Update app.py

app.py

@@ -1,41 +1,191 @@
 import os
-#os.environ["TRANSFORMERS_NO_TF"] = "1"
-
 import gradio as gr
 import requests
-import inspect
+import openai
+from smolagents import OpenAIServerModel, DuckDuckGoSearchTool, CodeAgent, WikipediaSearchTool
+from pathlib import Path
+import tempfile
+from smolagents.tools import PipelineTool, Tool
+import pathlib
+from typing import Union, Optional
 import pandas as pd
-import time
-#from transformers import pipeline
-from smolagents import CodeAgent, DuckDuckGoSearchTool, HfApiModel
+from tabulate import tabulate  # pragma: no cover – fallback path
+import re
+
 # (Keep Constants as is)
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
+class SpeechToTextTool(PipelineTool):
+    """
+    Transcribes an audio file to text using the OpenAI Whisper API.
+    Only local file paths are supported.
+    """
+    default_checkpoint = "openai/whisper-1"   # purely informational here
+    description = (
+        "This tool sends an audio file to OpenAI Whisper and returns the "
+        "transcribed text."
+    )
+    name = "transcriber"
+    inputs = {
+        "audio": {
+            "type": "string",
+            "description": "Absolute or relative path to a local audio file.",
+        }
+    }
+    output_type = "string"
+
+    # ──────────────────────────────────────────────────────────────────────────
+    # Public interface
+    # ──────────────────────────────────────────────────────────────────────────
+    def __call__(self, audio: str) -> str:
+        """
+        Convenience wrapper so the tool can be used like a regular function:
+            text = SpeechToTextTool()(path_to_audio)
+        """
+        return self._transcribe(audio)
+
+    # ──────────────────────────────────────────────────────────────────────────
+    # Internal helpers
+    # ──────────────────────────────────────────────────────────────────────────
+    @staticmethod
+    def _transcribe(audio_path: str) -> str:
+        # ----- validation ----------------------------------------------------
+        if not isinstance(audio_path, str):
+            raise TypeError(
+                "Parameter 'audio' must be a string containing the file path."
+            )
+        path = Path(audio_path).expanduser().resolve()
+        if not path.is_file():
+            raise FileNotFoundError(f"No such audio file: {path}")
+
+        # ----- API call ------------------------------------------------------
+        with path.open("rb") as fp:
+            response = openai.audio.transcriptions.create(
+                file=fp,
+                model="whisper-1",        # currently the only Whisper model
+                response_format="text"    # returns plain text instead of JSON
+            )
+
+        # For response_format="text", `response` is already the raw transcript
+        return response
+
+class ExcelToTextTool(Tool):
+    """Render an Excel worksheet as Markdown text."""
+
+    # ------------------------------------------------------------------
+    # Required smol‑agents metadata
+    # ------------------------------------------------------------------
+    name = "excel_to_text"
+    description = (
+        "Read an Excel file and return a Markdown table of the requested sheet. "
+        "Accepts either the sheet name or the zero-based index."
+    )
+
+    inputs = {
+        "excel_path": {
+            "type": "string",
+            "description": "Path to the Excel file (.xlsx / .xls).",
+        },
+        "sheet_name": {
+            "type": "string",
+            "description": (
+                "Worksheet name or zero‑based index *as a string* (optional; default first sheet)."
+            ),
+            "nullable": True,
+        },
+    }
+
+    output_type = "string"
+
+    # ------------------------------------------------------------------
+    # Core logic
+    # ------------------------------------------------------------------
+    def forward(
+            self,
+            excel_path: str,
+            sheet_name: Optional[str] = None,
+    ) -> str:
+        """Load *excel_path* and return the sheet as a Markdown table."""
+
+        path = pathlib.Path(excel_path).expanduser().resolve()
+        if not path.exists():
+            return f"Error: Excel file not found at {path}"
+
+        try:
+            # Interpret sheet identifier -----------------------------------
+            sheet: Union[str, int]
+            if sheet_name is None or sheet_name == "":
+                sheet = 0  # first sheet
+            else:
+                # If the user passed a numeric string (e.g. "1"), cast to int
+                sheet = int(sheet_name) if sheet_name.isdigit() else sheet_name
+
+            # Load worksheet ----------------------------------------------
+            df = pd.read_excel(path, sheet_name=sheet)
+
+            # Render to Markdown; fall back to tabulate if needed ---------
+            if hasattr(pd.DataFrame, "to_markdown"):
+                return df.to_markdown(index=False)
+            from tabulate import tabulate  # pragma: no cover – fallback path
+
+            return tabulate(df, headers="keys", tablefmt="github", showindex=False)
+
+        except Exception as exc:  # broad catch keeps the agent chat‑friendly
+            return f"Error reading Excel file: {exc}"
+
+
+def download_file_if_any(base_api_url: str, task_id: str) -> str | None:
+    """
+    Try GET /files/{task_id}.
+    • On HTTP 200 → save to a temp dir and return local path.
+    • On 404 → return None.
+    • On other errors → raise so caller can log / handle.
+    """
+    url = f"{base_api_url}/files/{task_id}"
+    try:
+        resp = requests.get(url, timeout=30)
+        if resp.status_code == 404:
+            return None          # no file
+        resp.raise_for_status()   # raise on 4xx/5xx ≠ 404
+    except requests.exceptions.HTTPError as e:
+        # propagate non-404 errors (403, 500, …)
+        raise e
+
+    # ▸ Save bytes to a named file inside the system temp dir
+    #    Try to keep original extension from Content-Disposition if present.
+    cdisp = resp.headers.get("content-disposition", "")
+    filename = task_id                                 # default base name
+    if "filename=" in cdisp:
+        m = re.search(r'filename="([^"]+)"', cdisp)
+        if m:
+            filename = m.group(1)                      # keep provided name
+
+    tmp_dir = Path(tempfile.gettempdir()) / "gaia_files"
+    tmp_dir.mkdir(exist_ok=True)
+    file_path = tmp_dir / filename
+    with open(file_path, "wb") as f:
+        f.write(resp.content)
+    return str(file_path)
 
 # --- Basic Agent Definition ---
 # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
 class BasicAgent:
-    def __init__(self, HF_API_KEY):
-        print("BasicAgent initialized.")
-        # Initialize a small local model instead of API model
-        self.model = HfApiModel("text-generation", model="mistralai/Mistral-7B-v0.1")
-        self.HF_API_KEY = HF_API_KEY
-        # Initialize the search tool (you can keep DuckDuckGoSearchTool)
-        search_tool = DuckDuckGoSearchTool()
-
-        # Initialize Agent
+    def __init__(self):
         self.agent = CodeAgent(
-            model=self.model,
-            api_key = self.HF_API_KEY,
-            tools=[search_tool]
+            model=OpenAIServerModel(model_id="gpt-4o"),
+            tools=[DuckDuckGoSearchTool(), WikipediaSearchTool(), SpeechToTextTool(), ExcelToTextTool()],
+            add_base_tools=True,
+            additional_authorized_imports=['pandas','numpy','csv','subprocess']
         )
 
+        print("BasicAgent initialized.")
+
     def __call__(self, question: str) -> str:
         print(f"Agent received question (first 50 chars): {question[:50]}...")
-        generated_text = self.model(question, max_length=200, num_return_sequences=1)[0]['generated_text']
-        print(f"Agent returning fixed answer: {generated_text}")
-        return generated_text
+        fixed_answer = self.agent.run(question)
+        print(f"Agent returning answer: {fixed_answer}")
+        return fixed_answer
 
 def run_and_submit_all( profile: gr.OAuthProfile | None):
     """
@@ -43,7 +193,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
     and displays the results.
     """
     # --- Determine HF Space Runtime URL and Repo URL ---
-    space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code
+    space_id = "l3xv/Final_Assignment_Template"
 
     if profile:
         username= f"{profile.username}"
@@ -94,14 +244,33 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
     for item in questions_data:
         task_id = item.get("task_id")
         question_text = item.get("question")
+
+        # ----------fetch any attached file ----------
+        try:
+            file_path = download_file_if_any(api_url, task_id)
+        except Exception as e:
+            file_path = None
+            print(f"[file fetch error] {task_id}: {e}")
+
+        # ---------- Build the prompt sent to the agent ----------
+        if file_path:
+            q_for_agent = (
+                f"{question_text}\n\n"
+                f"---\n"
+                f"A file was downloaded for this task and saved locally at:\n"
+                f"{file_path}\n"
+                f"---\n\n"
+            )
+        else:
+            q_for_agent = question_text
+
         if not task_id or question_text is None:
             print(f"Skipping item with missing task_id or question: {item}")
             continue
         try:
-            submitted_answer = agent(question_text)
+            submitted_answer = agent(q_for_agent)
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
-            time.sleep(2)
         except Exception as e:
              print(f"Error running agent on task {task_id}: {e}")
              results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
@@ -176,7 +345,7 @@ with gr.Blocks() as demo:
     )
 
     gr.LoginButton()
-        
+
     run_button = gr.Button("Run Evaluation & Submit All Answers")
 
     status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
@@ -185,8 +354,6 @@ with gr.Blocks() as demo:
 
     run_button.click(
         fn=run_and_submit_all,
-
-
         outputs=[status_output, results_table]
     )
 
@@ -194,7 +361,7 @@ if __name__ == "__main__":
     print("\n" + "-"*30 + " App Starting " + "-"*30)
     # Check for SPACE_HOST and SPACE_ID at startup for information
     space_host_startup = os.getenv("SPACE_HOST")
-    space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
+    space_id_startup = "l3xv/Final_Assignment_Template"
 
     if space_host_startup:
         print(f"✅ SPACE_HOST found: {space_host_startup}")