Refactoring

requirements.txt

@@ -4,4 +4,7 @@ requests~=2.32.3
 smolagents~=1.13.0
 python-dotenv~=1.1.0
 pandas~=2.2.3
-litellm~=1.66.1
+openpyxl~=3.1.5
+litellm~=1.66.1
+easyocr~=1.7.2
+wikipedia-api

src/agent/base_agent.py

@@ -1,24 +1,64 @@
-import os
-from smolagents import CodeAgent, LiteLLMModel, DuckDuckGoSearchTool
+from abc import abstractmethod, ABC
+
+from smolagents import CodeAgent, Tool, DuckDuckGoSearchTool, WikipediaSearchTool
+
+from tools.extract_text_from_image import ExtractTextFromImage
 from tools.weater_info_tool import WeatherInfoTool
 
 
-class BasicAgent:
-    def __init__(self):
-        model = LiteLLMModel(
-            model_id="gemini/gemini-2.0-flash-lite",
-            api_key=os.getenv("GEMINI_API_KEY")
-        )
-        self.agent = CodeAgent(
-            tools=[WeatherInfoTool(), DuckDuckGoSearchTool()],
-            model=model,
-            add_base_tools=True,
-            planning_interval=3
-        )
-        print("Agent initialized.")
+class BaseAgent(ABC):
+    def __init__(self, model_name: str, tools: list[Tool] | None = None, planning_interval: int = 3, max_steps: int = 10, use_all_custom_tools: bool = True):
+        self.model_name: str = model_name
+        self.planning_interval = planning_interval
+        self.max_steps = max_steps
+        self.use_all_custom_tools = use_all_custom_tools
+        self.tools: list[Tool] = self.init_tools(tools)
+        self.agent = self.init_agent()
 
     def __call__(self, question: str) -> str:
         print(f"Agent received question (first 50 chars): {question[:50]}...")
         fixed_answer = self.agent.run(question)
         print(f"Agent returning fixed answer: {fixed_answer}")
-        return fixed_answer
+        return fixed_answer
+
+    @abstractmethod
+    def get_model(self):
+        pass
+
+    def init_tools(self, tools: list[Tool] | None = None):
+        if tools is None:
+            tools = []
+        if self.use_all_custom_tools:
+            tools = [
+                ExtractTextFromImage(),
+                WeatherInfoTool(),
+                DuckDuckGoSearchTool(),
+                WikipediaSearchTool()
+            ]
+        return tools
+
+    def add_tool(self, tool: Tool):
+        self.tools.append(tool)
+
+    def init_agent(self):
+        agent = CodeAgent(
+            model=self.get_model(),
+            tools=[t for t in self.tools],
+            add_base_tools=True,
+            verbosity_level=1,
+            additional_authorized_imports=[
+                "pandas",
+                "numpy",
+                "datetime",
+                "json",
+                "re",
+                "math",
+                "os",
+                "requests",
+                "csv",
+                "urllib"],
+            planning_interval=self.planning_interval,
+            max_steps=self.max_steps
+        )
+        print("Agent initialized.")
+        return agent

src/agent/google_vertex_ai_agent.py

@@ -0,0 +1,19 @@
+import os
+from smolagents import Tool, LiteLLMModel
+from agent.base_agent import BaseAgent
+
+
+class OpenAiAgent(BaseAgent):
+    def __init__(self, model_name: str = "gemini/gemini-2.0-flash-lite", tools: list[Tool] | None = None,
+                 use_all_custom_tools: bool = True):
+        super().__init__(model_name=model_name, tools=tools, use_all_custom_tools=use_all_custom_tools)
+        self.model_name: str = model_name
+        self.agent = self.init_agent()
+
+    def get_model(self):
+        model = LiteLLMModel(
+            model_id=self.model_name,
+            temperature=0.2,
+            api_key=os.getenv("GEMINI_API_KEY")
+        )
+        return model

src/agent/open_ai_agent.py

@@ -0,0 +1,18 @@
+import os
+from smolagents import OpenAIServerModel, Tool
+from agent.base_agent import BaseAgent
+
+
+class OpenAiAgent(BaseAgent):
+    def __init__(self, model_name: str = "gpt-4o-mini", tools: list[Tool] | None = None, use_all_custom_tools: bool = True):
+        super().__init__(model_name=model_name, tools=tools, use_all_custom_tools=use_all_custom_tools)
+        self.model_name: str = model_name
+        self.agent = self.init_agent()
+
+    def get_model(self):
+        model = OpenAIServerModel(
+            model_id=self.model_name,
+            temperature=0.2,
+            api_key=os.getenv("OPENAI_API_KEY")
+        )
+        return model

src/core/evaluator.py

@@ -1,8 +1,8 @@
 import os
 import pandas as pd
 
+from agent.open_ai_agent import OpenAiAgent
 from rest_clients.hs_evaluator_client import HsEvaluatorClient
-from src.agent.base_agent import BasicAgent
 
 
 class Evaluator:
@@ -10,8 +10,8 @@ class Evaluator:
     def __init__(self, profile):
         self.profile = profile
         self.username = profile.username if profile else None
-        self.space_id = os.getenv("SPACE_ID")
-        self.agent = BasicAgent()
+        self.space_id = os.getenv("SPACE_ID", "tommaso1288/Final_Assignment_Template")
+        self.agent = OpenAiAgent()
         self.hs_evaluator_client: HsEvaluatorClient | None = None
 
     def run_and_submit(self):
@@ -31,16 +31,39 @@ class Evaluator:
     def _run_agent(self, questions):
         results_log = []
         answers_payload = []
+        print(f"Running agent on {len(questions)} questions...")
         for item in questions:
             task_id = item.get("task_id")
             question_text = item.get("question")
+
+            # ----------fetch any attached file ----------
+            try:
+                file_path = self.get_hs_evaluator_client().download_file_if_any(task_id)
+            except Exception as e:
+                file_path = None
+                print(f"[file fetch error] {task_id}: {e}")
+
+            # ---------- Build the prompt sent to the agent ----------
+            if file_path:
+                q_for_agent = (
+                    f"{question_text}\n\n"
+                    f"---\n"
+                    f"A file was downloaded for this task and saved locally at:\n"
+                    f"{file_path}\n"
+                    f"---\n\n"
+                )
+            else:
+                q_for_agent = question_text
+
             if not task_id or question_text is None:
+                print(f"Skipping item with missing task_id or question: {item}")
                 continue
             try:
-                submitted_answer = self.agent(question_text)
+                submitted_answer = self.agent(q_for_agent)
                 answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
                 results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
             except Exception as e:
+                print(f"Error running agent on task {task_id}: {e}")
                 results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
         return results_log, answers_payload
 

src/rest_clients/hs_evaluator_client.py

@@ -1,3 +1,7 @@
+import re
+import tempfile
+from pathlib import Path
+
 import pandas as pd
 import requests
 
@@ -18,6 +22,32 @@ class HsEvaluatorClient:
             print(f"Error fetching questions: {e}")
             return None
 
+    def download_file_if_any(self, task_id: str) -> str | None:
+        url = f"{self.base_url}/files/{task_id}"
+        try:
+            resp = requests.get(url, timeout=30)
+            if resp.status_code == 404:
+                return None
+            resp.raise_for_status()
+        except requests.exceptions.HTTPError as e:
+            raise e
+
+        # ▸ Save bytes to a named file inside the system temp dir
+        #    Try to keep original extension from Content-Disposition if present.
+        cdisp = resp.headers.get("content-disposition", "")
+        filename = task_id  # default base name
+        if "filename=" in cdisp:
+            m = re.search(r'filename="([^"]+)"', cdisp)
+            if m:
+                filename = m.group(1)  # keep provided name
+
+        tmp_dir = Path(tempfile.gettempdir()) / "gaia_files"
+        tmp_dir.mkdir(exist_ok=True)
+        file_path = tmp_dir / filename
+        with open(file_path, "wb") as f:
+            f.write(resp.content)
+        return str(file_path)
+
     def submit_answers(self, answers_payload, results_log):
         agent_code = f"https://huggingface.co/spaces/{self.space_id}/tree/main"
         submission_data = {

src/tools/extract_text_from_image.py

@@ -0,0 +1,43 @@
+import os
+import easyocr
+from smolagents import Tool
+
+
+class ExtractTextFromImage(Tool):
+    name = "extract_text_from_image"
+    description = "A tool for extracting text from an image using the EasyOCR library."
+    inputs = {
+        "image_path": {
+            "type": "string",
+            "description": "The file path to the image to be processed."
+        }
+    }
+    output_type = "string"
+
+    def forward(self, image_path: str) -> str:
+        """
+        Extract text from an image file using EasyOCR.
+
+        Args:
+            image_path (str): The path to the image file to be processed.
+
+        Returns:
+            str: The extracted text from the image or an error message.
+        """
+        try:
+            if not os.path.exists(image_path):
+                return f"Error: File '{image_path}' does not exist."
+
+            reader = easyocr.Reader(['en'], gpu=False)  # Use GPU=True for faster execution if available
+
+            results = reader.readtext(image_path, detail=1)
+
+            if not results:
+                return "No text detected in the image."
+            extracted_texts = [result[1] for result in results]  # Extract the text field from results
+            extracted_text = "\n".join(extracted_texts)
+            return f"Extracted text from image:\n\n{extracted_text}"
+        except ImportError:
+            return "Error: easyocr is not installed. Please install it with 'pip install easyocr'."
+        except Exception as e:
+            return f"Error extracting text from image: {str(e)}"