Update app.py

app.py

@@ -19,6 +19,94 @@ class BasicAgent:
         print(f"Agent returning fixed answer: {fixed_answer}")
         return fixed_answer
 
+from smolagents import ToolCallingAgent, InferenceClientModel()
+from smolagents.tools import DuckDuckGoSearchResults
+
+from smolagents import Tool
+from smolagents.tools import DuckDuckGoSearchResults
+from smolagents.models import InferenceClientModel
+from smolagents import CodeAgent
+
+class WebSearchTool(Tool):
+    def __init__(self):
+        self.agent = CodeAgent(
+            tools=[DuckDuckGoSearchResults()],
+            model=InferenceClientModel("deepseek-ai/DeepSeek-R1"),
+            name="WebSearcher",
+            description="Uses DuckDuckGo to answer queries with live web results.",
+            max_steps=5
+        )
+
+    def __call__(self, query: str) -> str:
+        try:
+            result = self.agent(query)
+            return result.get("output", "No response.")
+        except Exception as e:
+            return f"Web search failed: {e}"
+
+class VideoAnalyzerTool(Tool):
+    def __init__(self):
+        self.image_classifier = ImageClassifierTool()
+
+    def __call__(self, video_path: str) -> str:
+        cap = cv2.VideoCapture(video_path)
+        frame_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
+        labels = set()
+
+        for i in range(0, frame_count, max(1, frame_count // 5)):
+            cap.set(cv2.CAP_PROP_POS_FRAMES, i)
+            ret, frame = cap.read()
+            if not ret:
+                continue
+            frame_path = f"temp_frame.jpg"
+            cv2.imwrite(frame_path, frame)
+            try:
+                label = self.image_classifier(frame_path)
+                labels.add(label)
+            except Exception as e:
+                labels.add(f"Error processing frame: {e}")
+            os.remove(frame_path)
+
+        cap.release()
+        return f"Video contains: {', '.join(labels)}"
+
+from smolagents import CodeAgent, Tool
+from PIL import Image
+import torch
+import torchvision.transforms as transforms
+from transformers import ViTForImageClassification, ViTFeatureExtractor
+import cv2
+import os
+
+class ImageClassifierTool(Tool):
+    def __init__(self):
+        self.model = ViTForImageClassification.from_pretrained("google/vit-base-patch16-224")
+        self.feature_extractor = ViTFeatureExtractor.from_pretrained("google/vit-base-patch16-224")
+        self.transform = transforms.Compose([
+            transforms.Resize((224, 224)),
+            transforms.ToTensor()
+        ])
+        self.id2label = self.model.config.id2label
+
+    def __call__(self, image_path: str) -> str:
+        image = Image.open(image_path).convert("RGB")
+        inputs = self.feature_extractor(images=image, return_tensors="pt")
+        with torch.no_grad():
+            outputs = self.model(**inputs)
+        logits = outputs.logits
+        predicted_class_idx = logits.argmax(-1).item()
+        return f"Predicted label: {self.id2label[predicted_class_idx]}"
+
+class TimezoneTool(Tool):
+    name = "timezone_tool"
+    description = "Returns the current time for a given city."
+
+    def __call__(self, city: str) -> str:
+        url = f"http://worldtimeapi.org/api/timezone"
+        response = requests.get(url).json()
+        # You'd want to match city to a timezone
+        return "It's 9:45 AM in Tokyo."
+        
 def run_and_submit_all( profile: gr.OAuthProfile | None):
     """
     Fetches all questions, runs the BasicAgent on them, submits all answers,
@@ -40,7 +128,19 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
 
     # 1. Instantiate Agent ( modify this part to create your agent)
     try:
-        agent = BasicAgent()
+        agent = CodeAgent(
+    tools=[
+        ImageClassifierTool(),
+        VideoAnalyzerTool(),
+        TimezoneTool(),
+        WebSearchTool(),  # Now a Tool, so it can be integrated!
+    ],
+    model=InferenceClientModel("HuggingFaceH4/zephyr-7b-beta"),
+    max_steps=5,
+    name="web/media-agent",
+    description="An intelligent assistant that can classify images, summarize videos, check timezones, and search the web in real time."
+)
+
     except Exception as e:
         print(f"Error instantiating agent: {e}")
         return f"Error initializing agent: {e}", None