modularization

config.py

@@ -1,3 +0,0 @@
-TARS_LOGO = "images/TARS-Logo.png"
-LOGO_IMAGE_WIDTH = 360
-LOGO_IMAGE_HEIGHT = 360

discourse.py

@@ -2,31 +2,15 @@ import openai
 import os
 import azure.cognitiveservices.speech as speechsdk
 
-import csv
-from huggingface_hub import Repository
-
-DATASET_REPO_URL = "https://huggingface.co/datasets/FranklinWillemen/persistent-space-dataset"
-DATA_FILENAME = "data.csv"
-DATA_FILE = os.path.join("data", DATA_FILENAME)
-HF_TOKEN = os.environ.get("HF_TOKEN")
-
 LANGUAGE = "nl"
 AUDIO_FILE_NAME = "audio_response.wav"
 
-repo = Repository(
-    local_dir="data", clone_from=DATASET_REPO_URL, use_auth_token=HF_TOKEN
-)
-
 openai.api_key = os.environ['OPEN_AI_KEY']
 
 speech_config = speechsdk.SpeechConfig(subscription=os.environ['AZURE_SPEECH_KEY'], region="westeurope")
 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)
@@ -40,18 +24,8 @@ def gen_voice(response, response_filename):
     reponse_audio = speech_synthesizer.speak_text_async(response['content']).get()
     stream = speechsdk.AudioDataStream(reponse_audio)
     stream.save_to_wav_file(response_filename)
-
-def memory():
-    with open(DATA_FILE, "a") as csvfile:
-        for message in context:
-            writer = csv.DictWriter(csvfile, fieldnames=["name", "message"])
-            writer.writerow(
-                {"name": message['role'], "message": message['content']}
-            )
-        commit_url = repo.push_to_hub()
-        print(commit_url)
     
-def respond(audio:str):
+def respond(audio:str, context: list):
     transcript = transcribe("whisper-1", audio)
     context.append({"role": "user", "content": transcript['text']})
 
@@ -60,9 +34,10 @@ def respond(audio:str):
     
     gen_voice(response, AUDIO_FILE_NAME)
 
+    return AUDIO_FILE_NAME
+
+def transcript(context: list):
     transcript = ""
     for m in context:
         if m["role"] != "system":
-            transcript += m["role"] + " : " + m["content"] + "\n\n"
-
-    return AUDIO_FILE_NAME, transcript
+            transcript += m["role"] + " : " + m["content"] + "\n\n"

gradio-ui.py

@@ -1,5 +1,6 @@
 import gradio as gr
 import discourse as d
+import memory as m
 
 # set a custom theme
 theme = gr.themes.Default().set(
@@ -20,7 +21,7 @@ with gr.Blocks(theme=theme) as ui:
     with gr.Row():
         btn2 = gr.Button("Save Conversation")
 
-    btn1.click(fn=d.respond, inputs=message, outputs=[audio_response, text_response])
-    btn2.click(fn=d.memory)
+    btn1.click(fn=d.respond, inputs=[message, m.context], outputs=[audio_response, text_response])
+    btn2.click(fn=m.save_as_hf_dataset)
 
 ui.launch()

memory.py

@@ -0,0 +1,26 @@
+import os
+import csv
+from huggingface_hub import Repository
+
+DATASET_REPO_URL = "https://huggingface.co/datasets/FranklinWillemen/persistent-space-dataset"
+DATA_FILENAME = "data.csv"
+DATA_FILE = os.path.join("data", DATA_FILENAME)
+HF_TOKEN = os.environ.get("HF_TOKEN")
+
+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:
+            writer = csv.DictWriter(csvfile, fieldnames=["name", "message"])
+            writer.writerow(
+                {"name": message['role'], "message": message['content']}
+            )
+        commit_url = repo.push_to_hub()
+        print(commit_url)