wip: added tool to convert audio to text and improve fetch file and save to temp location

agent.py

@@ -11,7 +11,8 @@ model = {
     "grok": xAI(id="grok-3-beta"),
     "gpt": OpenAIChat(id="gpt-4.1-nano"),
     "gemini": Gemini(id="gemini-2.5-pro-preview-03-25"),
-    "open_router": OpenRouter(id="qwen/qwen3-235b-a22b")
+    # "open_router": OpenRouter(id="qwen/qwen3-235b-a22b")
+    "open_router": OpenRouter(id="google/gemini-2.5-pro-exp-03-25")
 }
 
 

requirements.txt

@@ -6,4 +6,5 @@ duckduckgo-search
 wikipedia
 google
 google-genai
-httpx
+httpx
+openai-whisper

tools.py

@@ -1,14 +1,13 @@
 import os
 
-import httpx
+import whisper
 from agno.tools.duckduckgo import DuckDuckGoTools
 from agno.tools.reasoning import ReasoningTools
+from utils import fetch_file_content
 from agno.tools.wikipedia import WikipediaTools
 
 BASE_STORAGE_ROOT = os.getenv("AGENT_STORAGE_ROOT", os.path.join(os.getcwd(), "agent_storage"))
 
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
 
 def get_file_from_task_id(task_id: str) -> str:
     """
@@ -18,28 +17,24 @@ def get_file_from_task_id(task_id: str) -> str:
         task_id (str): Identifier that points to the remote file.
 
     Returns:
-        str: task_id to be used by other tools to read the file
+        str: Path to the downloaded file.
     """
-    # ensure storage directory exists
     task_dir = os.path.join(BASE_STORAGE_ROOT, task_id)
     os.makedirs(task_dir, exist_ok=True)
 
-    # filename derived from task_id
     filename = task_id
     file_path = os.path.join(task_dir, filename)
 
-    # if file already exists, return
     if os.path.exists(file_path):
         print("[INFO] Using cached file:", file_path)
         return file_path
 
-    # fetch content from remote
-    response = httpx.get(f"{DEFAULT_API_URL}/files/{task_id}", timeout=15, follow_redirects=True)
-    response.raise_for_status()
+    # Use the utility function to fetch content
+    result = fetch_file_content(task_id)
+    content = result["content"]
 
-    # write content to file
     with open(file_path, "wb") as f:
-        f.write(response.content)
+        f.write(content)
 
     return file_path
 
@@ -60,10 +55,29 @@ def read_file_from_task_id(task_id: str) -> str:
         return f.read()
 
 
+def convert_audio_to_text(task_id: str) -> str:
+    """
+    Use this to download an audio and convert it to text
+
+     Args:
+        task_id (str): Identifier that points to the remote file.
+
+    Returns:
+        str: the transcript of the audio in text
+    """
+
+    result = fetch_file_content(task_id, temp=True)
+    model = whisper.load_model("turbo")
+    result_whisper = model.transcribe(audio=result["path"])
+    print("[convert_audio_to_text]", result_whisper["text"])
+    return result_whisper["text"]
+
+
 tools = [
     ReasoningTools(think=True, add_few_shot=True),
     DuckDuckGoTools(fixed_max_results=5),
     WikipediaTools(),
     get_file_from_task_id,
-    read_file_from_task_id
+    read_file_from_task_id,
+    convert_audio_to_text
 ]

utils.py

@@ -0,0 +1,32 @@
+import httpx
+import tempfile
+import mimetypes
+import os
+
+DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
+
+
+def fetch_file_content(task_id: str, temp: bool = False) -> dict:
+    """
+    Downloads file content for the given task_id.
+    Returns a dict with:
+      - content: bytes of the file
+      - path: filesystem path to a temp file if temp=True, else None
+    """
+    url = f"{DEFAULT_API_URL}/files/{task_id}"
+    resp = httpx.get(url, timeout=15, follow_redirects=True)
+    resp.raise_for_status()
+    content = resp.content
+
+    result = {"content": content, "path": None}
+
+    if temp:
+        ctype = resp.headers.get("content-type", "")
+        ext = mimetypes.guess_extension(ctype) or os.path.splitext(task_id)[1] or ""
+        tmp = tempfile.NamedTemporaryFile(delete=False, suffix=ext)
+        tmp.write(content)
+        tmp.close()
+        result["path"] = tmp.file.name
+
+    print("[fetch_file_content]", result)
+    return result