Update app.py

app.py

@@ -3,15 +3,17 @@ import subprocess
 import random
 from huggingface_hub import InferenceClient
 import gradio as gr
-from safe_search import safe_search  # Make sure you have this function defined
+from safe_search import safe_search
 from i_search import google
 from i_search import i_search as i_s
 from datetime import datetime
 import logging
 import json
-import nltk  # Import nltk for the generate_text_chunked function
+import nltk
+from transformers import pipeline
 
-nltk.download('punkt')  # Download the punkt tokenizer if you haven't already
+# Ensure NLTK data is downloaded
+nltk.download('punkt')
 
 now = datetime.now()
 date_time_str = now.strftime("%Y-%m-%d %H:%M:%S")
@@ -32,11 +34,9 @@ agents = [
     "AI_SYSTEM_PROMPT",
     "PYTHON_CODE_DEV"
 ]
-############################################
 
 VERBOSE = True
 MAX_HISTORY = 5
-# MODEL = "gpt-3.5-turbo"  # "gpt-4"
 
 PREFIX = """
 {date_time_str}
@@ -82,14 +82,14 @@ What do you want to know about the test results?
 thought: 
 """
 
-def format_prompt(message, history, max_history_turns=2):
-  prompt = "<s>"
-  # Keep only the last 'max_history_turns' turns
-  for user_prompt, bot_response in history[-max_history_turns:]:
-    prompt += f"[INST] {user_prompt} [/INST]"
-    prompt += f" {bot_response}</s> "
-  prompt += f"[INST] {message} [/INST]"
-  return prompt
+def format_prompt(message, history, max_history_turns=5):
+    prompt = "<s>"
+    # Keep only the last 'max_history_turns' turns
+    for user_prompt, bot_response in history[-max_history_turns:]:
+        prompt += f"[INST] {user_prompt} [/INST]"
+        prompt += f" {bot_response}</s> "
+    prompt += f"[INST] {message} [/INST]"
+    return prompt
 
 def run_gpt(
     prompt_template,
@@ -111,12 +111,31 @@ def run_gpt(
     
     resp = client.text_generation(content, max_new_tokens=max_tokens, stop_sequences=stop_tokens, temperature=0.7, top_p=0.8, repetition_penalty=1.5)
     if VERBOSE:
-        logging.info(LOG_RESPONSE.format(resp))  # Log the response
+        logging.info(LOG_RESPONSE.format(resp=resp))  # Log the response
     return resp
 
-def generate(prompt, history, agent_name=agents[0], sys_prompt="", temperature=0.7, max_new_tokens=2048, top_p=0.8, repetition_penalty=1.5, model="mistralai/Mixtral-8x7B-Instruct-v0.1"):
-    # Use 'prompt' here instead of 'message'
-    formatted_prompt = format_prompt(prompt, history, max_history_turns=5)  # Truncated history
+def generate(
+    prompt, history, agent_name=agents[0], sys_prompt="", temperature=0.7, max_new_tokens=2048, top_p=0.8, repetition_penalty=1.5, model="mistralai/Mixtral-8x7B-Instruct-v0.1"
+):
+    seed = random.randint(1,1111111111111111)
+
+    if agent_name == "WEB_DEV":
+        agent = "You are a helpful AI assistant. You are a web developer."
+    elif agent_name == "AI_SYSTEM_PROMPT":
+        agent = "You are a helpful AI assistant. You are an AI system."
+    elif agent_name == "PYTHON_CODE_DEV":
+        agent = "You are a helpful AI assistant. You are a Python code developer."
+    else:
+        agent = "You are a helpful AI assistant."
+
+    system_prompt = agent
+    temperature = float(temperature)
+    if temperature < 1e-2:
+        temperature = 1e-2
+    top_p = float(top_p)
+
+    formatted_prompt = f"{system_prompt} {prompt}"
+    formatted_prompt = format_prompt(formatted_prompt, history, max_history_turns=5)  # Truncated history
     logging.info(f"Formatted Prompt: {formatted_prompt}")
     stream = client.text_generation(formatted_prompt, temperature=temperature, max_new_tokens=max_new_tokens, top_p=top_p, repetition_penalty=repetition_penalty, stream=True, details=True, return_full_text=False)
     resp = ""
@@ -124,10 +143,9 @@ def generate(prompt, history, agent_name=agents[0], sys_prompt="", temperature=0
         resp += response.token.text
 
     if VERBOSE:
-        logging.info(LOG_RESPONSE.format(resp))  # Log the response
+        logging.info(LOG_RESPONSE.format(resp=resp))  # Log the response
     return resp
 
-
 def compress_history(purpose, task, history, directory):
     resp = run_gpt(
         COMPRESS_HISTORY_PROMPT,
@@ -143,7 +161,6 @@ def compress_history(purpose, task, history, directory):
 def call_search(purpose, task, history, directory, action_input):
     logging.info(f"CALLING SEARCH: {action_input}")
     try:
-        
         if "http" in action_input:
             if "<" in action_input:
                 action_input = action_input.strip("<")
@@ -151,7 +168,6 @@ def call_search(purpose, task, history, directory, action_input):
                 action_input = action_input.strip(">")
             
             response = i_s(action_input)
-            #response = google(search_return)
             logging.info(f"Search Result: {response}")
             history += "observation: search result is: {}\n".format(response)
         else:
@@ -178,7 +194,6 @@ def call_main(purpose, task, history, directory, action_input):
             history += "{}\n".format(line)
             logging.info(f"Thought: {line}")
         elif line.startswith("action: "):
-            
             action_name, action_input = parse_action(line)
             logging.info(f"Action: {action_name} - {action_input}")
             history += "{}\n".format(line)
@@ -190,13 +205,8 @@ def call_main(purpose, task, history, directory, action_input):
         else:
             history += "{}\n".format(line)
             logging.info(f"Other Output: {line}")
-            #history += "observation: the following command did not produce any useful output: '{}', I need to check the commands syntax, or use a different command\n".format(line)
-            
-            #return action_name, action_input, history, task
-            #assert False, "unknown action: {}".format(line)
     return "MAIN", None, history, task
 
-
 def call_set_task(purpose, task, history, directory, action_input):
     logging.info(f"CALLING SET_TASK: {action_input}")
     task = run_gpt(
@@ -220,7 +230,6 @@ NAME_TO_FUNC = {
     "UPDATE-TASK": call_set_task,
     "SEARCH": call_search,
     "COMPLETE": end_fn,
-
 }
 
 def run_action(purpose, task, history, directory, action_name, action_input):
@@ -248,14 +257,11 @@ def run_action(purpose, task, history, directory, action_name, action_input):
         logging.error(f"Error in run_action: {e}")
         return "MAIN", None, history, task
 
-def run(purpose,history):
-    
-    #print(purpose)
-    #print(hist)
-    task=None
-    directory="./"
+def run(purpose, history):
+    task = None
+    directory = "./"
     if history:
-        history=str(history).strip("[]")
+        history = str(history).strip("[]")
     if not history:
         history = ""
     
@@ -278,61 +284,8 @@ def run(purpose,history):
             action_input,
         )
         yield (history)
-        #yield ("",[(purpose,history)])
         if task == "END":
             return (history)
-            #return ("", [(purpose,history)])
-
-
-
-################################################
-
-def format_prompt(message, history, max_history_turns=5):
-  prompt = "<s>"
-  # Keep only the last 'max_history_turns' turns
-  for user_prompt, bot_response in history[-max_history_turns:]:
-    prompt += f"[INST] {user_prompt} [/INST]"
-    prompt += f" {bot_response}</s> "
-  prompt += f"[INST] {message} [/INST]"
-  return prompt
-agents =[
-    "WEB_DEV",
-    "AI_SYSTEM_PROMPT",
-    "PYTHON_CODE_DEV"
-]
-def generate(
-    prompt, history, agent_name=agents[0], sys_prompt="", temperature=0.9, max_new_tokens=2048, top_p=0.95, repetition_penalty=1.0, model="mistralai/Mixtral-8x7B-Instruct-v0.1"
-):
-    seed = random.randint(1,1111111111111111)
-
-    # Correct the line:
-    if agent_name == "WEB_DEV":
-        agent = "You are a helpful AI assistant. You are a web developer."
-    if agent_name == "AI_SYSTEM_PROMPT":
-        agent = "You are a helpful AI assistant. You are an AI system."
-    if agent_name == "PYTHON_CODE_DEV":
-        agent = "You are a helpful AI assistant. You are a Python code developer."
-    system_prompt = agent
-    temperature = float(temperature)
-    if temperature < 1e-2:
-        temperature = 1e-2
-    top_p = float(top_p)
-
-    # Add the system prompt to the beginning of the prompt
-    formatted_prompt = f"{system_prompt} {prompt}"
-
-    # Use 'prompt' here instead of 'message'
-    formatted_prompt = format_prompt(formatted_prompt, history, max_history_turns=5)  # Truncated history
-    logging.info(f"Formatted Prompt: {formatted_prompt}")
-    stream = client.text_generation(formatted_prompt, temperature=temperature, max_new_tokens=max_new_tokens, top_p=top_p, repetition_penalty=repetition_penalty, stream=True, details=True, return_full_text=False)
-    resp = ""
-    for response in stream:
-        resp += response.token.text
-
-    if VERBOSE:
-        logging.info(LOG_RESPONSE.format(resp))  # Log the response
-    return resp
-
 
 def generate_text_chunked(input_text, model, generation_parameters, max_tokens_to_generate):
     """Generates text in chunks to avoid token limit errors."""
@@ -349,234 +302,26 @@ def generate_text_chunked(input_text, model, generation_parameters, max_tokens_t
             generated_text.append(response[0]['generated_text'])
         else:
             # Handle cases where the sentence is too long
-            # You could split the sentence further or skip it
             print(f"Sentence too long: {sentence}") 
 
     return ''.join(generated_text)
 
-    formatted_prompt = format_prompt(prompt, history, max_history_turns=5)  # Truncated history
-    logging.info(f"Formatted Prompt: {formatted_prompt}")
-    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
-    output = ""
-
-    for response in stream:
-        output += response.token.text
-        yield output
-    return output
-
-
-additional_inputs=[
-    gr.Dropdown(
-        label="Agents",
-        choices=[s for s in agents],
-        value=agents[0],
-        interactive=True,
-        ),
-    gr.Textbox(
-        label="System Prompt",
-        max_lines=1,
-        interactive=True,
-    ),
-    gr.Slider(
-        label="Temperature",
-        value=0.9,
-        minimum=0.0,
-        maximum=1.0,
-        step=0.05,
-        interactive=True,
-        info="Higher values produce more diverse outputs",
-    ),
-
-    gr.Slider(
-        label="Max new tokens",
-        value=1048*10,
-        minimum=0,
-        maximum=1048*10,
-        step=64,
-        interactive=True,
-        info="The maximum numbers of new tokens",
-    ),
-    gr.Slider(
-        label="Top-p (nucleus sampling)",
-        value=0.90,
-        minimum=0.0,
-        maximum=1,
-        step=0.05,
-        interactive=True,
-        info="Higher values sample more low-probability tokens",
-    ),
-    gr.Slider(
-        label="Repetition penalty",
-        value=1.2,
-        minimum=1.0,
-        maximum=2.0,
-        step=0.05,
-        interactive=True,
-        info="Penalize repeated tokens",
-    ),
-
-
-]
-
-examples = [
-    ["Help me set up TypeScript configurations and integrate ts-loader in my existing React project.",
-"Update Webpack Configurations",
-"Install Dependencies",
-"Configure Ts-Loader",
-"TypeChecking Rules Setup",
-"React Specific Settings",
-"Compilation Options",
-"Test Runner Configuration"],
-
-["Guide me through building a serverless microservice using AWS Lambda and API Gateway, connecting to DynamoDB for storage.",
-"Set Up AWS Account",
-"Create Lambda Function",
-"APIGateway Integration",
-"Define DynamoDB Table Scheme",
-"Connect Service To DB",
-"Add Authentication Layers",
-"Monitor Metrics and Set Alarms"],
-
-["Migrate our current monolithic PHP application towards containerized services using Docker and Kubernetes for scalability.",
-"Architectural Restructuring Plan",
-"Containerisation Process With Docker",
-"Service Orchestration With Kubernetes",
-"Load Balancing Strategies",
-"Persistent Storage Solutions",
-"Network Policies Enforcement",
-"Continuous Integration / Continuous Delivery"],
-
-["Provide guidance on integrating WebAssembly modules compiled from C++ source files into an ongoing web project.",
-"Toolchain Selection (Emscripten vs. LLVM)",
-"Setting Up Compiler Environment",
-".cpp Source Preparation",
-"Module Building Approach",
-"Memory Management Considerations",
-"Performance Tradeoffs",
-"Seamless Web Assembly Embedding"]
-]
-
-def parse_action(line):
-    action_name, action_input = line.strip("action: ").split("=")
-    action_input = action_input.strip()
-    return action_name, action_input
-
-def get_file_tree(path):
-    """
-    Recursively explores a directory and returns a nested dictionary representing its file tree.
-    """
-    tree = {}
-    for item in os.listdir(path):
-        item_path = os.path.join(path, item)
-        if os.path.isdir(item_path):
-            tree[item] = get_file_tree(item_path)
-        else:
-            tree[item] = None
-    return tree
-
-def display_file_tree(tree, indent=0):
-    """
-    Prints a formatted representation of the file tree.
-    """
-    for name, subtree in tree.items():
-        print(f"{'  ' * indent}{name}")
-        if subtree is not None:
-            display_file_tree(subtree, indent + 1)
-
-def project_explorer(path):
-    """
-    Displays the file tree of a given path in a Streamlit app.
-    """
-    tree = get_file_tree(path)
-    tree_str = json.dumps(tree, indent=4)  # Convert the tree to a string for display
-    return tree_str
-
-def chat_app_logic(message, history, purpose, agent_name, sys_prompt, temperature, max_new_tokens, top_p, repetition_penalty, model):
-    # Your existing code here
-    
-    try:
-        # Pass 'message' as 'prompt'
-        response = ''.join(generate(
-            model=model,
-            prompt=message,  # Use 'prompt' here
-            history=history,
-            agent_name=agent_name,
-            sys_prompt=sys_prompt,
-            temperature=temperature,
-            max_new_tokens=max_new_tokens,
-            top_p=top_p,
-            repetition_penalty=repetition_penalty,
-        ))
-    except TypeError:
-        # ... (rest of the exception handling)
-
-        response_parts = []
-        for part in generate(
-            model=model,
-            prompt=message,  # Use 'prompt' here
-            history=history,
-            agent_name=agent_name,
-            sys_prompt=sys_prompt,
-            temperature=temperature,
-            max_new_tokens=max_new_tokens,
-            top_p=top_p,
-            repetition_penalty=repetition_penalty,
-        ):
-            if isinstance(part, str):
-                response_parts.append(part)
-            elif isinstance(part, dict) and 'content' in part:
-                response_parts.append(part['content'])
-
-        response = ''.join(response_parts)
-        history.append((message, response))
-        return history
-
-    history.append((message, response))
-    return history
-
-def main():
-    with gr.Blocks() as demo:
-        gr.Markdown("## FragMixt")
-        gr.Markdown("### Agents w/ Agents")
-
-        # Chat Interface
-        chatbot = gr.Chatbot(show_label=False, show_share_button=False, show_copy_button=True, likeable=True, layout="panel")
-        #chatbot.load(examples)
-
-        # Input Components
-        message = gr.Textbox(label="Enter your message", placeholder="Ask me anything!")
-        purpose = gr.Textbox(label="Purpose", placeholder="What is the purpose of this interaction?")
-        agent_name = gr.Dropdown(label="Agents", choices=[s for s in agents], value=agents[0], interactive=True)
-        sys_prompt = gr.Textbox(label="System Prompt", max_lines=1, interactive=True)
-        temperature = gr.Slider(label="Temperature", value=0.9, minimum=0.0, maximum=1.0, step=0.05, interactive=True, info="Higher values produce more diverse outputs")
-        max_new_tokens = gr.Slider(label="Max new tokens", value=1048*10, minimum=0, maximum=1048*10, step=64, interactive=True, info="The maximum numbers of new tokens")
-        top_p = gr.Slider(label="Top-p (nucleus sampling)", value=0.90, minimum=0.0, maximum=1, step=0.05, interactive=True, info="Higher values sample more low-probability tokens")
-        repetition_penalty = gr.Slider(label="Repetition penalty", value=1.2, minimum=1.0, maximum=2.0, step=0.05, interactive=True, info="Penalize repeated tokens")
-        model_input = gr.Textbox(label="Model", value="mistralai/Mixtral-8x7B-Instruct-v0.1", visible=False)
-
-        # Button to submit the message
-        submit_button = gr.Button(value="Send")
-
-        # Project Explorer Tab
-        with gr.Tab("Project Explorer"):
-            project_path = gr.Textbox(label="Project Path", placeholder="/home/user/app/current_project")
-            explore_button = gr.Button(value="Explore")
-            project_output = gr.Textbox(label="File Tree", lines=20)
-
-        # Chat App Logic Tab
-        with gr.Tab("Chat App"):
-            history = gr.State([])
-            for example in examples:
-                gr.Button(value=example[0]).click(lambda: chat_app_logic(example[0], history, purpose, agent_name, sys_prompt, temperature, max_new_tokens, top_p, repetition_penalty, model=model_input), outputs=chatbot)
-
-            # Connect components to the chat app logic
-            submit_button.click(chat_app_logic, inputs=[message, history, purpose, agent_name, sys_prompt, temperature, max_new_tokens, top_p, repetition_penalty, model_input], outputs=chatbot)
-            message.submit(chat_app_logic, inputs=[message, history, purpose, agent_name, sys_prompt, temperature, max_new_tokens, top_p, repetition_penalty, model_input], outputs=chatbot)
-
-        # Connect components to the project explorer
-        explore_button.click(project_explorer, inputs=project_path, outputs=project_output)
-
-    demo.launch(show_api=True)
+# Gradio Interface
+def gradio_interface(purpose, history):
+    history = json.loads(history) if history else []
+    result = run(purpose, history)
+    return next(result)
+
+iface = gr.Interface(
+    fn=gradio_interface,
+    inputs=[
+        gr.inputs.Textbox(lines=2, placeholder="Enter the purpose here..."),
+        gr.inputs.Textbox(lines=10, placeholder="Enter the history here (JSON format)...")
+    ],
+    outputs="text",
+    title="AI Assistant",
+    description="An AI assistant that helps with various tasks."
+)
 
 if __name__ == "__main__":
-    main()
+    iface.launch()