app.py

import os
import sys
import json
import argparse
import time
import io
import uuid
from PIL import Image
from typing import List, Dict, Any, Iterator

import gradio as gr
from gradio import ChatMessage

# Add the project root to the Python path
current_dir = os.path.dirname(os.path.abspath(__file__))
project_root = os.path.dirname(os.path.dirname(os.path.dirname(current_dir)))
sys.path.insert(0, project_root)

from octotools.models.initializer import Initializer
from octotools.models.planner import Planner
from octotools.models.memory import Memory
from octotools.models.executor import Executor
from octotools.models.utils import make_json_serializable

from utils import save_feedback


########### Test Huggingface Dataset ###########
from pathlib import Path
from huggingface_hub import CommitScheduler

# Add these near the top of the file with other constants
DATASET_DIR = Path("feedback_dataset")
DATASET_DIR.mkdir(parents=True, exist_ok=True)
DATASET_PATH = DATASET_DIR / f"feedback-{time.strftime('%Y%m%d_%H%M%S')}.json"

# Get Huggingface token from environment variable
HF_TOKEN = os.getenv("HUGGINGFACE_TOKEN")

scheduler = CommitScheduler(
    repo_id="lupantech/OctoTools-Gradio-Demo-User-Data",
    repo_type="dataset",
    folder_path=DATASET_DIR,
    path_in_repo="data",
    token=HF_TOKEN
)

def save_feedback(root_cache_dir: str, feedback_type: str, comment: str = None) -> None:
    """Save user feedback to Huggingface dataset"""
    with scheduler.lock:
        with DATASET_PATH.open("a") as f:
            feedback_data = {
                "query_id": os.path.basename(root_cache_dir),
                "feedback_type": feedback_type,
                "comment": comment,
                "datetime": time.strftime("%Y%m%d_%H%M%S")
            }
            json.dump(feedback_data, f)
            f.write("\n")
########### End of Test Huggingface Dataset ###########

class Solver:
    def __init__(
        self,
        planner,
        memory,
        executor,
        task: str,
        task_description: str,
        output_types: str = "base,final,direct",
        index: int = 0,
        verbose: bool = True,
        max_steps: int = 10,
        max_time: int = 60,
        root_cache_dir: str = "cache"
    ):
        self.planner = planner
        self.memory = memory
        self.executor = executor
        self.task = task
        self.task_description = task_description
        self.index = index
        self.verbose = verbose
        self.max_steps = max_steps
        self.max_time = max_time
        self.root_cache_dir = root_cache_dir

        self.output_types = output_types.lower().split(',')
        assert all(output_type in ["base", "final", "direct"] for output_type in self.output_types), "Invalid output type. Supported types are 'base', 'final', 'direct'."


    def stream_solve_user_problem(self, user_query: str, user_image: Image.Image, api_key: str, messages: List[ChatMessage]) -> Iterator[List[ChatMessage]]:
        """
        Streams intermediate thoughts and final responses for the problem-solving process based on user input.
        
        Args:
            user_query (str): The text query input from the user.
            user_image (Image.Image): The uploaded image from the user (PIL Image object).
            messages (list): A list of ChatMessage objects to store the streamed responses.
        """

        if user_image:
            # # Convert PIL Image to bytes (for processing)
            # img_bytes_io = io.BytesIO()
            # user_image.save(img_bytes_io, format="PNG")  # Convert image to PNG bytes
            # img_bytes = img_bytes_io.getvalue()  # Get bytes
            
            # Use image paths instead of bytes,
            # os.makedirs(os.path.join(self.root_cache_dir, 'images'), exist_ok=True)
            # img_path = os.path.join(self.root_cache_dir, 'images', str(uuid.uuid4()) + '.jpg')

            img_path = os.path.join(self.root_cache_dir, 'query_image.jpg')
            user_image.save(img_path)
        else:
            img_path = None

        # Set tool cache directory
        _cache_dir = os.path.join(self.root_cache_dir, "tool_cache") # NOTE: This is the directory for tool cache
        self.executor.set_query_cache_dir(_cache_dir)
        
        # Step 1: Display the received inputs
        if user_image:
            messages.append(ChatMessage(role="assistant", content=f"📝 Received Query: {user_query}\n🖼️ Image Uploaded"))
        else:
            messages.append(ChatMessage(role="assistant", content=f"📝 Received Query: {user_query}"))
        yield messages

        # # Step 2: Add "thinking" status while processing
        # messages.append(ChatMessage(
        #     role="assistant",
        #     content="",
        #     metadata={"title": "⏳ Thinking: Processing input..."}
        # ))

        # Step 3: Initialize problem-solving state
        start_time = time.time()
        step_count = 0
        json_data = {"query": user_query, "image": "Image received as bytes"}

        # Step 4: Query Analysis
        query_analysis = self.planner.analyze_query(user_query, img_path)
        json_data["query_analysis"] = query_analysis
        messages.append(ChatMessage(role="assistant", 
                                    content=f"{query_analysis}", 
                                    metadata={"title": "🔍 Query Analysis"}))
        yield messages

        # Step 5: Execution loop (similar to your step-by-step solver)
        while step_count < self.max_steps and (time.time() - start_time) < self.max_time:
            step_count += 1
            # messages.append(ChatMessage(role="assistant", 
            #                             content=f"Generating next step...",
            #                             metadata={"title": f"🔄 Step {step_count}"}))
            yield messages

            # Generate the next step
            next_step = self.planner.generate_next_step(
                user_query, img_path, query_analysis, self.memory, step_count, self.max_steps
            )
            context, sub_goal, tool_name = self.planner.extract_context_subgoal_and_tool(next_step)

            # Display the step information
            messages.append(ChatMessage(
                role="assistant",
                content=f"- Context: {context}\n- Sub-goal: {sub_goal}\n- Tool: {tool_name}",
                metadata={"title": f"📌 Step {step_count}: {tool_name}"}
            ))
            yield messages

            # Handle tool execution or errors
            if tool_name not in self.planner.available_tools:
                messages.append(ChatMessage(
                    role="assistant", 
                    content=f"⚠️ Error: Tool '{tool_name}' is not available."))
                yield messages
                continue

            # Execute the tool command
            tool_command = self.executor.generate_tool_command(
                user_query, img_path, context, sub_goal, tool_name, self.planner.toolbox_metadata[tool_name]
            )
            explanation, command = self.executor.extract_explanation_and_command(tool_command)
            result = self.executor.execute_tool_command(tool_name, command)
            result = make_json_serializable(result)

            messages.append(ChatMessage(
                role="assistant", 
                content=f"{json.dumps(result, indent=4)}",
                metadata={"title": f"✅ Step {step_count} Result: {tool_name}"}))
            yield messages

            # Step 6: Memory update and stopping condition
            self.memory.add_action(step_count, tool_name, sub_goal, tool_command, result)
            stop_verification = self.planner.verificate_memory(user_query, img_path, query_analysis, self.memory)
            conclusion = self.planner.extract_conclusion(stop_verification)

            messages.append(ChatMessage(
                role="assistant", 
                content=f"🛑 Step {step_count} Conclusion: {conclusion}"))
            yield messages

            if conclusion == 'STOP':
                break

        # Step 7: Generate Final Output (if needed)
        if 'final' in self.output_types:
            final_output = self.planner.generate_final_output(user_query, img_path, self.memory)
            messages.append(ChatMessage(role="assistant", content=f"🎯 Final Output:\n{final_output}"))
            yield messages

        if 'direct' in self.output_types:
            direct_output = self.planner.generate_direct_output(user_query, img_path, self.memory)
            messages.append(ChatMessage(role="assistant", content=f"🔹 Direct Output:\n{direct_output}"))
            yield messages

        # Step 8: Completion Message
        messages.append(ChatMessage(role="assistant", content="✅ Problem-solving process completed."))
        yield messages
            

def parse_arguments():
    parser = argparse.ArgumentParser(description="Run the OctoTools demo with specified parameters.")
    parser.add_argument("--llm_engine_name", default="gpt-4o", help="LLM engine name.")
    parser.add_argument("--max_tokens", type=int, default=2000, help="Maximum tokens for LLM generation.")
    parser.add_argument("--task", default="minitoolbench", help="Task to run.")
    parser.add_argument("--task_description", default="", help="Task description.")
    parser.add_argument(
        "--output_types",
        default="base,final,direct",
        help="Comma-separated list of required outputs (base,final,direct)"
    )
    parser.add_argument("--enabled_tools", default="Generalist_Solution_Generator_Tool", help="List of enabled tools.")
    parser.add_argument("--root_cache_dir", default="demo_solver_cache", help="Path to solver cache directory.")
    parser.add_argument("--verbose", type=bool, default=True, help="Enable verbose output.")

    # NOTE: Add new arguments
    parser.add_argument("--run_baseline_only", type=bool, default=False, help="Run only the baseline (no toolbox).")
    parser.add_argument("--openai_api_source", default="we_provided", choices=["we_provided", "user_provided"], help="Source of OpenAI API key.")
    return parser.parse_args()


def solve_problem_gradio(user_query, user_image, max_steps=10, max_time=60, api_key=None, llm_model_engine=None, enabled_tools=None):
    """
    Wrapper function to connect the solver to Gradio.
    Streams responses from `solver.stream_solve_user_problem` for real-time UI updates.
    """

    # Generate shorter ID (Date and first 8 characters of UUID)
    query_id = time.strftime("%Y%m%d_%H%M%S") + "_" + str(uuid.uuid4())[:8] # e.g, 20250217_062225_612f2474
    print(f"Query ID: {query_id}")

    # Create a directory for the query ID
    query_dir = os.path.join(args.root_cache_dir, query_id)
    os.makedirs(query_dir, exist_ok=True)
    args.root_cache_dir = query_dir

    if api_key is None:
        return [["assistant", "⚠️ Error: OpenAI API Key is required."]]
    
    # # Initialize Tools
    # enabled_tools = args.enabled_tools.split(",") if args.enabled_tools else []

    # # Hack enabled_tools
    # enabled_tools = ["Generalist_Solution_Generator_Tool"]

    # Instantiate Initializer
    initializer = Initializer(
        enabled_tools=enabled_tools,
        model_string=llm_model_engine,
        api_key=api_key
    )

    # Instantiate Planner
    planner = Planner(
        llm_engine_name=llm_model_engine,
        toolbox_metadata=initializer.toolbox_metadata,
        available_tools=initializer.available_tools,
        api_key=api_key
    )

    # Instantiate Memory
    memory = Memory()

    # Instantiate Executor
    executor = Executor(
        llm_engine_name=llm_model_engine,
        root_cache_dir=args.root_cache_dir, # NOTE
        enable_signal=False,
        api_key=api_key
    )

    # Instantiate Solver
    solver = Solver(
        planner=planner,
        memory=memory,
        executor=executor,
        task=args.task,
        task_description=args.task_description,
        output_types=args.output_types,  # Add new parameter
        verbose=args.verbose,
        max_steps=max_steps,
        max_time=max_time,
        root_cache_dir=args.root_cache_dir # NOTE
    )

    if solver is None:
        return [["assistant", "⚠️ Error: Solver is not initialized. Please restart the application."]]

    messages = []  # Initialize message list
    for message_batch in solver.stream_solve_user_problem(user_query, user_image, api_key, messages):
        yield [msg for msg in message_batch]  # Ensure correct format for Gradio Chatbot


def main(args):
    #################### Gradio Interface ####################
    with gr.Blocks() as demo:
    # with gr.Blocks(theme=gr.themes.Soft()) as demo:
        # Theming https://www.gradio.app/guides/theming-guide

        gr.Markdown("# 🐙 Chat with OctoTools: An Agentic Framework for Complex Reasoning")  # Title
        # gr.Markdown("[![OctoTools](https://img.shields.io/badge/OctoTools-Agentic%20Framework%20for%20Complex%20Reasoning-blue)](https://octotools.github.io/)")  # Title
        gr.Markdown("""
        **OctoTools** is a training-free, user-friendly, and easily extensible open-source agentic framework designed to tackle complex reasoning across diverse domains. 
        It introduces standardized **tool cards** to encapsulate tool functionality, a **planner** for both high-level and low-level planning, and an **executor** to carry out tool usage. 
                    
        [Website](https://octotools.github.io/) | 
        [Github](https://github.com/octotools/octotools) | 
        [arXiv](https://arxiv.org/abs/2502.xxxxx) | 
        [Paper](https://arxiv.org/pdf/2502.xxxxx) | 
        [Tool Cards](https://octotools.github.io/#tool-cards) | 
        [Example Visualizations](https://octotools.github.io/#visualization) | 
        [Discord](https://discord.gg/NMJx66DC)
        """)

        with gr.Row():
            # Left column for settings
            with gr.Column(scale=1):
                with gr.Row():
                    if args.openai_api_source == "user_provided":
                        print("Using API key from user input.")
                        api_key = gr.Textbox(
                            show_label=True,
                            placeholder="Your API key will not be stored in any way.",
                            type="password", 
                            label="OpenAI API Key",
                            # container=False
                        )
                    else:
                        print(f"Using local API key from environment variable: {os.getenv('OPENAI_API_KEY')[:4]}...")
                        api_key = gr.Textbox(
                            value=os.getenv("OPENAI_API_KEY"),
                            visible=False,
                            interactive=False
                        )

                with gr.Row():
                    llm_model_engine = gr.Dropdown(
                        choices=["gpt-4o", "gpt-4o-2024-11-20", "gpt-4o-2024-08-06", "gpt-4o-2024-05-13",
                                "gpt-4o-mini", "gpt-4o-mini-2024-07-18"], 
                        value="gpt-4o", 
                        label="LLM Model"
                    )
                with gr.Row():
                    max_steps = gr.Slider(value=5, minimum=1, maximum=10, step=1, label="Max Steps")
                
                with gr.Row():
                    max_time = gr.Slider(value=180, minimum=60, maximum=300, step=30, label="Max Time (seconds)")

                with gr.Row():
                    # Container for tools section
                    with gr.Column():

                        # First row for checkbox group
                        enabled_tools = gr.CheckboxGroup(
                            choices=all_tools,
                            value=all_tools,
                            label="Selected Tools",
                        )

                        # Second row for buttons
                        with gr.Row():
                            enable_all_btn = gr.Button("Select All Tools")
                            disable_all_btn = gr.Button("Clear All Tools")
                        
                        # Add click handlers for the buttons
                        enable_all_btn.click(
                            lambda: all_tools,
                            outputs=enabled_tools
                        )
                        disable_all_btn.click(
                            lambda: [],
                            outputs=enabled_tools
                        )

            with gr.Column(scale=5):
                
                with gr.Row():
                    # Middle column for the query
                    with gr.Column(scale=2):
                        user_image = gr.Image(type="pil", label="Upload an Image (Optional)", height=500)  # Accepts multiple formats
                        
                        with gr.Row():
                            user_query = gr.Textbox( placeholder="Type your question here...", label="Question (Required)")

                        with gr.Row():
                            run_button = gr.Button("🐙 Submit and Run", variant="primary")  # Run button with blue color

                    # Right column for the output
                    with gr.Column(scale=3):
                        chatbot_output = gr.Chatbot(type="messages", label="Step-wise Problem-Solving Output (Deep Thinking)", height=500)

                        # TODO: Add actions to the buttons
                        with gr.Row(elem_id="buttons") as button_row:
                            upvote_btn = gr.Button(value="👍  Upvote", interactive=True, variant="primary") # TODO
                            downvote_btn = gr.Button(value="👎  Downvote", interactive=True, variant="primary") # TODO
                            # stop_btn = gr.Button(value="⛔️  Stop", interactive=True) # TODO
                            # clear_btn = gr.Button(value="🗑️  Clear history", interactive=True) # TODO

                        # TODO: Add comment textbox
                        with gr.Row():
                            comment_textbox = gr.Textbox(value="", 
                                                        placeholder="Feel free to add any comments here. Thanks for using OctoTools!",
                                                        label="💬 Comment (Type and press Enter to submit.)", interactive=True) # TODO
                            
                        # Update the button click handlers
                        upvote_btn.click(
                            fn=lambda: save_feedback(args.root_cache_dir, "upvote"),
                            inputs=[],
                            outputs=[]
                        )
                        
                        downvote_btn.click(
                            fn=lambda: save_feedback(args.root_cache_dir, "downvote"),
                            inputs=[],
                            outputs=[]
                        )

                        # Add handler for comment submission
                        comment_textbox.submit(
                            fn=lambda comment: save_feedback(args.root_cache_dir, comment),
                            inputs=[comment_textbox],
                            outputs=[]
                        )

                # Bottom row for examples
                with gr.Row():
                    with gr.Column(scale=5):
                        gr.Markdown("")
                        gr.Markdown("""
                                    ## 💡 Try these examples with suggested tools.
                                    """)
                        gr.Examples(
                            examples=[
                                [ None, "Who is the president of the United States?", ["Google_Search_Tool"]],

                                [ "examples/baseball.png", "How many baseballs are there?", ["Object_Detector_Tool"]],

                                [ None, "Using the numbers [1, 1, 6, 9], create an expression that equals 24. You must use basic arithmetic operations (+, -, ×, /) and parentheses. For example, one solution for [1, 2, 3, 4] is (1+2+3)×4.", ["Python_Code_Generator_Tool"]],

                                [None, "What are the research trends in tool agents with large language models for scientific discovery? Please consider the latest literature from ArXiv, PubMed, Nature, and news sources.", ["ArXiv_Paper_Searcher_Tool", "Pubmed_Search_Tool", "Nature_News_Fetcher_Tool"]],

                                [ "examples/rotting_kiwi.png", "You are given a 3 x 3 grid in which each cell can contain either no kiwi, one fresh kiwi, or one rotten kiwi. Every minute, any fresh kiwi that is 4-directionally adjacent to a rotten kiwi also becomes rotten. What is the minimum number of minutes that must elapse until no cell has a fresh kiwi?", ["Image_Captioner_Tool"]]

                            ],
                            inputs=[user_image, user_query, enabled_tools],
                            # label="Try these examples with suggested tools."
                        )

        # Link button click to function
        run_button.click(
            fn=solve_problem_gradio, 
            inputs=[user_query, user_image, max_steps, max_time, api_key, llm_model_engine, enabled_tools], 
            outputs=chatbot_output
        )
    #################### Gradio Interface ####################

    # Launch the Gradio app
    demo.launch()


if __name__ == "__main__":
    args = parse_arguments()

    # Manually set enabled tools
    # args.enabled_tools = "Generalist_Solution_Generator_Tool"

    # All tools
    all_tools = [
        "Generalist_Solution_Generator_Tool",

        "Image_Captioner_Tool",
        "Object_Detector_Tool",
        "Relevant_Patch_Zoomer_Tool",
        "Text_Detector_Tool",

        "Python_Code_Generator_Tool",

        "ArXiv_Paper_Searcher_Tool",
        "Google_Search_Tool",
        "Nature_News_Fetcher_Tool",
        "Pubmed_Search_Tool",
        "URL_Text_Extractor_Tool",
        "Wikipedia_Knowledge_Searcher_Tool"
    ]
    args.enabled_tools = ",".join(all_tools)

    main(args)