speech + assistant fix

main.py

@@ -2,6 +2,8 @@ import logging
 import os
 import json
 import re
+import uuid
+import random
 
 import openai
 import requests
@@ -54,6 +56,19 @@ def download_file(name):
 def test():
     return jsonify({"status": "0"})
 
+def roulette() -> str:
+    roulette_arr = [(80, "gpt-3.5-turbo"), (20, "gpt-4o")]
+    rand_num = random.randrange(0, 99)
+    model_name = ""
+    n = 0
+    j = 0
+    while rand_num > n:
+        n += roulette_arr[j][0]
+        model_name = roulette_arr[j][1]
+        print(model_name)
+        j += 1
+    return model_name
+
 
 def recommend(headlines, category):
     chat_messages = app.recommend_messages.copy()
@@ -124,6 +139,7 @@ def vision(message, image_url=None, image_b64=None):
 @app.route('/gpt', methods=['POST'])
 def gpt():
     assistant_id = ""
+    assistant = None
     chat_messages = []
     chat_model = "gpt-3.5-turbo"
     use_video = False
@@ -145,10 +161,10 @@ def gpt():
         if 'payload' in json_payload:
             if 'predict_q' in json_payload:
                 predict_q = 5 if json_payload['predict_q'] > 4 else 0 if json_payload['predict_q'] < 1 else \
-                json_payload['predict_q']
+                    json_payload['predict_q']
             if 'num_choices' in json_payload:
                 num_choices = 5 if json_payload['num_choices'] > 4 else 1 if json_payload['num_choices'] < 2 else \
-                json_payload['num_choices']
+                    json_payload['num_choices']
             if 'use_video' in json_payload:
                 use_video = json_payload['use_video'] == "1"
             if 'chat_model' in json_payload and 'assistant_id' not in json_payload:
@@ -178,6 +194,8 @@ def gpt():
                                                          json_payload['payload'][-1]['content']
             elif 'assistant_id' in json_payload:
                 assistant_id = json_payload['assistant_id']
+                assistant = app.openai_client.beta.assistants.retrieve(assistant_id=assistant_id)
+                chat_model = assistant.model
             else:
                 chat_messages = app.chat_messages.copy()
             json_payload = json_payload['payload']
@@ -239,6 +257,9 @@ def gpt():
     result = {}
     try:
         n = num_choices
+        if "gpt-3.5-turbo" in chat_model:
+            chat_model = roulette()
+        app.logger.info(f"Model used: {chat_model}")
         if assistant_id and not suggest:
             runs = app.openai_client.beta.threads.create_and_run_poll(
                 assistant_id=assistant_id,
@@ -275,17 +296,56 @@ def gpt():
             else:
                 result = {"role": "assistant", "content": json_response.choices[0].message.content}
         if predict_q:
-            query_q = {
-                "role": "user",
-                "content": f"Berikan {predict_q} pertanyaan lain yang akan saya ajukan berdasarkan percakapan kali ini dalam bentuk json array"
-            }
+            if assistant_id:
+                query_q = {
+                    "role": "user",
+                    "content": f"Berikan {predict_q} pertanyaan random yang akan saya ajukan sesuai topik asisten dalam bentuk json array"
+                }
+            else:
+                query_q = {
+                    "role": "user",
+                    "content": f"Berikan {predict_q} pertanyaan lain yang akan saya ajukan berdasarkan percakapan kali ini dalam bentuk json array"
+                }
             chat_messages.append(result)
             chat_messages.append(query_q)
-            json_response_q = app.openai_client.chat.completions.create(model=chat_model,
-                                                                        messages=chat_messages,
-                                                                        max_tokens=max_resp_token,
-                                                                        temperature=0.2,
-                                                                        response_format={"type": "json_object"})
+            if assistant_id:
+                runs = app.openai_client.beta.threads.create_and_run_poll(
+                    assistant_id=assistant_id,
+                    thread={
+                        "messages": chat_messages
+                    }
+                )
+                messages = list(app.openai_client.beta.threads.messages.list(thread_id=runs.thread_id, run_id=runs.id))
+                message_content = messages[0].content[0].text
+                app.logger.info(message_content.value)
+                pattern = re.compile(r"【\d+:\d+†\(?source\)?】")
+                filtered_message = pattern.sub("", message_content.value)
+                predict_q_arr = [
+                    {
+                        "role": "system",
+                        "content": assistant.instructions
+                    },
+                    {
+                        "role": "assistant",
+                        "content": filtered_message
+                    },
+                    {
+                        "role": "user",
+                        "content": f"Ekstrak {predict_q} pertanyaan tersebut dalam bentuk json array"
+                    }
+                ]
+                json_response_q = app.openai_client.chat.completions.create(
+                    model=chat_model,
+                    messages=predict_q_arr,
+                    temperature=0.2,
+                    response_format={"type": "json_object"}
+                )
+            else:
+                json_response_q = app.openai_client.chat.completions.create(model=chat_model,
+                                                                            messages=chat_messages,
+                                                                            max_tokens=max_resp_token,
+                                                                            temperature=0.2,
+                                                                            response_format={"type": "json_object"})
             json_response_dict = json.loads(json_response_q.choices[0].message.content)
             if json_response_dict is not None:
                 if isinstance(json_response_dict, dict):
@@ -384,6 +444,14 @@ def assistant_create():
     model_name = request_form.pop('model_name', model_name)
     vector_store_id = request_form.pop('vector_store_id', "")
     file_batch_id = ""
+    try:
+        temperature = float(request_form.pop('temperature', 1.0))
+        if temperature < 0.0:
+            temperature = 0.0
+        elif temperature > 1.0:
+            temperature = 1.0
+    except ValueError:
+        temperature = 1.0
     tool_resources = {"tool_resources": {"file_search": {"vector_store_ids": [vector_store_id]}}} \
         if vector_store_id \
         else {}
@@ -393,6 +461,7 @@ def assistant_create():
             instructions=assistant_ins,
             model=model_name,
             tools=[{"type": "file_search"}],
+            temperature=temperature,
             **tool_resources,
             **request_form
         )
@@ -465,7 +534,8 @@ def assistant_att():
             )
             return {"status": file_batch.status, "vector_store_id": vector_store.id, "file_batch_id": file_batch.id}
         else:
-            file_batch = app.openai_client.beta.vector_stores.file_batches.retrieve(file_batch_id, vector_store_id=vector_store_id)
+            file_batch = app.openai_client.beta.vector_stores.file_batches.retrieve(file_batch_id,
+                                                                                    vector_store_id=vector_store_id)
             return {"status": file_batch.status}
     except Exception as e:
         app.logger.exception("error")
@@ -495,6 +565,7 @@ def assistant_update(aid=None, vid=None):
         app.logger.exception("error")
         return {"status": "error", "message": "Update assistant failed, please try again"}
 
+
 @app.route('/llama', methods=['POST'])
 def llama():
     max_char_msg = 500
@@ -541,6 +612,29 @@ def llama():
     return result
 
 
+@app.route('/speech', methods=['POST'])
+def speech(text=""):
+    if not text and 'text' not in request.form:
+        audio_file = request.files.get('audio')
+        res = app.openai_client.audio.transcriptions.create(
+            model="whisper-1",
+            file=(audio_file.filename, audio_file.stream.read())
+        )
+        return {"status": "ok", "message": res.text}
+    elif 'text' in request.form or text:
+        text = request.form['text'] if 'text' in request.form else text
+        uu_id = str(uuid.uuid4())
+        print(text)
+        with app.openai_client.audio.speech.with_streaming_response.create(
+            model="tts-1-hd",
+            voice="echo",
+            speed=0.8,
+            input=text
+        ) as res:
+            res.stream_to_file(os.path.join(app.config['UPLOAD_FOLDER'], f"{uu_id}.mp3"))
+        return download_file(f"{uu_id}.mp3")
+
+
 # Press the green button in the gutter to run the script.
 if __name__ == '__main__':
     app.run(host='0.0.0.0', port=8348, debug=True, ssl_context=ssl)