def gen_response(model: str, context: list):
+def gen_response(model: str):
     response = openai.ChatCompletion.create(model=model, messages=context)
     return response["choices"][0]["message"]
 
@@ -25,7 +29,7 @@ def gen_voice(response, response_filename):
     stream = speechsdk.AudioDataStream(reponse_audio)
     stream.save_to_wav_file(response_filename)
     
-def respond(audio:str, context: list):
+def respond(audio:str):
     transcript = transcribe("whisper-1", audio)
     context.append({"role": "user", "content": transcript['text']})
 
@@ -36,8 +40,10 @@ def respond(audio:str, context: list):
 
     return AUDIO_FILE_NAME
 
-def transcript(context: list):
+def transcript():
     transcript = ""
     for m in context:
         if m["role"] != "system":
-            transcript += m["role"] + " : " + m["content"] + "\n\n"
+            transcript += m["role"] + " : " + m["content"] + "\n\n"
+
+    return transcript

gradio-ui.py

@@ -12,16 +12,19 @@ with gr.Blocks(theme=theme) as ui:
         with gr.Column(scale=1):
             message = gr.Audio(source="microphone", type="filepath")
     with gr.Row():
-        btn1 = gr.Button("Respond")
+        btn1 = gr.Button("Generate Reponse")
     with gr.Row():
         with gr.Column(scale=1):
             audio_response = gr.Audio()
     with gr.Row():
         text_response = gr.Textbox(label="Transcript", max_lines=10)
     with gr.Row():
-        btn2 = gr.Button("Save Conversation")
+        btn3 = gr.Button("Show Transcript")
+    with gr.Row():
+        btn2 = gr.Button("Save Transcript")
 
-    btn1.click(fn=d.respond, inputs=[message, m.context], outputs=[audio_response, text_response])
+    btn1.click(fn=d.respond, inputs=message, outputs=audio_response)
     btn2.click(fn=m.save_as_hf_dataset)
+    btn3.click(fn=d.transcript, output=text_response)
 
 ui.launch()

memory.py

@@ -11,10 +11,6 @@ repo = Repository(
     local_dir="data", clone_from=DATASET_REPO_URL, use_auth_token=HF_TOKEN
 )
 
-context = [{"role": "system", "content": 'Je bent een slimme en behulpzame gesprekspartner. \
-                                          Antwoord beknopt en ter zake.\
-                                          Vermeld niet dat je een AI of een soort service bent.'}]
-
 def save_as_hf_dataset():
     with open(DATA_FILE, "a") as csvfile:
         for message in context: