final touches

app.py

@@ -1,9 +1,12 @@
+import logging
 import gradio as gr
 import openai
 from constants import *
+import string
 
 
 openai.api_key = OPENAI_API_KEY
+openai.organization = OPENAI_ORGANIZATION
 
 title = "Car Seats Voice Commands"
 
@@ -30,25 +33,27 @@ Command Classification phase. For Transcription I used The OpenAi whisper model,
 I Fine-Tuned the OpenAi **ada** model on Car Seats Command.
 """
 
+def remove_punctuation(input_string):
+    translator = str.maketrans('', '', string.punctuation)
+    clean_string = input_string.translate(translator)
 
-# def get_command(command, model, id2label):
-#     """
-#     This function get the classification outputs from openai API
-#     """
-#     completion = openai.Completion.create(
-#         model=model, prompt=f"{command}->", max_tokens=1, temperature=0
-#     )
-#     id = int(completion["choices"][0]["text"].strip())
-#     result = id2label[id] if id in id2label else "unknown"
-#     return result
+    return clean_string
 
+id2label = {
+    1:"massage_seats_on",
+    2:"massage_seats_off",
+    3:"heated_seats_on",
+    4:"heated_seats_off",
+    5:"cooled_seats_on",
+    6:"cooled_seats_off"
+}
 
-def get_command(command, model, id2label):
+def get_command(command, id2label, model = "text-davinci-003"):
     """
     This function get the classification outputs from openai API
     """
     prompt = f"""
-    We want to control the seats of a car which has features to cool, heat, or massage a seat. The user said "{command}", Which feature we should use to ensure user comfort? Give just the number of the feature.
+    We want to control the seats of a car which has features to cool, heat, or massage a seat. The user said "{command}", Which feature we should use to ensure user comfort? Give just the number of the feature without any punctuation.
     Mapping:
     1: "massage_seats_on"
     2: "massage_seats_off"
@@ -61,51 +66,72 @@ def get_command(command, model, id2label):
     """
 
     completion = openai.Completion.create(
-        model="text-davinci-003", prompt=prompt, max_tokens=2, temperature=0
+        model=model, prompt=prompt, max_tokens=2, temperature=0
     )
-    id = int(completion["choices"][0]["text"].strip())
+
+    print("result")
+    print(completion["choices"][0]["text"].strip())
+
+    id = int(remove_punctuation(completion["choices"][0]["text"]).strip())
 
     result = id2label[id] if id in id2label else "unknown"
     return result
 
 
-def transcribe(audio, text):
+def command_tokens(command, model = "text-davinci-003"):
+    """
+    This function get the classification outputs from openai API
+    """
+    prompt = f"""
+        Give an array of the same length of the input, for every element of the returned array use  one of the labels in the label-list
+
+        label-list :
+        - unit if belongs to the International System of Units
+        - value
+        - none if none of the above
+
+        input : [{",".join(command.split(" "))}]
+        output :
+    """
+
+    completion = openai.Completion.create(
+        model=model, prompt=prompt, max_tokens=128, temperature=0
+    )
+
+    result = completion["choices"][0]["text"].strip()
+
+    result_list = result.replace("[", "").replace("]", "").replace("'", "").split(',')
+
+    return list(zip(command.split(" "), result_list))
+
+
+def transcribe(audio):
     """
     if text provided the function will classify the input directly.
     if not the audio will be transcribed then the transcription will be classified.
-    """
 
-    if text:
-        result = get_command(text, MODEL, id2label)
-        return "Text provided by the user", text_respnses[result], None
+    return a tuple of (transcription, command, audio to be played)
+    """
 
     # getting text transcription
     audio_file = open(audio, "rb")
     transcription = openai.Audio.transcribe("whisper-1", audio_file, language="en")
     transcription = transcription["text"]
 
-    result = get_command(transcription, MODEL, id2label)
-    audio_res = resoponses.get(result)()
+    result = get_command(transcription, id2label)
+    tokens = command_tokens(transcription)
+
+    print("result", result)
+    print("tokens", tokens)
 
-    return transcription, text_respnses[result], audio_res
+    return result, tokens
 
 
-if __name__ == "__main__":
+if __name__=="__main__":
     gr.Interface(
-        fn=transcribe,
-        inputs=[
-            gr.Audio(label="", source="microphone", type="filepath"),
-            gr.Textbox(label="If you prefer type your command (more accurate)"),
-        ],
-        outputs=[
-            gr.Textbox(
-                label="Input Transcription (Please check that this matches what you've said)"
-            ),
-            gr.Textbox(label="Machine Response (Text Version)"),
-            gr.Audio(label="Machine Response (Audio Version)"),
-        ],
-        allow_flagging="auto",
-        title=title,
-        description=description,
-        article=article,
+    fn=transcribe,
+    inputs=gr.Audio(source="microphone", type="filepath"),
+    outputs=["text", "highlight"],
+    title=title,
+    description=description
     ).launch()

constants.py

@@ -1,3 +1,4 @@
+from token import OP
 from commands import ResponseManager
 from dotenv import load_dotenv
 import os
@@ -5,6 +6,7 @@ import os
 load_dotenv()
 
 OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
+OPENAI_ORGANIZATION = os.getenv("OPENAI_ORGANIZATION")
 MODEL = os.getenv("MODEL")
 
 rs = ResponseManager()

flagged/log.csv

@@ -1,38 +0,0 @@
-audio,output,flag,username,timestamp
-C:\Users\admin\Desktop\DS JRP\Programming\Python\Portfolio\Voice_Commands\flagged\audio\tmp7zujjp6r.wav,,,,2023-02-18 19:27:17.163146
-C:\Users\admin\Desktop\DS JRP\Programming\Python\Portfolio\Voice_Commands\flagged\audio\tmpxmty_reu.wav,,,,2023-02-18 19:27:19.327085
-C:\Users\admin\Desktop\DS JRP\Programming\Python\Portfolio\Voice_Commands\flagged\component 0\tmp4mx6cxo6.wav,,,,2023-02-19 19:21:00.276808
-C:\Users\admin\Desktop\DS JRP\Programming\Python\Portfolio\Voice_Commands\flagged\component 0\tmpj5vph7ui.wav,,,,2023-02-19 19:22:46.434333
-,my neck is cold,,,2023-02-19 19:28:56.486805
-,my neck  need no more heating,,,2023-02-19 19:29:16.234959
-C:\Users\admin\Desktop\DS JRP\Programming\Python\Portfolio\Voice_Commands\flagged\component 0\tmp7zfrxwn1.wav,,,,2023-02-19 20:28:51.150010
-C:\Users\admin\Desktop\DS JRP\Programming\Python\Portfolio\Voice_Commands\flagged\component 0\tmpj9jcosbt.wav,,,,2023-02-19 20:29:16.601422
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmpgp91h4bs.wav,,,,2023-03-03 10:48:05.204488
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmpzxnv34e9.wav,,,,2023-03-03 11:00:04.470118
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp293krtcc.wav,,,,2023-03-03 11:05:18.409853
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp8w_dg0my.wav,,,,2023-03-03 11:06:13.151883
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp8uwv1dlh.wav,,,,2023-03-03 11:08:34.437979
-,,,,2023-03-03 11:09:59.046011
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmpwerjyrhm.wav,,,,2023-03-03 11:10:14.290251
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp1msp3gk7.wav,,,,2023-03-03 11:11:04.122287
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmpic0klrj5.wav,,,,2023-03-03 11:11:17.671385
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmpvb2y2kk5.wav,,,,2023-03-03 11:11:31.502951
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp_sv03ta9.wav,,,,2023-03-03 11:11:52.200257
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmpl6bjw0sw.wav,,,,2023-03-03 11:17:37.575339
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp7272239u.wav,,,,2023-03-03 11:18:02.333990
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp4fwibr34.wav,,,,2023-03-03 11:18:18.551143
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp3qcgcacy.wav,,,,2023-03-03 11:18:35.923342
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmplih8n03h.wav,,,,2023-03-03 11:51:57.253346
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmpx7nz_jih.wav,,,,2023-03-03 11:52:11.455135
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmpflbx1ja4.wav,,,,2023-03-03 11:52:28.437100
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmpcrult4_q.wav,,,,2023-03-03 11:52:39.450544
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmps1x6dtf2.wav,,,,2023-03-03 11:53:52.432486
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp_sseh3hk.wav,,,,2023-03-03 11:54:09.292486
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp68h39ev5.wav,,,,2023-03-03 11:54:26.488504
-,I'm cold,,,2023-03-04 10:42:40.559129
-,I'm cold,,,2023-03-04 10:44:49.418487
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmpz0hq94rg.wav,,,,2023-03-04 11:32:53.276143
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmpr_cd3uqa.wav,,,,2023-03-04 11:33:22.605201
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp8pxcjozd.wav,,,,2023-03-04 11:34:02.384882
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp7vog58pt.wav,,,,2023-03-04 11:34:32.302966
-/home/boumaza/Desktop/Programming/Python/Voice_Commands/flagged/component 0/tmp5v5ipo64.wav,,,,2023-03-04 11:34:50.524003