[apps/hal9] Run generic replies in parallel follow up optimizations (#…

…465)

apps/hal9/app.py

@@ -1,21 +1,20 @@
 from utils import generate_response, load_messages, insert_message, execute_function, save_messages, insert_tool_message, is_url, download_file, generate_text_embeddings_parquet
 from tools.calculator import solve_math_problem_description, solve_math_problem
-from tools.generic import answer_generic_wrapper_description, answer_generic_wrapper, answer_generic_question_submit
-from tools.generic import answer_generic_question, answer_generic_question_description
+from tools.generic import answer_generic_question, answer_generic_question_description, answer_generic_future
 from tools.csv_agent import analyze_csv_description, analyze_csv
 from tools.image_agent import images_management_system, images_management_system_description, add_images_descriptions
 from tools.hal9 import answer_hal9_questions_description, answer_hal9_questions
 from tools.text_agent import analyze_text_file_description, analyze_text_file
 from tools.streamlit import streamlit_generator, streamlit_generator_description
 from tools.website import website_generator, website_generator_description
-from concurrent.futures import ThreadPoolExecutor, Future
+from concurrent.futures import ThreadPoolExecutor
 
 # load messages
 messages = load_messages()
 
 # load tools
-tools_descriptions = [answer_generic_wrapper_description, solve_math_problem_description, analyze_csv_description, images_management_system_description, answer_hal9_questions_description, analyze_text_file_description, streamlit_generator_description, website_generator_description]
-tools_functions = [answer_generic_wrapper, solve_math_problem, analyze_csv, images_management_system, answer_hal9_questions, analyze_text_file, streamlit_generator, website_generator]
+tools_descriptions = [answer_generic_question_description, solve_math_problem_description, analyze_csv_description, images_management_system_description, answer_hal9_questions_description, analyze_text_file_description, streamlit_generator_description, website_generator_description]
+tools_functions = [answer_generic_question, solve_math_problem, analyze_csv, images_management_system, answer_hal9_questions, analyze_text_file, streamlit_generator, website_generator]
 
 if len(messages) < 1:
     messages = insert_message(messages, "system", "You are Hal9, a helpful and highly capable AI assistant. Your primary responsibility is to analyze user questions and select the most appropriate tool to provide precise, relevant, and actionable responses. Always prioritize using the right tool to ensure efficiency and clarity in your answers.")
@@ -34,16 +33,18 @@
     print(f"I'm ready to answer questions about your file: {filename}")
 else:
     user_input = user_input.replace("\f", "\n")
+    messages = insert_message(messages, "user", user_input)
 
     with ThreadPoolExecutor() as executor:
-        answer_generic_question_submit(executor, user_input)
+        answer_generic_future = executor.submit(answer_generic_question, user_input)
 
-        messages = insert_message(messages, "user", user_input)
+        response = executor.submit(generate_response, "openai", "gpt-4-turbo", messages, tools_descriptions, tool_choice="required", parallel_tool_calls=False)
 
-        response = generate_response("openai", "gpt-4-turbo", messages, tools_descriptions, tool_choice = "required", parallel_tool_calls=False)
+    response_result = response.result()
 
-        tool_result = execute_function(response, tools_functions)
+    tool_result = execute_function(response_result, tools_functions)
 
-        insert_tool_message(messages, response, tool_result)
+    if tool_result is not None:
+        insert_tool_message(messages, response_result, tool_result)
 
 save_messages(messages)

apps/hal9/tools/generic.py

@@ -1,21 +1,33 @@
 from groq import Groq
-from utils import stream_print, load_messages, insert_message, save_messages
+from utils import load_messages, insert_message, save_messages
+from concurrent.futures import ThreadPoolExecutor
 
-def answer_generic_question(user_input, show = True):
-    # load messages
+answer_generic_future = None
+
+def answer_generic_question(user_input):
+    global answer_generic_future
+    if answer_generic_future is not None:
+        text_response=answer_generic_future.result()
+        print(text_response)
+        return text_response
+    else:
+        with ThreadPoolExecutor() as executor:
+            answer_generic_future = executor.submit(process_request, user_input)
+            text_response = answer_generic_future.result()
+            return text_response
+
+def process_request(user_input):
     messages = load_messages(file_path="./.storage/.generic_agent_messages.json")
     messages = insert_message(messages, "user", user_input)
-    stream = Groq().chat.completions.create(
-        model = "llama3-70b-8192",
-        messages = messages,
-        temperature = 0,
-        seed = 1,
-        stream = True)
-
-    text_response = stream_print(stream, show)
+    response = Groq().chat.completions.create(
+        model="llama3-70b-8192",
+        messages=messages,
+        temperature=0,
+        seed=1
+    )
+    text_response = response.choices[0].message.content
     messages = insert_message(messages, "assistant", text_response)
     save_messages(messages, file_path="./.storage/.generic_agent_messages.json")
-
     return text_response
 
 answer_generic_question_description = {
@@ -36,33 +48,4 @@ def answer_generic_question(user_input, show = True):
             "additionalProperties": False,
         },
     }
-}
-
-answer_generic_future = None
-def answer_generic_wrapper(user_input):
-    result = answer_generic_future.result()
-    print(result)
-
-def answer_generic_question_submit(executor, user_input):
-    global answer_generic_future
-    answer_generic_future = executor.submit(answer_generic_question, user_input, False)
-
-answer_generic_wrapper_description = {
-    "type": "function",
-    "function": {
-        "name": "answer_generic_wrapper",
-        "description": "Handles general questions or queries provided by the user by taking their input and generating a meaningful response.",
-        "strict": True,
-        "parameters": {
-            "type": "object",
-            "properties": {
-                "user_input": {
-                    "type": "string",
-                    "description": "Take the user input and pass the same string to the function",
-                },
-            },
-            "required": ["user_input"],
-            "additionalProperties": False,
-        },
-    }
 }

apps/hal9/tools/hal9.py

@@ -3,14 +3,13 @@
 from utils import stream_print
 
 def answer_hal9_questions(user_input):
-    stream = Groq().chat.completions.create(
+    response = Groq().chat.completions.create(
         model = "llama3-70b-8192",
         messages = [{"role": "system", "content": DATA["hal9"]},{"role": "user", "content": user_input}],
         temperature = 0,
-        seed = 1,
-        stream = True)
+        seed = 1,)
 
-    text_response = stream_print(stream)
+    text_response = print(response.choices[0].message.content)
     return text_response
 
 answer_hal9_questions_description = {

apps/hal9/utils.py

@@ -3,7 +3,6 @@
 import urllib.parse
 import urllib.request
 import requests
-import time
 from typing import Literal, List, Dict, Any, Union, Optional
 from clients import openai_client, azure_openai_client
 from groq import Groq
@@ -90,7 +89,7 @@ def generate_response(
         "n": n
     }
 
-    if tools is not None and not client_type == "groq":
+    if tools is not None:
         payload["parallel_tool_calls"] = parallel_tool_calls
 
     # Generate the response using the client's completion API.
@@ -105,9 +104,6 @@ def load_messages(file_path="./.storage/.messages.json") -> List[Dict[str, Any]]
     Returns:
         List[Dict[str, Any]]: A list of messages if the file exists and is valid.
     """
-    # Create the .storage directory if it doesn't exist
-    if not os.path.exists("./.storage"):
-        os.makedirs("./.storage")
     if not os.path.exists(file_path):
         return []
     else :
@@ -124,9 +120,6 @@ def save_messages(messages: List[Dict[str, Any]], file_path="./.storage/.message
         messages (List[Dict[str, Any]]): A list of messages to be saved.
     """
 
-    # Create the .storage directory if it doesn't exist
-    if not os.path.exists("./.storage"):
-        os.makedirs("./.storage")
     with open(file_path, "w", encoding="utf-8") as file:
         json.dump(messages, file, ensure_ascii=False, indent=4)
 
@@ -219,9 +212,6 @@ def is_url(prompt):
   return all([result.scheme, result.netloc])
 
 def download_file(url):
-    # Create the .storage directory if it doesn't exist
-    if not os.path.exists("./.storage"):
-        os.makedirs("./.storage")
     filename = url.split("/")[-1]
     modified_filename = f"./.storage/.{filename}"
 
@@ -283,9 +273,6 @@ def process_chunk(chunk_info):
     }
 
 def generate_text_embeddings_parquet(url, model="text-embedding-3-small", client_type="azure", n_words=300, overlap=0, max_threads=8):
-    # Create the .storage directory if it doesn't exist
-    if not os.path.exists("./.storage"):
-        os.makedirs("./.storage")
     # Download and read the PDF
     response = requests.get(url)
     pdf_document = fitz.open(stream=BytesIO(response.content))