Update app.py

Added a FetchFileTool and an AudioDescriptionTool

app.py

@@ -30,6 +30,67 @@ model = LiteLLMModel(
 # Import tool from Hub
 image_generation_tool = load_tool("agents-course/text-to-image", trust_remote_code=True)
 
+class AudioDescriptionTool(Tool):
+    name = "audio_description"
+    description = "This is a tool that will describe an audio clip."
+    inputs = {
+        "file_name": {
+            "type": "string",
+            "description": "The name of the local file to describe, for example: /files/99c9cc74-fdc8-46c6-8f8d-3ce2d3bfeea3",
+        }
+    }
+    output_type = "string"
+    def __init__(self):
+        self.client = genai.Client(api_key=os.environ.get("GEMINI_KEY"))
+
+    def forward(self, file_name: str):
+        try:
+            mp3_file = self.client.files.upload(file=f"{file_name}")
+            audio_description = self.client.models.generate_content(
+                model="gemini-2.0-flash", contents=["Describe this audio clip", mp3_file]
+            )
+            return audio_description.text
+        except Exception as e:
+            print(f"Error getting audio description: {e}")
+            return False
+
+audio_description_tool = AudioDescriptionTool()
+class FetchFileTool(Tool):
+    name = "file_saver"
+    description = """
+    This is a tool that will fetch an internet file and save content locally.
+    It takes a location, file path, file name, and content to save."""
+    inputs = {
+        "location": {
+            "type": "string",
+            "description": "The internet location of the file to fetch",
+        },
+        "file_path": {
+            "type": "string",
+            "description": "The path to the directory where the file will be saved",
+        },
+        "file_name": {
+            "type": "string",
+            "description": "The name of the file where content will be saved",
+        },
+    }
+    output_type = "boolean"
+
+    def forward(self, location: str, file_path: str, file_name: str):
+        try:
+            response = requests.get(location)
+            response.raise_for_status()
+            content = response.content
+            full_path = os.path.join(file_path, file_name)
+            with open(full_path, 'wb') as out_file:
+                out_file.write(content)
+            return True
+        except Exception as e:
+            print(f"Error saving content to file: {e}")
+            return False
+
+fetch_file_tool = FetchFileTool()
+
 class WikipediaSearchTool(Tool):
     name = "wikipedia_search"
     description = "Fetches a summary of a Wikipedia page based on a given search query (only one word or group of words)."
@@ -85,7 +146,7 @@ class BasicAgent:
             prompt_templates = yaml.safe_load(stream) 
         self.agent = CodeAgent(
             model=model,
-            tools=[final_answer, search_tool, wiki_tool],
+            tools=[final_answer, search_tool, wiki_tool, fetch_file_tool, audio_description_tool],
             max_steps=7,
             verbosity_level=1,
             grammar=None,
@@ -159,25 +220,17 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
     answers_payload = []
     print(f"Running agent on {len(questions_data)} questions...")
     for item in questions_data:
-        if not item.get("file_name"):
-            continue
-            
         task_id = item.get("task_id")
-        question_text = item.get("question") + " think hard to answer."            
+        question_text = item.get("question") + " think hard to answer."
         if not task_id or question_text is None:
             print(f"Skipping item with missing task_id or question: {item}")
             continue
         try:
             # check if question_text contains an mp3 filename
-            if item.get("file_name") and ".mp3" in question_text:
-                client = genai.Client(api_key=os.environ.get("GEMINI_KEY"))
-                mp3_file = client.files.upload(file=f"/files/{item.get('task_id')}")
-                audio_description = client.models.generate_content(
-                    model="gemini-2.0-flash", contents=["Describe this audio clip", mp3_file]
-                )
-                question_text = f"{question_text} {audio_description.text}"
+            if not item.get("file_name") and ".mp3" in question_text:
+                continue
             # check if the file_name is not empty
-            elif item.get("file_name"):
+            if item.get("file_name"):
                 question_text = f"{question_text} Here is the file: https://agents-course-unit4-scoring.hf.space/files/{item.get('task_id')}"
             submitted_answer = agent(question_text)
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
@@ -192,7 +245,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
         print("Agent did not produce any answers to submit.")
         return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
 
-    return "Questions parsed.", pd.DataFrame(results_log)
+    # return "Questions parsed.", pd.DataFrame(results_log)
     
     # 4. Prepare Submission 
     # submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}