Update tools.py

tools.py

@@ -6,6 +6,7 @@ except ImportError:
     Tool = object
 
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
+HF_DATASET_BASE = "https://huggingface.co/datasets/gaia-benchmark/GAIA/resolve/main/2023/test"
 
 # Tool to download and read file attachments (text) from the scoring API
 # Plain class (no Tool inheritance) — called directly, not via CodeAgent
@@ -16,19 +17,24 @@ class FileReaderTool:
         self.api_url = api_url
         print("FileReaderTool initialized.")
 
-    def __call__(self, task_id: str) -> str:
-        file_url = f"{self.api_url}/files/{task_id}"
-        print(f"FileReaderTool downloading from: {file_url}")
-        try:
-            resp = requests.get(file_url, timeout=30)
-            resp.raise_for_status()
-            content = resp.text
-            print(f"FileReaderTool downloaded {len(content)} chars.")
-            return content
-        except requests.exceptions.RequestException as e:
-            error_msg = f"Failed to download file for task {task_id}: {e}"
-            print(error_msg)
-            return error_msg
+    def __call__(self, task_id: str, file_name: str = "") -> str:
+        # Try HF dataset URL first (actual file location), then scoring API fallback
+        urls_to_try = []
+        if file_name:
+            urls_to_try.append(f"{HF_DATASET_BASE}/{file_name}")
+        urls_to_try.append(f"{self.api_url}/files/{task_id}")
+
+        for url in urls_to_try:
+            print(f"FileReaderTool trying: {url}")
+            try:
+                resp = requests.get(url, timeout=30)
+                if resp.status_code == 200:
+                    content = resp.text
+                    print(f"FileReaderTool downloaded {len(content)} chars.")
+                    return content
+            except requests.exceptions.RequestException:
+                continue
+        return f"Failed to download file for task {task_id}"
 
 
 # Tool to download image attachments and return them as base64
@@ -40,21 +46,25 @@ class ImageReaderTool:
         self.api_url = api_url
         print("ImageReaderTool initialized.")
 
-    def __call__(self, task_id: str) -> str:
-        file_url = f"{self.api_url}/files/{task_id}"
-        print(f"ImageReaderTool downloading from: {file_url}")
-        try:
-            resp = requests.get(file_url, timeout=30)
-            resp.raise_for_status()
-            # Detect content type for the base64 data URI
-            content_type = resp.headers.get("Content-Type", "image/png")
-            image_b64 = base64.b64encode(resp.content).decode("utf-8")
-            print(f"ImageReaderTool downloaded image ({len(resp.content)} bytes, {content_type}).")
-            return f"data:{content_type};base64,{image_b64}"
-        except requests.exceptions.RequestException as e:
-            error_msg = f"Failed to download image for task {task_id}: {e}"
-            print(error_msg)
-            return error_msg
+    def __call__(self, task_id: str, file_name: str = "") -> str:
+        # Try HF dataset URL first (actual file location), then scoring API fallback
+        urls_to_try = []
+        if file_name:
+            urls_to_try.append(f"{HF_DATASET_BASE}/{file_name}")
+        urls_to_try.append(f"{self.api_url}/files/{task_id}")
+
+        for url in urls_to_try:
+            print(f"ImageReaderTool trying: {url}")
+            try:
+                resp = requests.get(url, timeout=30)
+                if resp.status_code == 200:
+                    content_type = resp.headers.get("Content-Type", "image/png")
+                    image_b64 = base64.b64encode(resp.content).decode("utf-8")
+                    print(f"ImageReaderTool downloaded image ({len(resp.content)} bytes, {content_type}).")
+                    return f"data:{content_type};base64,{image_b64}"
+            except requests.exceptions.RequestException:
+                continue
+        return f"Failed to download image for task {task_id}"
 
 
 #search web tool — uses ddgs directly to avoid smolagents DuckDuckGoSearchTool package check