submit

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Luong Huu Thanh
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -1,12 +1,104 @@
 ---
-title: Agent Fc
-emoji: 🐢
-colorFrom: yellow
-colorTo: pink
+title: Template Final Assignment
+emoji: 🕵🏻‍♂️
+colorFrom: indigo
+colorTo: indigo
 sdk: gradio
-sdk_version: 5.35.0
+sdk_version: 5.25.2
 app_file: app.py
 pinned: false
+hf_oauth: true
+# optional, default duration is 8 hours/480 minutes. Max duration is 30 days/43200 minutes.
+hf_oauth_expiration_minutes: 480
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# **GAIA Agent**
+
+## **Introduction**
+
+**GAIA Agent** is an automated system built to tackle and submit solutions for the GAIA benchmark, which tests the capabilities of general-purpose AI agents on diverse and challenging tasks. These tasks require a combination of reasoning, code execution, information retrieval, data interpretation, and multimodal understanding. Powered by advanced language models (such as HuggingFace, and Groq), the agent incorporates a versatile set of tools including browser tools, code interpreter tools, mathematical tools, document processing tools, image processing and generation tools. It is designed for seamless interaction with the benchmark, offering automatic evaluation, submission, and result display through a user-friendly Gradio interface.
+
+## **Tools Implementation**
+
+### **Browser tools** 
+- **Wikipedia Search:** Search Wikipedia for a query and return maximum 2 results.
+- **Web Search:** Search the web for a query and return maximum 2 results.
+- **Arxiv Search:** Search arXiv for a query and return maximum 2 results.
+
+### **Code interpreter tools**
+- **Execute Multi-programming Language:** Execute code in multiple languages (Python, Bash, SQL, C, Java) and return results.
+
+### **Mathematical tools**
+- **Multiplication Tools:** Multiplies 2 numbers 
+- **Addition:** Adds 2 numbers
+- **Subtraction:** Subtracts 2 numbers 
+- **Division:** Divides 2 numbers 
+- **Modulus:** Get the modulus of 2 numbers
+- **Power:** Get the power of 2 numbers 
+- **Square root:** Get the square root of a number
+
+### **Document processing tools**
+- **Save and Read File:** Save content to a file and return the path 
+- **Download a File from URL:** Download a file from a URL and save it to a temporary location
+- **Extract Text from Image:** Extract text from an image using OCR library pytesseract (if available)
+- **Analyze CSV File:** Analyze a CSV file using pandas and answer a question about it 
+- **Analyze Excel File:** Analyze an Excel file using pandas and answer a question about it
+
+### **Image processing and generation tools**
+- **Analyze Image:** Analyze basic properties of an image (size, mode, color analysis, thumbnail preview)
+- **Transform Image:** Apply transformations: resize, rotate, crop, flip, brightness, contrast, blur, sharpen, grayscale
+- **Draw on Image:** Draw shapes (rectangle, circle, line) or text onto an image
+- **Generate Simple Image:** Generate a simple image (gradient, noise, pattern, chart)
+- **Combine Images:** Combine multiple images (collage, stack, blend)
+
+
+## **Installation**
+Clone the repository, change the current working directory to this repository's root folder:
+
+```
+git clone https://github.com/fisherman611/gaia-agent.git
+```
+```
+cd gaia-agent
+```
+
+Install ```requirements.txt``` (replace `3.11` with your installed Python version):
+
+```
+py -3.11 -m pip install -r requirements.txt
+```
+
+## **Environment Variables**
+Store some API keys an variables in the `.env` file and load it in your code using `load_dotenv`
+
+```
+SUPABASE_URL=...
+SUPABASE_SERVICE_ROLE_KEY=...
+SUPABASE_SERVICE_KEY=...
+HUGGINGFACEHUB_API_TOKEN=...
+GROQ_API_KEY=...
+TAVILY_API_KEY=...
+LANGSMITH_API_KEY=...
+
+LANGSMITH_TRACING=true
+LANGSMITH_PROJECT=ai_agent_course
+LANGSMITH_ENDPOINT=https://api.smith.langchain.com
+```
+
+## **Demo**
+To run the application using the command line, use the following command (replace `3.11` with your installed Python version):
+```
+py -3.11 app.py
+```
+Or run in the [Hugging Face Space](https://huggingface.co/spaces/fisherman611/gaia-agent)
+## **Resources**
+- [GAIA Benchmark](https://huggingface.co/spaces/gaia-benchmark/leaderboard)
+- [Hugging Face Agents Course](https://huggingface.co/agents-course)
+- [Langgraph Agents](https://langchain-ai.github.io/langgraph/)
+
+
+## **Contributing**
+Contributions are welcome! If you find any issues or have suggestions for improvements, please open an issue or submit a pull request.
+
+## **License**
+This project is licensed under the [MIT License](https://mit-license.org/).

agent.py

@@ -1,15 +1,36 @@
-"""LangGraph Agent"""
 import os
 from dotenv import load_dotenv
+from typing import List, Dict, Any, Optional
+import tempfile
+import re
+import json
+import requests
+from urllib.parse import urlparse
+import pytesseract
+from PIL import Image, ImageDraw, ImageFont, ImageEnhance, ImageFilter
+import cmath
+import pandas as pd
+import uuid
+import numpy as np
+from code_interpreter import CodeInterpreter
+
+interpreter_instance = CodeInterpreter()
+
+from image_processing import *
+
+"""Langraph"""
 from langgraph.graph import START, StateGraph, MessagesState
-from langgraph.prebuilt import tools_condition
-from langgraph.prebuilt import ToolNode
-from langchain_google_genai import ChatGoogleGenerativeAI
-from langchain_groq import ChatGroq
-from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint, HuggingFaceEmbeddings
 from langchain_community.tools.tavily_search import TavilySearchResults
 from langchain_community.document_loaders import WikipediaLoader
 from langchain_community.document_loaders import ArxivLoader
+from langgraph.prebuilt import ToolNode, tools_condition
+from langchain_google_genai import ChatGoogleGenerativeAI
+from langchain_groq import ChatGroq
+from langchain_huggingface import (
+    ChatHuggingFace,
+    HuggingFaceEndpoint,
+    HuggingFaceEmbeddings,
+)
 from langchain_community.vectorstores import SupabaseVectorStore
 from langchain_core.messages import SystemMessage, HumanMessage
 from langchain_core.tools import tool
@@ -18,119 +39,666 @@ from supabase.client import Client, create_client
 
 load_dotenv()
 
+### =============== BROWSER TOOLS =============== ###
+
+
+@tool
+def wiki_search(query: str) -> str:
+    """Search Wikipedia for a query and return maximum 2 results.
+
+    Args:
+        query: The search query."""
+    search_docs = WikipediaLoader(query=query, load_max_docs=2).load()
+    formatted_search_docs = "\n\n---\n\n".join(
+        [
+            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content}\n</Document>'
+            for doc in search_docs
+        ]
+    )
+    return {"wiki_results": formatted_search_docs}
+
+
+@tool
+def web_search(query: str) -> str:
+    """Search Tavily for a query and return maximum 3 results.
+
+    Args:
+        query: The search query."""
+    search_docs = TavilySearchResults(max_results=3).invoke(query=query)
+    formatted_search_docs = "\n\n---\n\n".join(
+        [
+            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content}\n</Document>'
+            for doc in search_docs
+        ]
+    )
+    return {"web_results": formatted_search_docs}
+
+
 @tool
-def multiply(a: int, b: int) -> int:
-    """Multiply two numbers.
+def arxiv_search(query: str) -> str:
+    """Search Arxiv for a query and return maximum 3 result.
 
     Args:
-        a: first int
-        b: second int
+        query: The search query."""
+    search_docs = ArxivLoader(query=query, load_max_docs=3).load()
+    formatted_search_docs = "\n\n---\n\n".join(
+        [
+            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content[:1000]}\n</Document>'
+            for doc in search_docs
+        ]
+    )
+    return {"arxiv_results": formatted_search_docs}
+
+
+### =============== CODE INTERPRETER TOOLS =============== ###
+
+
+@tool
+def execute_code_multilang(code: str, language: str = "python") -> str:
+    """Execute code in multiple languages (Python, Bash, SQL, C, Java) and return results.
+
+    Args:
+        code (str): The source code to execute.
+        language (str): The language of the code. Supported: "python", "bash", "sql", "c", "java".
+
+    Returns:
+        A string summarizing the execution results (stdout, stderr, errors, plots, dataframes if any).
+    """
+    supported_languages = ["python", "bash", "sql", "c", "java"]
+    language = language.lower()
+
+    if language not in supported_languages:
+        return f"❌ Unsupported language: {language}. Supported languages are: {', '.join(supported_languages)}"
+
+    result = interpreter_instance.execute_code(code, language=language)
+
+    response = []
+
+    if result["status"] == "success":
+        response.append(f"✅ Code executed successfully in **{language.upper()}**")
+
+        if result.get("stdout"):
+            response.append(
+                "\n**Standard Output:**\n```\n" + result["stdout"].strip() + "\n```"
+            )
+
+        if result.get("stderr"):
+            response.append(
+                "\n**Standard Error (if any):**\n```\n"
+                + result["stderr"].strip()
+                + "\n```"
+            )
+
+        if result.get("result") is not None:
+            response.append(
+                "\n**Execution Result:**\n```\n"
+                + str(result["result"]).strip()
+                + "\n```"
+            )
+
+        if result.get("dataframes"):
+            for df_info in result["dataframes"]:
+                response.append(
+                    f"\n**DataFrame `{df_info['name']}` (Shape: {df_info['shape']})**"
+                )
+                df_preview = pd.DataFrame(df_info["head"])
+                response.append("First 5 rows:\n```\n" + str(df_preview) + "\n```")
+
+        if result.get("plots"):
+            response.append(
+                f"\n**Generated {len(result['plots'])} plot(s)** (Image data returned separately)"
+            )
+
+    else:
+        response.append(f"❌ Code execution failed in **{language.upper()}**")
+        if result.get("stderr"):
+            response.append(
+                "\n**Error Log:**\n```\n" + result["stderr"].strip() + "\n```"
+            )
+
+    return "\n".join(response)
+
+
+### =============== MATHEMATICAL TOOLS =============== ###
+
+
+@tool
+def multiply(a: float, b: float) -> float:
+    """
+    Multiplies two numbers.
+
+    Args:
+        a (float): the first number
+        b (float): the second number
     """
     return a * b
 
+
 @tool
-def add(a: int, b: int) -> int:
-    """Add two numbers.
-    
+def add(a: float, b: float) -> float:
+    """
+    Adds two numbers.
+
     Args:
-        a: first int
-        b: second int
+        a (float): the first number
+        b (float): the second number
     """
     return a + b
 
+
 @tool
-def subtract(a: int, b: int) -> int:
-    """Subtract two numbers.
-    
+def subtract(a: float, b: float) -> int:
+    """
+    Subtracts two numbers.
+
     Args:
-        a: first int
-        b: second int
+        a (float): the first number
+        b (float): the second number
     """
     return a - b
 
+
 @tool
-def divide(a: int, b: int) -> int:
-    """Divide two numbers.
-    
+def divide(a: float, b: float) -> float:
+    """
+    Divides two numbers.
+
     Args:
-        a: first int
-        b: second int
+        a (float): the first float number
+        b (float): the second float number
     """
     if b == 0:
-        raise ValueError("Cannot divide by zero.")
+        raise ValueError("Cannot divided by zero.")
     return a / b
 
+
 @tool
 def modulus(a: int, b: int) -> int:
-    """Get the modulus of two numbers.
-    
+    """
+    Get the modulus of two numbers.
+
     Args:
-        a: first int
-        b: second int
+        a (int): the first number
+        b (int): the second number
     """
     return a % b
 
+
 @tool
-def wiki_search(query: str) -> str:
-    """Search Wikipedia for a query and return maximum 2 results.
-    
+def power(a: float, b: float) -> float:
+    """
+    Get the power of two numbers.
+
     Args:
-        query: The search query."""
-    search_docs = WikipediaLoader(query=query, load_max_docs=2).load()
-    formatted_search_docs = "\n\n---\n\n".join(
-        [
-            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content}\n</Document>'
-            for doc in search_docs
-        ])
-    return {"wiki_results": formatted_search_docs}
+        a (float): the first number
+        b (float): the second number
+    """
+    return a**b
+
 
 @tool
-def web_search(query: str) -> str:
-    """Search Tavily for a query and return maximum 3 results.
-    
+def square_root(a: float) -> float | complex:
+    """
+    Get the square root of a number.
+
     Args:
-        query: The search query."""
-    search_docs = TavilySearchResults(max_results=3).invoke(query=query)
-    formatted_search_docs = "\n\n---\n\n".join(
-        [
-            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content}\n</Document>'
-            for doc in search_docs
-        ])
-    return {"web_results": formatted_search_docs}
+        a (float): the number to get the square root of
+    """
+    if a >= 0:
+        return a**0.5
+    return cmath.sqrt(a)
+
+
+### =============== DOCUMENT PROCESSING TOOLS =============== ###
+
 
 @tool
-def arvix_search(query: str) -> str:
-    """Search Arxiv for a query and return maximum 3 result.
-    
+def save_and_read_file(content: str, filename: Optional[str] = None) -> str:
+    """
+    Save content to a file and return the path.
+
     Args:
-        query: The search query."""
-    search_docs = ArxivLoader(query=query, load_max_docs=3).load()
-    formatted_search_docs = "\n\n---\n\n".join(
-        [
-            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content[:1000]}\n</Document>'
-            for doc in search_docs
-        ])
-    return {"arvix_results": formatted_search_docs}
+        content (str): the content to save to the file
+        filename (str, optional): the name of the file. If not provided, a random name file will be created.
+    """
+    temp_dir = tempfile.gettempdir()
+    if filename is None:
+        temp_file = tempfile.NamedTemporaryFile(delete=False, dir=temp_dir)
+        filepath = temp_file.name
+    else:
+        filepath = os.path.join(temp_dir, filename)
+
+    with open(filepath, "w") as f:
+        f.write(content)
+
+    return f"File saved to {filepath}. You can read this file to process its contents."
+
+
+@tool
+def download_file_from_url(url: str, filename: Optional[str] = None) -> str:
+    """
+    Download a file from a URL and save it to a temporary location.
+
+    Args:
+        url (str): the URL of the file to download.
+        filename (str, optional): the name of the file. If not provided, a random name file will be created.
+    """
+    try:
+        # Parse URL to get filename if not provided
+        if not filename:
+            path = urlparse(url).path
+            filename = os.path.basename(path)
+            if not filename:
+                filename = f"downloaded_{uuid.uuid4().hex[:8]}"
+
+        # Create temporary file
+        temp_dir = tempfile.gettempdir()
+        filepath = os.path.join(temp_dir, filename)
+
+        # Download the file
+        response = requests.get(url, stream=True)
+        response.raise_for_status()
+
+        # Save the file
+        with open(filepath, "wb") as f:
+            for chunk in response.iter_content(chunk_size=8192):
+                f.write(chunk)
+
+        return f"File downloaded to {filepath}. You can read this file to process its contents."
+    except Exception as e:
+        return f"Error downloading file: {str(e)}"
+
+
+@tool
+def extract_text_from_image(image_path: str) -> str:
+    """
+    Extract text from an image using OCR library pytesseract (if available).
+
+    Args:
+        image_path (str): the path to the image file.
+    """
+    try:
+        # Open the image
+        image = Image.open(image_path)
+
+        # Extract text from the image
+        text = pytesseract.image_to_string(image)
+
+        return f"Extracted text from image:\n\n{text}"
+    except Exception as e:
+        return f"Error extracting text from image: {str(e)}"
+
+
+@tool
+def analyze_csv_file(file_path: str, query: str) -> str:
+    """
+    Analyze a CSV file using pandas and answer a question about it.
+
+    Args:
+        file_path (str): the path to the CSV file.
+        query (str): Question about the data
+    """
+    try:
+        # Read the CSV file
+        df = pd.read_csv(file_path)
+
+        # Run various analyses based on the query
+        result = f"CSV file loaded with {len(df)} rows and {len(df.columns)} columns.\n"
+        result += f"Columns: {', '.join(df.columns)}\n\n"
+
+        # Add summary statistics
+        result += "Summary statistics:\n"
+        result += str(df.describe())
+
+        return result
+
+    except Exception as e:
+        return f"Error analyzing CSV file: {str(e)}"
+
+
+@tool
+def analyze_excel_file(file_path: str, query: str) -> str:
+    """
+    Analyze an Excel file using pandas and answer a question about it.
+
+    Args:
+        file_path (str): the path to the Excel file.
+        query (str): Question about the data
+    """
+    try:
+        # Read the Excel file
+        df = pd.read_excel(file_path)
+
+        # Run various analyses based on the query
+        result = (
+            f"Excel file loaded with {len(df)} rows and {len(df.columns)} columns.\n"
+        )
+        result += f"Columns: {', '.join(df.columns)}\n\n"
+
+        # Add summary statistics
+        result += "Summary statistics:\n"
+        result += str(df.describe())
+
+        return result
+
+    except Exception as e:
+        return f"Error analyzing Excel file: {str(e)}"
+
+
+### ============== IMAGE PROCESSING AND GENERATION TOOLS =============== ###
+
+
+@tool
+def analyze_image(image_base64: str) -> Dict[str, Any]:
+    """
+    Analyze basic properties of an image (size, mode, color analysis, thumbnail preview).
+
+    Args:
+        image_base64 (str): Base64 encoded image string
+
+    Returns:
+        Dictionary with analysis result
+    """
+    try:
+        img = decode_image(image_base64)
+        width, height = img.size
+        mode = img.mode
 
+        if mode in ("RGB", "RGBA"):
+            arr = np.array(img)
+            avg_colors = arr.mean(axis=(0, 1))
+            dominant = ["Red", "Green", "Blue"][np.argmax(avg_colors[:3])]
+            brightness = avg_colors.mean()
+            color_analysis = {
+                "average_rgb": avg_colors.tolist(),
+                "brightness": brightness,
+                "dominant_color": dominant,
+            }
+        else:
+            color_analysis = {"note": f"No color analysis for mode {mode}"}
+
+        thumbnail = img.copy()
+        thumbnail.thumbnail((100, 100))
+        thumb_path = save_image(thumbnail, "thumbnails")
+        thumbnail_base64 = encode_image(thumb_path)
+
+        return {
+            "dimensions": (width, height),
+            "mode": mode,
+            "color_analysis": color_analysis,
+            "thumbnail": thumbnail_base64,
+        }
+    except Exception as e:
+        return {"error": str(e)}
+
+
+@tool
+def transform_image(
+    image_base64: str, operation: str, params: Optional[Dict[str, Any]] = None
+) -> Dict[str, Any]:
+    """
+    Apply transformations: resize, rotate, crop, flip, brightness, contrast, blur, sharpen, grayscale.
+
+    Args:
+        image_base64 (str): Base64 encoded input image
+        operation (str): Transformation operation
+        params (Dict[str, Any], optional): Parameters for the operation
+
+    Returns:
+        Dictionary with transformed image (base64)
+    """
+    try:
+        img = decode_image(image_base64)
+        params = params or {}
+
+        if operation == "resize":
+            img = img.resize(
+                (
+                    params.get("width", img.width // 2),
+                    params.get("height", img.height // 2),
+                )
+            )
+        elif operation == "rotate":
+            img = img.rotate(params.get("angle", 90), expand=True)
+        elif operation == "crop":
+            img = img.crop(
+                (
+                    params.get("left", 0),
+                    params.get("top", 0),
+                    params.get("right", img.width),
+                    params.get("bottom", img.height),
+                )
+            )
+        elif operation == "flip":
+            if params.get("direction", "horizontal") == "horizontal":
+                img = img.transpose(Image.FLIP_LEFT_RIGHT)
+            else:
+                img = img.transpose(Image.FLIP_TOP_BOTTOM)
+        elif operation == "adjust_brightness":
+            img = ImageEnhance.Brightness(img).enhance(params.get("factor", 1.5))
+        elif operation == "adjust_contrast":
+            img = ImageEnhance.Contrast(img).enhance(params.get("factor", 1.5))
+        elif operation == "blur":
+            img = img.filter(ImageFilter.GaussianBlur(params.get("radius", 2)))
+        elif operation == "sharpen":
+            img = img.filter(ImageFilter.SHARPEN)
+        elif operation == "grayscale":
+            img = img.convert("L")
+        else:
+            return {"error": f"Unknown operation: {operation}"}
+
+        result_path = save_image(img)
+        result_base64 = encode_image(result_path)
+        return {"transformed_image": result_base64}
+
+    except Exception as e:
+        return {"error": str(e)}
+
+
+@tool
+def draw_on_image(
+    image_base64: str, drawing_type: str, params: Dict[str, Any]
+) -> Dict[str, Any]:
+    """
+    Draw shapes (rectangle, circle, line) or text onto an image.
+
+    Args:
+        image_base64 (str): Base64 encoded input image
+        drawing_type (str): Drawing type
+        params (Dict[str, Any]): Drawing parameters
+
+    Returns:
+        Dictionary with result image (base64)
+    """
+    try:
+        img = decode_image(image_base64)
+        draw = ImageDraw.Draw(img)
+        color = params.get("color", "red")
+
+        if drawing_type == "rectangle":
+            draw.rectangle(
+                [params["left"], params["top"], params["right"], params["bottom"]],
+                outline=color,
+                width=params.get("width", 2),
+            )
+        elif drawing_type == "circle":
+            x, y, r = params["x"], params["y"], params["radius"]
+            draw.ellipse(
+                (x - r, y - r, x + r, y + r),
+                outline=color,
+                width=params.get("width", 2),
+            )
+        elif drawing_type == "line":
+            draw.line(
+                (
+                    params["start_x"],
+                    params["start_y"],
+                    params["end_x"],
+                    params["end_y"],
+                ),
+                fill=color,
+                width=params.get("width", 2),
+            )
+        elif drawing_type == "text":
+            font_size = params.get("font_size", 20)
+            try:
+                font = ImageFont.truetype("arial.ttf", font_size)
+            except IOError:
+                font = ImageFont.load_default()
+            draw.text(
+                (params["x"], params["y"]),
+                params.get("text", "Text"),
+                fill=color,
+                font=font,
+            )
+        else:
+            return {"error": f"Unknown drawing type: {drawing_type}"}
+
+        result_path = save_image(img)
+        result_base64 = encode_image(result_path)
+        return {"result_image": result_base64}
+
+    except Exception as e:
+        return {"error": str(e)}
+
+
+@tool
+def generate_simple_image(
+    image_type: str,
+    width: int = 500,
+    height: int = 500,
+    params: Optional[Dict[str, Any]] = None,
+) -> Dict[str, Any]:
+    """
+    Generate a simple image (gradient, noise, pattern, chart).
+
+    Args:
+        image_type (str): Type of image
+        width (int), height (int)
+        params (Dict[str, Any], optional): Specific parameters
+
+    Returns:
+        Dictionary with generated image (base64)
+    """
+    try:
+        params = params or {}
+
+        if image_type == "gradient":
+            direction = params.get("direction", "horizontal")
+            start_color = params.get("start_color", (255, 0, 0))
+            end_color = params.get("end_color", (0, 0, 255))
+
+            img = Image.new("RGB", (width, height))
+            draw = ImageDraw.Draw(img)
+
+            if direction == "horizontal":
+                for x in range(width):
+                    r = int(
+                        start_color[0] + (end_color[0] - start_color[0]) * x / width
+                    )
+                    g = int(
+                        start_color[1] + (end_color[1] - start_color[1]) * x / width
+                    )
+                    b = int(
+                        start_color[2] + (end_color[2] - start_color[2]) * x / width
+                    )
+                    draw.line([(x, 0), (x, height)], fill=(r, g, b))
+            else:
+                for y in range(height):
+                    r = int(
+                        start_color[0] + (end_color[0] - start_color[0]) * y / height
+                    )
+                    g = int(
+                        start_color[1] + (end_color[1] - start_color[1]) * y / height
+                    )
+                    b = int(
+                        start_color[2] + (end_color[2] - start_color[2]) * y / height
+                    )
+                    draw.line([(0, y), (width, y)], fill=(r, g, b))
+
+        elif image_type == "noise":
+            noise_array = np.random.randint(0, 256, (height, width, 3), dtype=np.uint8)
+            img = Image.fromarray(noise_array, "RGB")
+
+        else:
+            return {"error": f"Unsupported image_type {image_type}"}
+
+        result_path = save_image(img)
+        result_base64 = encode_image(result_path)
+        return {"generated_image": result_base64}
+
+    except Exception as e:
+        return {"error": str(e)}
+
+
+@tool
+def combine_images(
+    images_base64: List[str], operation: str, params: Optional[Dict[str, Any]] = None
+) -> Dict[str, Any]:
+    """
+    Combine multiple images (collage, stack, blend).
+
+    Args:
+        images_base64 (List[str]): List of base64 images
+        operation (str): Combination type
+        params (Dict[str, Any], optional)
+
+    Returns:
+        Dictionary with combined image (base64)
+    """
+    try:
+        images = [decode_image(b64) for b64 in images_base64]
+        params = params or {}
+
+        if operation == "stack":
+            direction = params.get("direction", "horizontal")
+            if direction == "horizontal":
+                total_width = sum(img.width for img in images)
+                max_height = max(img.height for img in images)
+                new_img = Image.new("RGB", (total_width, max_height))
+                x = 0
+                for img in images:
+                    new_img.paste(img, (x, 0))
+                    x += img.width
+            else:
+                max_width = max(img.width for img in images)
+                total_height = sum(img.height for img in images)
+                new_img = Image.new("RGB", (max_width, total_height))
+                y = 0
+                for img in images:
+                    new_img.paste(img, (0, y))
+                    y += img.height
+        else:
+            return {"error": f"Unsupported combination operation {operation}"}
+
+        result_path = save_image(new_img)
+        result_base64 = encode_image(result_path)
+        return {"combined_image": result_base64}
+
+    except Exception as e:
+        return {"error": str(e)}
 
 
 # load the system prompt from the file
 with open("system_prompt.txt", "r", encoding="utf-8") as f:
     system_prompt = f.read()
+print(system_prompt)
 
 # System message
 sys_msg = SystemMessage(content=system_prompt)
 
 # build a retriever
-embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2") #  dim=768
+embeddings = HuggingFaceEmbeddings(
+    model_name="sentence-transformers/all-mpnet-base-v2"
+)  #  dim=768
 supabase: Client = create_client(
-    os.environ.get("SUPABASE_URL"), 
-    os.environ.get("SUPABASE_SERVICE_KEY"))
+    os.environ.get("SUPABASE_URL"), os.environ.get("SUPABASE_SERVICE_ROLE_KEY")
+)
 vector_store = SupabaseVectorStore(
     client=supabase,
-    embedding= embeddings,
-    table_name="documents",
-    query_name="match_documents_langchain",
+    embedding=embeddings,
+    table_name="documents2",
+    query_name="match_documents_2",
 )
 create_retriever_tool = create_retriever_tool(
     retriever=vector_store.as_retriever(),
@@ -139,38 +707,53 @@ create_retriever_tool = create_retriever_tool(
 )
 
 
-
 tools = [
+    web_search,
+    wiki_search,
+    arxiv_search,
     multiply,
     add,
     subtract,
     divide,
     modulus,
-    wiki_search,
-    web_search,
-    arvix_search,
+    power,
+    square_root,
+    save_and_read_file,
+    download_file_from_url,
+    extract_text_from_image,
+    analyze_csv_file,
+    analyze_excel_file,
+    execute_code_multilang,
+    analyze_image,
+    transform_image,
+    draw_on_image,
+    generate_simple_image,
+    combine_images,
 ]
 
+
 # Build graph function
 def build_graph(provider: str = "groq"):
     """Build the graph"""
     # Load environment variables from .env file
-    if provider == "google":
-        # Google Gemini
-        llm = ChatGoogleGenerativeAI(model="gemini-2.0-flash", temperature=0)
-    elif provider == "groq":
+    if provider == "groq":
         # Groq https://console.groq.com/docs/models
-        llm = ChatGroq(model="qwen-qwq-32b", temperature=0) # optional : qwen-qwq-32b gemma2-9b-it
+        llm = ChatGroq(model="qwen-qwq-32b", temperature=0)
     elif provider == "huggingface":
         # TODO: Add huggingface endpoint
         llm = ChatHuggingFace(
             llm=HuggingFaceEndpoint(
-                url="https://api-inference.huggingface.co/models/Meta-DeepLearning/llama-2-7b-chat-hf",
+                repo_id="TinyLlama/TinyLlama-1.1B-Chat-v1.0",
+                task="text-generation",  # for chat‐style use “text-generation”
+                max_new_tokens=1024,
+                do_sample=False,
+                repetition_penalty=1.03,
                 temperature=0,
             ),
+            verbose=True,
         )
     else:
-        raise ValueError("Invalid provider. Choose 'google', 'groq' or 'huggingface'.")
+        raise ValueError("Invalid provider. Choose 'groq' or 'huggingface'.")
     # Bind tools to LLM
     llm_with_tools = llm.bind_tools(tools)
 
@@ -178,14 +761,19 @@ def build_graph(provider: str = "groq"):
     def assistant(state: MessagesState):
         """Assistant node"""
         return {"messages": [llm_with_tools.invoke(state["messages"])]}
-    
+
     def retriever(state: MessagesState):
         """Retriever node"""
         similar_question = vector_store.similarity_search(state["messages"][0].content)
-        example_msg = HumanMessage(
-            content=f"Here I provide a similar question and answer for reference: \n\n{similar_question[0].page_content}",
-        )
-        return {"messages": [sys_msg] + state["messages"] + [example_msg]}
+
+        if similar_question:  # Check if the list is not empty
+            example_msg = HumanMessage(
+                content=f"Here I provide a similar question and answer for reference: \n\n{similar_question[0].page_content}",
+            )
+            return {"messages": [sys_msg] + state["messages"] + [example_msg]}
+        else:
+            # Handle the case when no similar questions are found
+            return {"messages": [sys_msg] + state["messages"]}
 
     builder = StateGraph(MessagesState)
     builder.add_node("retriever", retriever)
@@ -202,12 +790,11 @@ def build_graph(provider: str = "groq"):
     # Compile graph
     return builder.compile()
 
+
 # test
 if __name__ == "__main__":
     question = "When was a picture of St. Thomas Aquinas first added to the Wikipedia page on the Principle of double effect?"
-    # Build the graph
     graph = build_graph(provider="groq")
-    # Run the graph
     messages = [HumanMessage(content=question)]
     messages = graph.invoke({"messages": messages})
     for m in messages["messages"]:

app.py

@@ -4,6 +4,7 @@ import inspect
 import gradio as gr
 import requests
 import pandas as pd
+import time
 from langchain_core.messages import HumanMessage
 from agent import build_graph
 
@@ -92,6 +93,9 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
         if not task_id or question_text is None:
             print(f"Skipping item with missing task_id or question: {item}")
             continue
+        
+        # time.sleep(10)
+        
         try:
             submitted_answer = agent(question_text)
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
@@ -159,11 +163,9 @@ with gr.Blocks() as demo:
     gr.Markdown(
         """
         **Instructions:**
-
         1.  Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
         2.  Log in to your Hugging Face account using the button below. This uses your HF username for submission.
         3.  Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
-
         ---
         **Disclaimers:**
         Once clicking on the "submit button, it can take quite some time ( this is the time for the agent to go through all the questions).
@@ -206,4 +208,4 @@ if __name__ == "__main__":
     print("-"*(60 + len(" App Starting ")) + "\n")
 
     print("Launching Gradio Interface for Basic Agent Evaluation...")
-    demo.launch(debug=True, share=False)
+    demo.launch(debug=True, share=False)

code_interpreter.py

@@ -0,0 +1,281 @@
+import os
+import io
+import sys
+import uuid
+import base64
+import traceback
+import contextlib
+import tempfile
+import subprocess
+import sqlite3
+from typing import Dict, List, Any, Optional, Union
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+from PIL import Image
+
+class CodeInterpreter:
+    def __init__(self, allowed_modules=None, max_execution_time=30, working_directory=None):
+        """Initialize the code interpreter with safety measures."""
+        self.allowed_modules = allowed_modules or [
+            "numpy", "pandas", "matplotlib", "scipy", "sklearn", 
+            "math", "random", "statistics", "datetime", "collections",
+            "itertools", "functools", "operator", "re", "json",
+            "sympy", "networkx", "nltk", "PIL", "pytesseract", 
+            "cmath", "uuid", "tempfile", "requests", "urllib"
+        ]
+        self.max_execution_time = max_execution_time
+        self.working_directory = working_directory or os.path.join(os.getcwd()) 
+        if not os.path.exists(self.working_directory):
+            os.makedirs(self.working_directory)
+        
+        self.globals = {
+            "__builtins__": __builtins__,
+            "np": np,
+            "pd": pd,
+            "plt": plt,
+            "Image": Image,
+        }
+        self.temp_sqlite_db = os.path.join(tempfile.gettempdir(), "code_exec.db")
+
+    def execute_code(self, code: str, language: str = "python") -> Dict[str, Any]:
+        """Execute the provided code in the selected programming language."""
+        language = language.lower()
+        execution_id = str(uuid.uuid4())
+        
+        result = {
+            "execution_id": execution_id,
+            "status": "error",
+            "stdout": "",
+            "stderr": "",
+            "result": None,
+            "plots": [],
+            "dataframes": []
+        }
+        
+        try:
+            if language == "python":
+                return self._execute_python(code, execution_id)
+            elif language == "bash":
+                return self._execute_bash(code, execution_id)
+            elif language == "sql":
+                return self._execute_sql(code, execution_id)
+            elif language == "c":
+                return self._execute_c(code, execution_id)
+            elif language == "java":
+                return self._execute_java(code, execution_id)
+            else:
+                result["stderr"] = f"Unsupported language: {language}"
+        except Exception as e:
+            result["stderr"] = str(e)
+        
+        return result
+
+    def _execute_python(self, code: str, execution_id: str) -> dict:
+        output_buffer = io.StringIO()
+        error_buffer = io.StringIO()
+        result = {
+            "execution_id": execution_id,
+            "status": "error",
+            "stdout": "",
+            "stderr": "",
+            "result": None,
+            "plots": [],
+            "dataframes": []
+        }
+        
+        try:
+            exec_dir = os.path.join(self.working_directory, execution_id)
+            os.makedirs(exec_dir, exist_ok=True)
+            plt.switch_backend('Agg')
+            
+            with contextlib.redirect_stdout(output_buffer), contextlib.redirect_stderr(error_buffer):
+                exec_result = exec(code, self.globals)
+
+                if plt.get_fignums():
+                    for i, fig_num in enumerate(plt.get_fignums()):
+                        fig = plt.figure(fig_num)
+                        img_path = os.path.join(exec_dir, f"plot_{i}.png")
+                        fig.savefig(img_path)
+                        with open(img_path, "rb") as img_file:
+                            img_data = base64.b64encode(img_file.read()).decode('utf-8')
+                            result["plots"].append({
+                                "figure_number": fig_num,
+                                "data": img_data
+                            })
+
+                for var_name, var_value in self.globals.items():
+                    if isinstance(var_value, pd.DataFrame) and len(var_value) > 0:
+                        result["dataframes"].append({
+                            "name": var_name,
+                            "head": var_value.head().to_dict(),
+                            "shape": var_value.shape,
+                            "dtypes": str(var_value.dtypes)
+                        })
+                
+            result["status"] = "success"
+            result["stdout"] = output_buffer.getvalue()
+            result["result"] = exec_result
+            
+        except Exception as e:
+            result["status"] = "error"
+            result["stderr"] = f"{error_buffer.getvalue()}\n{traceback.format_exc()}"
+        
+        return result
+
+    def _execute_bash(self, code: str, execution_id: str) -> dict:
+        try:
+            completed = subprocess.run(
+                code, shell=True, capture_output=True, text=True, timeout=self.max_execution_time
+            )
+            return {
+                "execution_id": execution_id,
+                "status": "success" if completed.returncode == 0 else "error",
+                "stdout": completed.stdout,
+                "stderr": completed.stderr,
+                "result": None,
+                "plots": [],
+                "dataframes": []
+            }
+        except subprocess.TimeoutExpired:
+            return {
+                "execution_id": execution_id,
+                "status": "error",
+                "stdout": "",
+                "stderr": "Execution timed out.",
+                "result": None,
+                "plots": [],
+                "dataframes": []
+            }
+
+    def _execute_sql(self, code: str, execution_id: str) -> dict:
+        result = {
+            "execution_id": execution_id,
+            "status": "error",
+            "stdout": "",
+            "stderr": "",
+            "result": None,
+            "plots": [],
+            "dataframes": []
+        }
+        try:
+            conn = sqlite3.connect(self.temp_sqlite_db)
+            cur = conn.cursor()
+            cur.execute(code)
+            if code.strip().lower().startswith("select"):
+                columns = [description[0] for description in cur.description]
+                rows = cur.fetchall()
+                df = pd.DataFrame(rows, columns=columns)
+                result["dataframes"].append({
+                    "name": "query_result",
+                    "head": df.head().to_dict(),
+                    "shape": df.shape,
+                    "dtypes": str(df.dtypes)
+                })
+            else:
+                conn.commit()
+
+            result["status"] = "success"
+            result["stdout"] = "Query executed successfully."
+
+        except Exception as e:
+            result["stderr"] = str(e)
+        finally:
+            conn.close()
+
+        return result
+
+    def _execute_c(self, code: str, execution_id: str) -> dict:
+        temp_dir = tempfile.mkdtemp()
+        source_path = os.path.join(temp_dir, "program.c")
+        binary_path = os.path.join(temp_dir, "program")
+
+        try:
+            with open(source_path, "w") as f:
+                f.write(code)
+
+            compile_proc = subprocess.run(
+                ["gcc", source_path, "-o", binary_path],
+                capture_output=True, text=True, timeout=self.max_execution_time
+            )
+            if compile_proc.returncode != 0:
+                return {
+                    "execution_id": execution_id,
+                    "status": "error",
+                    "stdout": compile_proc.stdout,
+                    "stderr": compile_proc.stderr,
+                    "result": None,
+                    "plots": [],
+                    "dataframes": []
+                }
+
+            run_proc = subprocess.run(
+                [binary_path],
+                capture_output=True, text=True, timeout=self.max_execution_time
+            )
+            return {
+                "execution_id": execution_id,
+                "status": "success" if run_proc.returncode == 0 else "error",
+                "stdout": run_proc.stdout,
+                "stderr": run_proc.stderr,
+                "result": None,
+                "plots": [],
+                "dataframes": []
+            }
+        except Exception as e:
+            return {
+                "execution_id": execution_id,
+                "status": "error",
+                "stdout": "",
+                "stderr": str(e),
+                "result": None,
+                "plots": [],
+                "dataframes": []
+            }
+
+    def _execute_java(self, code: str, execution_id: str) -> dict:
+        temp_dir = tempfile.mkdtemp()
+        source_path = os.path.join(temp_dir, "Main.java")
+
+        try:
+            with open(source_path, "w") as f:
+                f.write(code)
+
+            compile_proc = subprocess.run(
+                ["javac", source_path],
+                capture_output=True, text=True, timeout=self.max_execution_time
+            )
+            if compile_proc.returncode != 0:
+                return {
+                    "execution_id": execution_id,
+                    "status": "error",
+                    "stdout": compile_proc.stdout,
+                    "stderr": compile_proc.stderr,
+                    "result": None,
+                    "plots": [],
+                    "dataframes": []
+                }
+
+            run_proc = subprocess.run(
+                ["java", "-cp", temp_dir, "Main"],
+                capture_output=True, text=True, timeout=self.max_execution_time
+            )
+            return {
+                "execution_id": execution_id,
+                "status": "success" if run_proc.returncode == 0 else "error",
+                "stdout": run_proc.stdout,
+                "stderr": run_proc.stderr,
+                "result": None,
+                "plots": [],
+                "dataframes": []
+            }
+        except Exception as e:
+            return {
+                "execution_id": execution_id,
+                "status": "error",
+                "stdout": "",
+                "stderr": str(e),
+                "result": None,
+                "plots": [],
+                "dataframes": []
+            }

explore_metadata.ipynb

@@ -0,0 +1,332 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "a600d7fc",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import json \n",
+    "with open('metadata.jsonl', 'r') as f: \n",
+    "    json_list = list(f)\n",
+    "\n",
+    "json_QA = []\n",
+    "for json_str in json_list: \n",
+    "    json_data = json.loads(json_str)\n",
+    "    json_QA.append(json_data)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "fa5d8eb8",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "==================================================\n",
+      "Task ID: d1af70ea-a9a4-421a-b9cc-94b5e02f1788\n",
+      "Question: As of the 2020 census, what was the population difference between the largest county seat and smallest county seat, by land area of the county seat, in Washington state? For population figures, please use the official data from data.census.gov. Please report the integer difference.\n",
+      "Level: 2\n",
+      "Final Answer: 736455\n",
+      "Annotator Metadata: \n",
+      "  ├── Steps: \n",
+      "  │      ├── Step 1: Using a web browser, access a search engine and conduct a search, \"Washington cities by area\"\n",
+      "  │      ├── Step 2: Navigate to the second search result, https://en.wikipedia.org/wiki/List_of_municipalities_in_Washington\n",
+      "  │      ├── Step 3: Evaluate the page contents, finding the largest and smallest county seats by land area, Seattle and Cathlamet\n",
+      "  │      ├── Step 4: Using a web browser, navigate to https://data.census.gov/\n",
+      "  │      ├── Step 5: Using the website's search area, conduct a search, Seattle, Washington\n",
+      "  │      ├── Step 6: Record the reported 2020 Decennial Census population of Seattle, Washington, 737,015\n",
+      "  │      ├── Step 7: Using the website's search area, conduct a search, Cathlamet, Washington\n",
+      "  │      ├── Step 8: Record the reported 2020 Decennial Census population of Cathlamet, Washington, 560\n",
+      "  │      ├── Step 9: Using a calculator, find the difference in populations,\n",
+      "  │      ├── \n",
+      "  │      ├── 737,015 - 560\n",
+      "  │      ├── 736,455\n",
+      "  │      ├── Step 10: Report the correct answer to my user in the requested format, \"736,455\"\n",
+      "  ├── Number of steps: 10\n",
+      "  ├── How long did this take?: 5 minutes\n",
+      "  ├── Tools:\n",
+      "  │      ├── 1. A web browser\n",
+      "  │      ├── 2. A search engine\n",
+      "  │      ├── 3. A calculator\n",
+      "  └── Number of tools: 3\n",
+      "==================================================\n"
+     ]
+    }
+   ],
+   "source": [
+    "import random\n",
+    "random_samples = random.sample(json_QA, 1)\n",
+    "for sample in random_samples:\n",
+    "    print(\"=\" * 50)\n",
+    "    print(f\"Task ID: {sample['task_id']}\")\n",
+    "    print(f\"Question: {sample['Question']}\")\n",
+    "    print(f\"Level: {sample['Level']}\")\n",
+    "    print(f\"Final Answer: {sample['Final answer']}\")\n",
+    "    print(f\"Annotator Metadata: \")\n",
+    "    print(f\"  ├── Steps: \")\n",
+    "    for step in sample['Annotator Metadata']['Steps'].split('\\n'):\n",
+    "        print(f\"  │      ├── {step}\")\n",
+    "    print(f\"  ├── Number of steps: {sample['Annotator Metadata']['Number of steps']}\")\n",
+    "    print(f\"  ├── How long did this take?: {sample['Annotator Metadata']['How long did this take?']}\")\n",
+    "    print(f\"  ├── Tools:\")\n",
+    "    for tool in sample['Annotator Metadata']['Tools'].split('\\n'):\n",
+    "        print(f\"  │      ├── {tool}\")\n",
+    "    print(f\"  └── Number of tools: {sample['Annotator Metadata']['Number of tools']}\")\n",
+    "print(\"=\" * 50)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "05076516",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "from dotenv import load_dotenv\n",
+    "from langchain_huggingface import HuggingFaceEmbeddings\n",
+    "from langchain_community.vectorstores import SupabaseVectorStore\n",
+    "from supabase.client import Client, create_client\n",
+    "\n",
+    "\n",
+    "load_dotenv()\n",
+    "embeddings = HuggingFaceEmbeddings(model_name=\"sentence-transformers/all-mpnet-base-v2\") #  dim=768\n",
+    "\n",
+    "supabase_url = os.environ.get(\"SUPABASE_URL\")\n",
+    "supabase_key = os.environ.get(\"SUPABASE_SERVICE_ROLE_KEY\")\n",
+    "supabase: Client = create_client(supabase_url, supabase_key)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 20,
+   "id": "aa1402e3",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from langchain.schema import Document\n",
+    "docs = []\n",
+    "cnt = 0 \n",
+    "for sample in json_QA:\n",
+    "    content = f\"Question : {sample['Question']}\\n\\nFinal answer : {sample['Final answer']}\"\n",
+    "    doc = {\n",
+    "        \"id\" : cnt,\n",
+    "        \"content\" : content,\n",
+    "        \"metadata\" : {\n",
+    "            \"source\" : sample['task_id']\n",
+    "        },\n",
+    "        \"embedding\" : embeddings.embed_query(content),\n",
+    "    }\n",
+    "    docs.append(doc)\n",
+    "    cnt += 1\n",
+    "\n",
+    "# upload the documents to the vector database\n",
+    "try:\n",
+    "    response = (\n",
+    "        supabase.table(\"documents2\")\n",
+    "        .insert(docs)\n",
+    "        .execute()\n",
+    "    )\n",
+    "except Exception as exception:\n",
+    "    print(\"Error inserting data into Supabase:\", exception)\n",
+    "\n",
+    "# # Save the documents (a list of dict) into a csv file, and manually upload it to Supabase\n",
+    "# import pandas as pd\n",
+    "# df = pd.DataFrame(docs)\n",
+    "# df.to_csv('supabase_docs.csv',index=False)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 41,
+   "id": "9aa7eb5e",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# add items to vector database\n",
+    "vector_store = SupabaseVectorStore(\n",
+    "    client=supabase,\n",
+    "    embedding= embeddings,\n",
+    "    table_name=\"documents2\",\n",
+    "    query_name=\"match_documents_2\",\n",
+    ")\n",
+    "retriever = vector_store.as_retriever()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 42,
+   "id": "9eecafd1",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "query = \"On June 6, 2023, an article by Carolyn Collins Petersen was published in Universe Today. This article mentions a team that produced a paper about their observations, linked at the bottom of the article. Find this paper. Under what NASA award number was the work performed by R. G. Arendt supported by?\"\n",
+    "# matched_docs = vector_store.similarity_search(query, k=2)\n",
+    "docs = retriever.invoke(query)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 43,
+   "id": "ff917840",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Document(metadata={'source': '840bfca7-4f7b-481a-8794-c560c340185d'}, page_content='Question : On June 6, 2023, an article by Carolyn Collins Petersen was published in Universe Today. This article mentions a team that produced a paper about their observations, linked at the bottom of the article. Find this paper. Under what NASA award number was the work performed by R. G. Arendt supported by?\\n\\nFinal answer : 80GSFC21M0002')"
+      ]
+     },
+     "execution_count": 43,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "docs[0]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 44,
+   "id": "01c8f337",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "List of tools used in all samples:\n",
+      "Total number of tools used: 83\n",
+      "  ├── web browser: 107\n",
+      "  ├── image recognition tools (to identify and parse a figure with three axes): 1\n",
+      "  ├── search engine: 101\n",
+      "  ├── calculator: 34\n",
+      "  ├── unlambda compiler (optional): 1\n",
+      "  ├── a web browser.: 2\n",
+      "  ├── a search engine.: 2\n",
+      "  ├── a calculator.: 1\n",
+      "  ├── microsoft excel: 5\n",
+      "  ├── google search: 1\n",
+      "  ├── ne: 9\n",
+      "  ├── pdf access: 7\n",
+      "  ├── file handling: 2\n",
+      "  ├── python: 3\n",
+      "  ├── image recognition tools: 12\n",
+      "  ├── jsonld file access: 1\n",
+      "  ├── video parsing: 1\n",
+      "  ├── python compiler: 1\n",
+      "  ├── video recognition tools: 3\n",
+      "  ├── pdf viewer: 7\n",
+      "  ├── microsoft excel / google sheets: 3\n",
+      "  ├── word document access: 1\n",
+      "  ├── tool to extract text from images: 1\n",
+      "  ├── a word reversal tool / script: 1\n",
+      "  ├── counter: 1\n",
+      "  ├── excel: 3\n",
+      "  ├── image recognition: 5\n",
+      "  ├── color recognition: 3\n",
+      "  ├── excel file access: 3\n",
+      "  ├── xml file access: 1\n",
+      "  ├── access to the internet archive, web.archive.org: 1\n",
+      "  ├── text processing/diff tool: 1\n",
+      "  ├── gif parsing tools: 1\n",
+      "  ├── a web browser: 7\n",
+      "  ├── a search engine: 7\n",
+      "  ├── a speech-to-text tool: 2\n",
+      "  ├── code/data analysis tools: 1\n",
+      "  ├── audio capability: 2\n",
+      "  ├── pdf reader: 1\n",
+      "  ├── markdown: 1\n",
+      "  ├── a calculator: 5\n",
+      "  ├── access to wikipedia: 3\n",
+      "  ├── image recognition/ocr: 3\n",
+      "  ├── google translate access: 1\n",
+      "  ├── ocr: 4\n",
+      "  ├── bass note data: 1\n",
+      "  ├── text editor: 1\n",
+      "  ├── xlsx file access: 1\n",
+      "  ├── powerpoint viewer: 1\n",
+      "  ├── csv file access: 1\n",
+      "  ├── calculator (or use excel): 1\n",
+      "  ├── computer algebra system: 1\n",
+      "  ├── video processing software: 1\n",
+      "  ├── audio processing software: 1\n",
+      "  ├── computer vision: 1\n",
+      "  ├── google maps: 1\n",
+      "  ├── access to excel files: 1\n",
+      "  ├── calculator (or ability to count): 1\n",
+      "  ├── a file interface: 3\n",
+      "  ├── a python ide: 1\n",
+      "  ├── spreadsheet editor: 1\n",
+      "  ├── tools required: 1\n",
+      "  ├── b browser: 1\n",
+      "  ├── image recognition and processing tools: 1\n",
+      "  ├── computer vision or ocr: 1\n",
+      "  ├── c++ compiler: 1\n",
+      "  ├── access to google maps: 1\n",
+      "  ├── youtube player: 1\n",
+      "  ├── natural language processor: 1\n",
+      "  ├── graph interaction tools: 1\n",
+      "  ├── bablyonian cuniform -> arabic legend: 1\n",
+      "  ├── access to youtube: 1\n",
+      "  ├── image search tools: 1\n",
+      "  ├── calculator or counting function: 1\n",
+      "  ├── a speech-to-text audio processing tool: 1\n",
+      "  ├── access to academic journal websites: 1\n",
+      "  ├── pdf reader/extracter: 1\n",
+      "  ├── rubik's cube model: 1\n",
+      "  ├── wikipedia: 1\n",
+      "  ├── video capability: 1\n",
+      "  ├── image processing tools: 1\n",
+      "  ├── age recognition software: 1\n",
+      "  ├── youtube: 1\n"
+     ]
+    }
+   ],
+   "source": [
+    "# list of the tools used in all the samples\n",
+    "from collections import Counter, OrderedDict\n",
+    "\n",
+    "tools = []\n",
+    "for sample in json_QA:\n",
+    "    for tool in sample['Annotator Metadata']['Tools'].split('\\n'):\n",
+    "        tool = tool[2:].strip().lower()\n",
+    "        if tool.startswith(\"(\"):\n",
+    "            tool = tool[11:].strip()\n",
+    "        tools.append(tool)\n",
+    "tools_counter = OrderedDict(Counter(tools))\n",
+    "print(\"List of tools used in all samples:\")\n",
+    "print(\"Total number of tools used:\", len(tools_counter))\n",
+    "for tool, count in tools_counter.items():\n",
+    "    print(f\"  ├── {tool}: {count}\")"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "env",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

gitignore

@@ -0,0 +1,177 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/latest/usage/project/#working-with-version-control
+.pdm.toml
+.pdm-python
+.pdm-build/
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+###
+/image_outputs

image_processing.py

@@ -0,0 +1,26 @@
+import os
+import io
+import base64
+import uuid
+from PIL import Image
+
+# Helper functions for image processing
+def encode_image(image_path: str) -> str:
+    """Convert an image file to base64 string."""
+    with open(image_path, "rb") as image_file:
+        return base64.b64encode(image_file.read()).decode("utf-8")
+
+
+def decode_image(base64_string: str) -> Image.Image:
+    """Convert a base64 string to a PIL Image."""
+    image_data = base64.b64decode(base64_string)
+    return Image.open(io.BytesIO(image_data))
+
+
+def save_image(image: Image.Image, directory: str = "image_outputs") -> str:
+    """Save a PIL Image to disk and return the path."""
+    os.makedirs(directory, exist_ok=True)
+    image_id = str(uuid.uuid4())
+    image_path = os.path.join(directory, f"{image_id}.png")
+    image.save(image_path)
+    return image_path

requirements.txt

@@ -15,4 +15,6 @@ arxiv
 pymupdf
 wikipedia
 pgvector
-python-dotenv
+python-dotenv
+pytesseract
+matplotlib

results.csv

@@ -1,15 +0,0 @@
-Task ID,Question,Submitted Answer
-8e867cd7-cff9-4e6c-867a-ff5ddc2550be,How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use the latest 2022 version of english wikipedia.,3
-2d83110e-a098-4ebb-9987-066c06fa42d0,".rewsna eht sa ""tfel"" drow eht fo etisoppo eht etirw ,ecnetnes siht dnatsrednu uoy fI",right
-4fc2f1ae-8625-45b5-ab34-ad4433bc21f8,Who nominated the only Featured Article on English Wikipedia about a dinosaur that was promoted in November 2016?,FunkMonk
-cabe07ed-9eca-40ea-8ead-410ef5e83f91,What is the surname of the equine veterinarian mentioned in 1.E Exercises from the chemistry materials licensed by Marisa Alviar-Agnew & Henry Agnew under the CK-12 license in LibreText's Introductory Chemistry materials as compiled 08/21/2023?,not found in materials
-3cef3a44-215e-4aed-8e3b-b1e3f08063b7,"I'm making a grocery list for my mom, but she's a professor of botany and she's a real stickler when it comes to categorizing things. I need to add different foods to different categories on the grocery list, but if I make a mistake, she won't buy anything inserted in the wrong category. Here's the list I have so far:
-
-milk, eggs, flour, whole bean coffee, Oreos, sweet potatoes, fresh basil, plums, green beans, rice, corn, bell pepper, whole allspice, acorns, broccoli, celery, zucchini, lettuce, peanuts
-
-I need to make headings for the fruits and vegetables. Could you please create a list of just the vegetables from my list? If you could do that, then I can figure out how to categorize the rest of the list into the appropriate categories. But remember that my mom is a real stickler, so make sure that no botanical fruits end up on the vegetable list, or she won't get them when she's at the store. Please alphabetize the list of vegetables, and place each item in a comma separated list.","broccoli, celery, green beans, lettuce, sweet potatoes, zucchini"
-305ac316-eef6-4446-960a-92d80d542f82,Who did the actor who played Ray in the Polish-language version of Everybody Loves Raymond play in Magda M.? Give only the first name.,Wojciech
-3f57289b-8c60-48be-bd80-01f8099ca449,How many at bats did the Yankee with the most walks in the 1977 regular season have that same season?,519
-cf106601-ab4f-4af9-b045-5295fe67b37d,"What country had the least number of athletes at the 1928 Summer Olympics? If there's a tie for a number of athletes, return the first in alphabetical order. Give the IOC country code as your answer.",CUB
-a0c07678-e491-4bbc-8f0b-07405144218f,"Who are the pitchers with the number before and after Taishō Tamai's number as of July 2023? Give them to me in the form Pitcher Before, Pitcher After, use their last names only, in Roman characters.","Yamasaki, Uehara"
-5a0c1adf-205e-4841-a666-7c3ef95def9d,What is the first name of the only Malko Competition recipient from the 20th Century (after 1977) whose nationality on record is a country that no longer exists?,Claus Peter

system_prompt.txt

@@ -1,5 +1,5 @@
 You are a helpful assistant tasked with answering questions using a set of tools. 
 Now, I will ask you a question. Report your thoughts, and finish your answer with the following template: 
 FINAL ANSWER: [YOUR FINAL ANSWER]. 
-YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings. If you are asked for a number, don't use comma to write your number neither use units such as $ or percent sign unless specified otherwise. If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise. If you are asked for a comma separated list, apply the above rules depending of whether the element to be put in the list is a number or a string.
-Your answer should only start with "FINAL ANSWER: ", then follows with the answer. 
+YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings. If you are asked for a number, don't use comma to write your number neither use units such as $ or percent sign unless specified otherwise. If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise. If you are asked for a comma separated list, Apply the rules above for each element (number or string), ensure there is exactly one space after each comma.
+Your answer should only start with "FINAL ANSWER: ", then follows with the answer. 

test.ipynb

@@ -1,684 +0,0 @@
-{
- "cells": [
-  {
-   "cell_type": "markdown",
-   "id": "d0cc4adf",
-   "metadata": {},
-   "source": [
-    "### Question data"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 2,
-   "id": "14e3f417",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# Load metadata.jsonl\n",
-    "import json\n",
-    "# Load the metadata.jsonl file\n",
-    "with open('metadata.jsonl', 'r') as jsonl_file:\n",
-    "    json_list = list(jsonl_file)\n",
-    "\n",
-    "json_QA = []\n",
-    "for json_str in json_list:\n",
-    "    json_data = json.loads(json_str)\n",
-    "    json_QA.append(json_data)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 3,
-   "id": "5e2da6fc",
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "==================================================\n",
-      "Task ID: ed58682d-bc52-4baa-9eb0-4eb81e1edacc\n",
-      "Question: What is the last word before the second chorus of the King of Pop's fifth single from his sixth studio album?\n",
-      "Level: 2\n",
-      "Final Answer: stare\n",
-      "Annotator Metadata: \n",
-      "  ├── Steps: \n",
-      "  │      ├── 1. Google searched \"King of Pop\".\n",
-      "  │      ├── 2. Clicked on Michael Jackson's Wikipedia.\n",
-      "  │      ├── 3. Scrolled down to \"Discography\".\n",
-      "  │      ├── 4. Clicked on the sixth album, \"Thriller\".\n",
-      "  │      ├── 5. Looked under \"Singles from Thriller\".\n",
-      "  │      ├── 6. Clicked on the fifth single, \"Human Nature\".\n",
-      "  │      ├── 7. Google searched \"Human Nature Michael Jackson Lyrics\".\n",
-      "  │      ├── 8. Looked at the opening result with full lyrics sourced by Musixmatch.\n",
-      "  │      ├── 9. Looked for repeating lyrics to determine the chorus.\n",
-      "  │      ├── 10. Determined the chorus begins with \"If they say\" and ends with \"Does he do me that way?\"\n",
-      "  │      ├── 11. Found the second instance of the chorus within the lyrics.\n",
-      "  │      ├── 12. Noted the last word before the second chorus - \"stare\".\n",
-      "  ├── Number of steps: 12\n",
-      "  ├── How long did this take?: 20 minutes\n",
-      "  ├── Tools:\n",
-      "  │      ├── Web Browser\n",
-      "  └── Number of tools: 1\n",
-      "==================================================\n"
-     ]
-    }
-   ],
-   "source": [
-    "# randomly select 3 samples\n",
-    "# {\"task_id\": \"c61d22de-5f6c-4958-a7f6-5e9707bd3466\", \"Question\": \"A paper about AI regulation that was originally submitted to arXiv.org in June 2022 shows a figure with three axes, where each axis has a label word at both ends. Which of these words is used to describe a type of society in a Physics and Society article submitted to arXiv.org on August 11, 2016?\", \"Level\": 2, \"Final answer\": \"egalitarian\", \"file_name\": \"\", \"Annotator Metadata\": {\"Steps\": \"1. Go to arxiv.org and navigate to the Advanced Search page.\\n2. Enter \\\"AI regulation\\\" in the search box and select \\\"All fields\\\" from the dropdown.\\n3. Enter 2022-06-01 and 2022-07-01 into the date inputs, select \\\"Submission date (original)\\\", and submit the search.\\n4. Go through the search results to find the article that has a figure with three axes and labels on each end of the axes, titled \\\"Fairness in Agreement With European Values: An Interdisciplinary Perspective on AI Regulation\\\".\\n5. Note the six words used as labels: deontological, egalitarian, localized, standardized, utilitarian, and consequential.\\n6. Go back to arxiv.org\\n7. Find \\\"Physics and Society\\\" and go to the page for the \\\"Physics and Society\\\" category.\\n8. Note that the tag for this category is \\\"physics.soc-ph\\\".\\n9. Go to the Advanced Search page.\\n10. Enter \\\"physics.soc-ph\\\" in the search box and select \\\"All fields\\\" from the dropdown.\\n11. Enter 2016-08-11 and 2016-08-12 into the date inputs, select \\\"Submission date (original)\\\", and submit the search.\\n12. Search for instances of the six words in the results to find the paper titled \\\"Phase transition from egalitarian to hierarchical societies driven by competition between cognitive and social constraints\\\", indicating that \\\"egalitarian\\\" is the correct answer.\", \"Number of steps\": \"12\", \"How long did this take?\": \"8 minutes\", \"Tools\": \"1. Web browser\\n2. Image recognition tools (to identify and parse a figure with three axes)\", \"Number of tools\": \"2\"}}\n",
-    "\n",
-    "import random\n",
-    "# random.seed(42)\n",
-    "random_samples = random.sample(json_QA, 1)\n",
-    "for sample in random_samples:\n",
-    "    print(\"=\" * 50)\n",
-    "    print(f\"Task ID: {sample['task_id']}\")\n",
-    "    print(f\"Question: {sample['Question']}\")\n",
-    "    print(f\"Level: {sample['Level']}\")\n",
-    "    print(f\"Final Answer: {sample['Final answer']}\")\n",
-    "    print(f\"Annotator Metadata: \")\n",
-    "    print(f\"  ├── Steps: \")\n",
-    "    for step in sample['Annotator Metadata']['Steps'].split('\\n'):\n",
-    "        print(f\"  │      ├── {step}\")\n",
-    "    print(f\"  ├── Number of steps: {sample['Annotator Metadata']['Number of steps']}\")\n",
-    "    print(f\"  ├── How long did this take?: {sample['Annotator Metadata']['How long did this take?']}\")\n",
-    "    print(f\"  ├── Tools:\")\n",
-    "    for tool in sample['Annotator Metadata']['Tools'].split('\\n'):\n",
-    "        print(f\"  │      ├── {tool}\")\n",
-    "    print(f\"  └── Number of tools: {sample['Annotator Metadata']['Number of tools']}\")\n",
-    "print(\"=\" * 50)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 56,
-   "id": "4bb02420",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "### build a vector database based on the metadata.jsonl\n",
-    "# https://python.langchain.com/docs/integrations/vectorstores/supabase/\n",
-    "import os\n",
-    "from dotenv import load_dotenv\n",
-    "from langchain_huggingface import HuggingFaceEmbeddings\n",
-    "from langchain_community.vectorstores import SupabaseVectorStore\n",
-    "from supabase.client import Client, create_client\n",
-    "\n",
-    "\n",
-    "load_dotenv()\n",
-    "embeddings = HuggingFaceEmbeddings(model_name=\"sentence-transformers/all-mpnet-base-v2\") #  dim=768\n",
-    "\n",
-    "supabase_url = os.environ.get(\"SUPABASE_URL\")\n",
-    "supabase_key = os.environ.get(\"SUPABASE_SERVICE_KEY\")\n",
-    "supabase: Client = create_client(supabase_url, supabase_key)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "a070b955",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# wrap the metadata.jsonl's questions and answers into a list of document\n",
-    "from langchain.schema import Document\n",
-    "docs = []\n",
-    "for sample in json_QA:\n",
-    "    content = f\"Question : {sample['Question']}\\n\\nFinal answer : {sample['Final answer']}\"\n",
-    "    doc = {\n",
-    "        \"content\" : content,\n",
-    "        \"metadata\" : { # meatadata的格式必须时source键，否则会报错\n",
-    "            \"source\" : sample['task_id']\n",
-    "        },\n",
-    "        \"embedding\" : embeddings.embed_query(content),\n",
-    "    }\n",
-    "    docs.append(doc)\n",
-    "\n",
-    "# upload the documents to the vector database\n",
-    "try:\n",
-    "    response = (\n",
-    "        supabase.table(\"documents\")\n",
-    "        .insert(docs)\n",
-    "        .execute()\n",
-    "    )\n",
-    "except Exception as exception:\n",
-    "    print(\"Error inserting data into Supabase:\", exception)\n",
-    "\n",
-    "# ALTERNATIVE : Save the documents (a list of dict) into a csv file, and manually upload it to Supabase\n",
-    "# import pandas as pd\n",
-    "# df = pd.DataFrame(docs)\n",
-    "# df.to_csv('supabase_docs.csv', index=False)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 54,
-   "id": "77fb9dbb",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# add items to vector database\n",
-    "vector_store = SupabaseVectorStore(\n",
-    "    client=supabase,\n",
-    "    embedding= embeddings,\n",
-    "    table_name=\"documents\",\n",
-    "    query_name=\"match_documents_langchain\",\n",
-    ")\n",
-    "retriever = vector_store.as_retriever()"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 55,
-   "id": "12a05971",
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...\n",
-      "To disable this warning, you can either:\n",
-      "\t- Avoid using `tokenizers` before the fork if possible\n",
-      "\t- Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)\n"
-     ]
-    },
-    {
-     "data": {
-      "text/plain": [
-       "Document(metadata={'source': '840bfca7-4f7b-481a-8794-c560c340185d'}, page_content='Question : On June 6, 2023, an article by Carolyn Collins Petersen was published in Universe Today. This article mentions a team that produced a paper about their observations, linked at the bottom of the article. Find this paper. Under what NASA award number was the work performed by R. G. Arendt supported by?\\n\\nFinal answer : 80GSFC21M0002')"
-      ]
-     },
-     "execution_count": 55,
-     "metadata": {},
-     "output_type": "execute_result"
-    }
-   ],
-   "source": [
-    "query = \"On June 6, 2023, an article by Carolyn Collins Petersen was published in Universe Today. This article mentions a team that produced a paper about their observations, linked at the bottom of the article. Find this paper. Under what NASA award number was the work performed by R. G. Arendt supported by?\"\n",
-    "# matched_docs = vector_store.similarity_search(query, 2)\n",
-    "docs = retriever.invoke(query)\n",
-    "docs[0]"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 31,
-   "id": "1eae5ba4",
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "List of tools used in all samples:\n",
-      "Total number of tools used: 83\n",
-      "  ├── web browser: 107\n",
-      "  ├── image recognition tools (to identify and parse a figure with three axes): 1\n",
-      "  ├── search engine: 101\n",
-      "  ├── calculator: 34\n",
-      "  ├── unlambda compiler (optional): 1\n",
-      "  ├── a web browser.: 2\n",
-      "  ├── a search engine.: 2\n",
-      "  ├── a calculator.: 1\n",
-      "  ├── microsoft excel: 5\n",
-      "  ├── google search: 1\n",
-      "  ├── ne: 9\n",
-      "  ├── pdf access: 7\n",
-      "  ├── file handling: 2\n",
-      "  ├── python: 3\n",
-      "  ├── image recognition tools: 12\n",
-      "  ├── jsonld file access: 1\n",
-      "  ├── video parsing: 1\n",
-      "  ├── python compiler: 1\n",
-      "  ├── video recognition tools: 3\n",
-      "  ├── pdf viewer: 7\n",
-      "  ├── microsoft excel / google sheets: 3\n",
-      "  ├── word document access: 1\n",
-      "  ├── tool to extract text from images: 1\n",
-      "  ├── a word reversal tool / script: 1\n",
-      "  ├── counter: 1\n",
-      "  ├── excel: 3\n",
-      "  ├── image recognition: 5\n",
-      "  ├── color recognition: 3\n",
-      "  ├── excel file access: 3\n",
-      "  ├── xml file access: 1\n",
-      "  ├── access to the internet archive, web.archive.org: 1\n",
-      "  ├── text processing/diff tool: 1\n",
-      "  ├── gif parsing tools: 1\n",
-      "  ├── a web browser: 7\n",
-      "  ├── a search engine: 7\n",
-      "  ├── a speech-to-text tool: 2\n",
-      "  ├── code/data analysis tools: 1\n",
-      "  ├── audio capability: 2\n",
-      "  ├── pdf reader: 1\n",
-      "  ├── markdown: 1\n",
-      "  ├── a calculator: 5\n",
-      "  ├── access to wikipedia: 3\n",
-      "  ├── image recognition/ocr: 3\n",
-      "  ├── google translate access: 1\n",
-      "  ├── ocr: 4\n",
-      "  ├── bass note data: 1\n",
-      "  ├── text editor: 1\n",
-      "  ├── xlsx file access: 1\n",
-      "  ├── powerpoint viewer: 1\n",
-      "  ├── csv file access: 1\n",
-      "  ├── calculator (or use excel): 1\n",
-      "  ├── computer algebra system: 1\n",
-      "  ├── video processing software: 1\n",
-      "  ├── audio processing software: 1\n",
-      "  ├── computer vision: 1\n",
-      "  ├── google maps: 1\n",
-      "  ├── access to excel files: 1\n",
-      "  ├── calculator (or ability to count): 1\n",
-      "  ├── a file interface: 3\n",
-      "  ├── a python ide: 1\n",
-      "  ├── spreadsheet editor: 1\n",
-      "  ├── tools required: 1\n",
-      "  ├── b browser: 1\n",
-      "  ├── image recognition and processing tools: 1\n",
-      "  ├── computer vision or ocr: 1\n",
-      "  ├── c++ compiler: 1\n",
-      "  ├── access to google maps: 1\n",
-      "  ├── youtube player: 1\n",
-      "  ├── natural language processor: 1\n",
-      "  ├── graph interaction tools: 1\n",
-      "  ├── bablyonian cuniform -> arabic legend: 1\n",
-      "  ├── access to youtube: 1\n",
-      "  ├── image search tools: 1\n",
-      "  ├── calculator or counting function: 1\n",
-      "  ├── a speech-to-text audio processing tool: 1\n",
-      "  ├── access to academic journal websites: 1\n",
-      "  ├── pdf reader/extracter: 1\n",
-      "  ├── rubik's cube model: 1\n",
-      "  ├── wikipedia: 1\n",
-      "  ├── video capability: 1\n",
-      "  ├── image processing tools: 1\n",
-      "  ├── age recognition software: 1\n",
-      "  ├── youtube: 1\n"
-     ]
-    }
-   ],
-   "source": [
-    "# list of the tools used in all the samples\n",
-    "from collections import Counter, OrderedDict\n",
-    "\n",
-    "tools = []\n",
-    "for sample in json_QA:\n",
-    "    for tool in sample['Annotator Metadata']['Tools'].split('\\n'):\n",
-    "        tool = tool[2:].strip().lower()\n",
-    "        if tool.startswith(\"(\"):\n",
-    "            tool = tool[11:].strip()\n",
-    "        tools.append(tool)\n",
-    "tools_counter = OrderedDict(Counter(tools))\n",
-    "print(\"List of tools used in all samples:\")\n",
-    "print(\"Total number of tools used:\", len(tools_counter))\n",
-    "for tool, count in tools_counter.items():\n",
-    "    print(f\"  ├── {tool}: {count}\")"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "id": "5efee12a",
-   "metadata": {},
-   "source": [
-    "#### Graph"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 55,
-   "id": "7fe573cc",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "system_prompt = \"\"\"\n",
-    "You are a helpful assistant tasked with answering questions using a set of tools.\n",
-    "If the tool is not available, you can try to find the information online. You can also use your own knowledge to answer the question. \n",
-    "You need to provide a step-by-step explanation of how you arrived at the answer.\n",
-    "==========================\n",
-    "Here is a few examples showing you how to answer the question step by step.\n",
-    "\"\"\"\n",
-    "for i, samples in enumerate(random_samples):\n",
-    "    system_prompt += f\"\\nQuestion {i+1}: {samples['Question']}\\nSteps:\\n{samples['Annotator Metadata']['Steps']}\\nTools:\\n{samples['Annotator Metadata']['Tools']}\\nFinal Answer: {samples['Final answer']}\\n\"\n",
-    "system_prompt += \"\\n==========================\\n\"\n",
-    "system_prompt += \"Now, please answer the following question step by step.\\n\"\n",
-    "\n",
-    "# save the system_prompt to a file\n",
-    "with open('system_prompt.txt', 'w') as f:\n",
-    "    f.write(system_prompt)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 56,
-   "id": "d6beb0da",
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "\n",
-      "You are a helpful assistant tasked with answering questions using a set of tools.\n",
-      "If the tool is not available, you can try to find the information online. You can also use your own knowledge to answer the question. \n",
-      "You need to provide a step-by-step explanation of how you arrived at the answer.\n",
-      "==========================\n",
-      "Here is a few examples showing you how to answer the question step by step.\n",
-      "\n",
-      "Question 1: In terms of geographical distance between capital cities, which 2 countries are the furthest from each other within the ASEAN bloc according to wikipedia? Answer using a comma separated list, ordering the countries by alphabetical order.\n",
-      "Steps:\n",
-      "1. Search the web for \"ASEAN bloc\".\n",
-      "2. Click the Wikipedia result for the ASEAN Free Trade Area.\n",
-      "3. Scroll down to find the list of member states.\n",
-      "4. Click into the Wikipedia pages for each member state, and note its capital.\n",
-      "5. Search the web for the distance between the first two capitals. The results give travel distance, not geographic distance, which might affect the answer.\n",
-      "6. Thinking it might be faster to judge the distance by looking at a map, search the web for \"ASEAN bloc\" and click into the images tab.\n",
-      "7. View a map of the member countries. Since they're clustered together in an arrangement that's not very linear, it's difficult to judge distances by eye.\n",
-      "8. Return to the Wikipedia page for each country. Click the GPS coordinates for each capital to get the coordinates in decimal notation.\n",
-      "9. Place all these coordinates into a spreadsheet.\n",
-      "10. Write formulas to calculate the distance between each capital.\n",
-      "11. Write formula to get the largest distance value in the spreadsheet.\n",
-      "12. Note which two capitals that value corresponds to: Jakarta and Naypyidaw.\n",
-      "13. Return to the Wikipedia pages to see which countries those respective capitals belong to: Indonesia, Myanmar.\n",
-      "Tools:\n",
-      "1. Search engine\n",
-      "2. Web browser\n",
-      "3. Microsoft Excel / Google Sheets\n",
-      "Final Answer: Indonesia, Myanmar\n",
-      "\n",
-      "Question 2: Review the chess position provided in the image. It is black's turn. Provide the correct next move for black which guarantees a win. Please provide your response in algebraic notation.\n",
-      "Steps:\n",
-      "Step 1: Evaluate the position of the pieces in the chess position\n",
-      "Step 2: Report the best move available for black: \"Rd5\"\n",
-      "Tools:\n",
-      "1. Image recognition tools\n",
-      "Final Answer: Rd5\n",
-      "\n",
-      "==========================\n",
-      "Now, please answer the following question step by step.\n",
-      "\n"
-     ]
-    }
-   ],
-   "source": [
-    "# load the system prompt from the file\n",
-    "with open('system_prompt.txt', 'r') as f:\n",
-    "    system_prompt = f.read()\n",
-    "print(system_prompt)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "42fde0f8",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "import dotenv\n",
-    "from langgraph.graph import MessagesState, START, StateGraph\n",
-    "from langgraph.prebuilt import tools_condition\n",
-    "from langgraph.prebuilt import ToolNode\n",
-    "from langchain_google_genai import ChatGoogleGenerativeAI\n",
-    "from langchain_huggingface import HuggingFaceEmbeddings\n",
-    "from langchain_community.tools.tavily_search import TavilySearchResults\n",
-    "from langchain_community.document_loaders import WikipediaLoader\n",
-    "from langchain_community.document_loaders import ArxivLoader\n",
-    "from langchain_community.vectorstores import SupabaseVectorStore\n",
-    "from langchain.tools.retriever import create_retriever_tool\n",
-    "from langchain_core.messages import HumanMessage, SystemMessage\n",
-    "from langchain_core.tools import tool\n",
-    "from supabase.client import Client, create_client\n",
-    "\n",
-    "# Define the retriever from supabase\n",
-    "load_dotenv()\n",
-    "embeddings = HuggingFaceEmbeddings(model_name=\"sentence-transformers/all-mpnet-base-v2\") #  dim=768\n",
-    "\n",
-    "supabase_url = os.environ.get(\"SUPABASE_URL\")\n",
-    "supabase_key = os.environ.get(\"SUPABASE_SERVICE_KEY\")\n",
-    "supabase: Client = create_client(supabase_url, supabase_key)\n",
-    "vector_store = SupabaseVectorStore(\n",
-    "    client=supabase,\n",
-    "    embedding= embeddings,\n",
-    "    table_name=\"documents\",\n",
-    "    query_name=\"match_documents_langchain\",\n",
-    ")\n",
-    "\n",
-    "question_retrieve_tool = create_retriever_tool(\n",
-    "    vector_store.as_retriever(),\n",
-    "    \"Question Retriever\",\n",
-    "    \"Find similar questions in the vector database for the given question.\",\n",
-    ")\n",
-    "\n",
-    "@tool\n",
-    "def multiply(a: int, b: int) -> int:\n",
-    "    \"\"\"Multiply two numbers.\n",
-    "\n",
-    "    Args:\n",
-    "        a: first int\n",
-    "        b: second int\n",
-    "    \"\"\"\n",
-    "    return a * b\n",
-    "\n",
-    "@tool\n",
-    "def add(a: int, b: int) -> int:\n",
-    "    \"\"\"Add two numbers.\n",
-    "    \n",
-    "    Args:\n",
-    "        a: first int\n",
-    "        b: second int\n",
-    "    \"\"\"\n",
-    "    return a + b\n",
-    "\n",
-    "@tool\n",
-    "def subtract(a: int, b: int) -> int:\n",
-    "    \"\"\"Subtract two numbers.\n",
-    "    \n",
-    "    Args:\n",
-    "        a: first int\n",
-    "        b: second int\n",
-    "    \"\"\"\n",
-    "    return a - b\n",
-    "\n",
-    "@tool\n",
-    "def divide(a: int, b: int) -> int:\n",
-    "    \"\"\"Divide two numbers.\n",
-    "    \n",
-    "    Args:\n",
-    "        a: first int\n",
-    "        b: second int\n",
-    "    \"\"\"\n",
-    "    if b == 0:\n",
-    "        raise ValueError(\"Cannot divide by zero.\")\n",
-    "    return a / b\n",
-    "\n",
-    "@tool\n",
-    "def modulus(a: int, b: int) -> int:\n",
-    "    \"\"\"Get the modulus of two numbers.\n",
-    "    \n",
-    "    Args:\n",
-    "        a: first int\n",
-    "        b: second int\n",
-    "    \"\"\"\n",
-    "    return a % b\n",
-    "\n",
-    "@tool\n",
-    "def wiki_search(query: str) -> str:\n",
-    "    \"\"\"Search Wikipedia for a query and return maximum 2 results.\n",
-    "    \n",
-    "    Args:\n",
-    "        query: The search query.\"\"\"\n",
-    "    search_docs = WikipediaLoader(query=query, load_max_docs=2).load()\n",
-    "    formatted_search_docs = \"\\n\\n---\\n\\n\".join(\n",
-    "        [\n",
-    "            f'<Document source=\"{doc.metadata[\"source\"]}\" page=\"{doc.metadata.get(\"page\", \"\")}\"/>\\n{doc.page_content}\\n</Document>'\n",
-    "            for doc in search_docs\n",
-    "        ])\n",
-    "    return {\"wiki_results\": formatted_search_docs}\n",
-    "\n",
-    "@tool\n",
-    "def web_search(query: str) -> str:\n",
-    "    \"\"\"Search Tavily for a query and return maximum 3 results.\n",
-    "    \n",
-    "    Args:\n",
-    "        query: The search query.\"\"\"\n",
-    "    search_docs = TavilySearchResults(max_results=3).invoke(query=query)\n",
-    "    formatted_search_docs = \"\\n\\n---\\n\\n\".join(\n",
-    "        [\n",
-    "            f'<Document source=\"{doc.metadata[\"source\"]}\" page=\"{doc.metadata.get(\"page\", \"\")}\"/>\\n{doc.page_content}\\n</Document>'\n",
-    "            for doc in search_docs\n",
-    "        ])\n",
-    "    return {\"web_results\": formatted_search_docs}\n",
-    "\n",
-    "@tool\n",
-    "def arvix_search(query: str) -> str:\n",
-    "    \"\"\"Search Arxiv for a query and return maximum 3 result.\n",
-    "    \n",
-    "    Args:\n",
-    "        query: The search query.\"\"\"\n",
-    "    search_docs = ArxivLoader(query=query, load_max_docs=3).load()\n",
-    "    formatted_search_docs = \"\\n\\n---\\n\\n\".join(\n",
-    "        [\n",
-    "            f'<Document source=\"{doc.metadata[\"source\"]}\" page=\"{doc.metadata.get(\"page\", \"\")}\"/>\\n{doc.page_content[:1000]}\\n</Document>'\n",
-    "            for doc in search_docs\n",
-    "        ])\n",
-    "    return {\"arvix_results\": formatted_search_docs}\n",
-    "\n",
-    "@tool\n",
-    "def similar_question_search(question: str) -> str:\n",
-    "    \"\"\"Search the vector database for similar questions and return the first results.\n",
-    "    \n",
-    "    Args:\n",
-    "        question: the question human provided.\"\"\"\n",
-    "    matched_docs = vector_store.similarity_search(query, 3)\n",
-    "    formatted_search_docs = \"\\n\\n---\\n\\n\".join(\n",
-    "        [\n",
-    "            f'<Document source=\"{doc.metadata[\"source\"]}\" page=\"{doc.metadata.get(\"page\", \"\")}\"/>\\n{doc.page_content[:1000]}\\n</Document>'\n",
-    "            for doc in matched_docs\n",
-    "        ])\n",
-    "    return {\"similar_questions\": formatted_search_docs}\n",
-    "\n",
-    "tools = [\n",
-    "    multiply,\n",
-    "    add,\n",
-    "    subtract,\n",
-    "    divide,\n",
-    "    modulus,\n",
-    "    wiki_search,\n",
-    "    web_search,\n",
-    "    arvix_search,\n",
-    "    question_retrieve_tool\n",
-    "]\n",
-    "\n",
-    "llm = ChatGoogleGenerativeAI(model=\"gemini-2.0-flash\")\n",
-    "llm_with_tools = llm.bind_tools(tools)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "7dd0716c",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# load the system prompt from the file\n",
-    "with open('system_prompt.txt', 'r') as f:\n",
-    "    system_prompt = f.read()\n",
-    "\n",
-    "\n",
-    "# System message\n",
-    "sys_msg = SystemMessage(content=system_prompt)\n",
-    "\n",
-    "# Node\n",
-    "def assistant(state: MessagesState):\n",
-    "    \"\"\"Assistant node\"\"\"\n",
-    "    return {\"messages\": [llm_with_tools.invoke([sys_msg] + state[\"messages\"])]}\n",
-    "\n",
-    "# Build graph\n",
-    "builder = StateGraph(MessagesState)\n",
-    "builder.add_node(\"assistant\", assistant)\n",
-    "builder.add_node(\"tools\", ToolNode(tools))\n",
-    "builder.add_edge(START, \"assistant\")\n",
-    "builder.add_conditional_edges(\n",
-    "    \"assistant\",\n",
-    "    # If the latest message (result) from assistant is a tool call -> tools_condition routes to tools\n",
-    "    # If the latest message (result) from assistant is a not a tool call -> tools_condition routes to END\n",
-    "    tools_condition,\n",
-    ")\n",
-    "builder.add_edge(\"tools\", \"assistant\")\n",
-    "\n",
-    "# Compile graph\n",
-    "graph = builder.compile()\n"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 49,
-   "id": "f4e77216",
-   "metadata": {},
-   "outputs": [
-    {
-     "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<IPython.core.display.Image object>"
-      ]
-     },
-     "metadata": {},
-     "output_type": "display_data"
-    }
-   ],
-   "source": [
-    "from IPython.display import Image, display\n",
-    "\n",
-    "display(Image(graph.get_graph(xray=True).draw_mermaid_png()))"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "5987d58c",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "question = \"\"\n",
-    "messages = [HumanMessage(content=question)]\n",
-    "messages = graph.invoke({\"messages\": messages})"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "330cbf17",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "for m in messages['messages']:\n",
-    "    m.pretty_print()"
-   ]
-  }
- ],
- "metadata": {
-  "kernelspec": {
-   "display_name": "aiagent",
-   "language": "python",
-   "name": "python3"
-  },
-  "language_info": {
-   "codemirror_mode": {
-    "name": "ipython",
-    "version": 3
-   },
-   "file_extension": ".py",
-   "mimetype": "text/x-python",
-   "name": "python",
-   "nbconvert_exporter": "python",
-   "pygments_lexer": "ipython3",
-   "version": "3.12.9"
-  }
- },
- "nbformat": 4,
- "nbformat_minor": 5
-}