Update app.py

app.py

@@ -3,37 +3,142 @@ import gradio as gr
 import requests
 import inspect
 import pandas as pd
-from transformers import pipeline
-import numpy as np
-
+from smolagents import CodeAgent, DuckDuckGoSearchTool, OpenAIServerModel, VisitWebpageTool, Tool, HfApiModel, ToolCallingAgent
+import io
 # (Keep Constants as is)
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
+class AttachmentDownloadTool(Tool):
+    name = "attachment-downloader"
+    description = "Downloads the file associated with the given task_id. If it does not exist, return None. input: task_id。output: attachment files or None"
+    inputs = {
+        "task_id": {
+            "type": "str",
+            "description": "task_id that needs to download attachment files."
+        }
+    }
+    output_type = io.BytesIO
+
+    def forward(self, task_id):
+        download_url = f"{api_url}/files/"
+        
+        try:
+            response = requests.get(download_url + task_id, stream=True, timeout=15)
+            if response.status_code != 200:
+                return None
+            file_obj = io.BytesIO(response.content)
+            file_obj.seek(0)
+            return file_obj
+        except Exception as e:
+            return None
+
+class ImageCaptionTool(Tool):
+    name = "image-captioner"
+    description = "Identify the content of the input image and describe it in natural language. Input: image. Output: description text."
+    inputs = {
+        "image": {
+            "type": "image",
+            "description": "Images that need to be identified and described"
+        }
+    }
+    output_type = str
+
+    def setup(self):
+        self.model = OpenAIServerModel(
+            model_id="Qwen/Qwen2.5-VL-32B-Instruct",
+            api_base="https://api.siliconflow.cn/v1/",
+            api_key=os.getenv('MODEL_TOKEN'),
+        )
+
+    def forward(self, image):
+        prompt = "Please describe the content of this picture in detail."
+        return self.model(prompt, images=[image])
+
+class AudioToTextTool(Tool):
+    name = "audio-to-text"
+    description = "Convert the input audio content to text. Input: audio. Output: text."
+    inputs = {
+        "audio": {
+            "type": "audio",
+            "description": "The audio file that needs to be transcribed"
+        }
+    }
+    output_type = str
+
+    def setup(self):
+        # 使用 HuggingFace Hub 上的 Whisper 大模型
+        self.model = HfApiModel(model_id="openai/whisper-large-v3")  # 或其他支持音频转写的模型
+
+    def forward(self, audio):
+        prompt = "Please transcribe this audio content into text."
+        return self.model(prompt, audios=[audio])
+
 # --- Basic Agent Definition ---
 # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
+# class BasicAgent:
+#     def __init__(self):
+#         print("BasicAgent initialized.")
+#     def __call__(self, question: str) -> str:
+#         print(f"Agent received question (first 50 chars): {question[:50]}...")
+#         fixed_answer = "This is a default answer."
+#         print(f"Agent returning fixed answer: {fixed_answer}")
+#         return fixed_answer
 class BasicAgent:
     def __init__(self):
-        print("Initializing BERT-based QA agent...")
-        # Load the BERT model fine-tuned on the SQuAD dataset
-        self.qa_pipeline = pipeline("question-answering", model="bert-large-uncased-whole-word-masking-finetuned-squad")
-    
-    def __call__(self, question: str) -> str:
-        """
-        Process the question using the BERT model fine-tuned on SQuAD, and return an answer based on the context.
-        """
-        try:
-            context = f"answer the following question as briefly as possible{question}"
-            # Use the QA pipeline to get an answer based on the context and the question
-            result = self.qa_pipeline(question=question, context=context)
-            answer = result["answer"]
-        except Exception as e:
-            print(f"Error during QA: {e}")
-            answer = "Error processing question."
+
+        self.think_model = OpenAIServerModel(
+            model_id="THUDM/GLM-Z1-32B-0414",
+            api_base="https://api.siliconflow.cn/v1/",
+            api_key=os.getenv('MODEL_TOKEN'),
+        )
+        self.base_model = OpenAIServerModel(
+            model_id="THUDM/GLM-4-32B-0414",
+            api_base="https://api.siliconflow.cn/v1/",
+            api_key=os.getenv('MODEL_TOKEN'),
+        )
+        # self.vision_model = OpenAIServerModel(
+        #     model_id="Qwen/Qwen2.5-VL-32B-Instruct",
+        #     api_base="https://api.siliconflow.cn/v1/",
+        #     api_key=os.getenv('MODEL_TOKEN'),
+        # )
+
+        self.tools = [AttachmentDownloadTool, ImageCaptionTool, AudioToTextTool]
+
+        web_agent = ToolCallingAgent(
+            tools=[DuckDuckGoSearchTool(), VisitWebpageTool()],
+            model=self.base_model,
+            max_steps=10,
+            name="web_search_agent",
+            description="Runs web searches for you.",
+        )
         
-        return answer
+        self.agent = CodeAgent(
+            tools=self.tools,
+            model=self.think_model,
+            managed_agents=[web_agent,],
+            additional_authorized_imports=["time", "numpy", "pandas"],
+            max_steps=20
+        )
+        print("BasicAgent initialized.")
 
+    def __call__(self, question: str, images=None) -> str:
         
+        print(f"Agent received question (first 50 chars): {question[:50]}...")
+        
+        try:
+            if images is not None:
+                result = self.agent.run(question, images=images)
+            else:
+                result = self.agent.run(question)
+            
+            print(f"Agent returning answer: {result}")
+            return result
+        except Exception as e:
+            print(f"Agent error: {e}")
+            return f"AGENT ERROR: {e}"
+
+
 def run_and_submit_all( profile: gr.OAuthProfile | None):
     """
     Fetches all questions, runs the BasicAgent on them, submits all answers,
@@ -161,11 +266,9 @@ 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).