Update app.py

app.py

@@ -8,6 +8,7 @@ import time
 import threading
 import requests
 import pandas as pd
+import pandas as pd
 import gradio as gr
 
 from langchain_core.messages import HumanMessage
@@ -30,7 +31,7 @@ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 # Single values — no per-difficulty branching
 TIMEOUT_SECONDS  = 60
 RECURSION_LIMIT  = 12
-PAUSE_SECONDS    = 1.5
+PAUSE_SECONDS    = 1.0
 
 # ==========================================================
 # LOGGING
@@ -104,28 +105,31 @@ def submit_answers(url: str, payload: dict) -> dict:
     return r.json()
 
 
-def fetch_task_file(task_id: str) -> str | None:
+def fetch_task_file(task_id: str) -> tuple[str | None, str | None]:
     """
-    Try to download the file attached to a task from GET /files/{task_id}.
-    Saves it to /tmp/<task_id>.<ext> and returns the local path.
-    Returns None if the task has no attachment or the download fails.
+    Download the file attached to a task from GET /files/{task_id}.
+    Returns (local_path, injected_text):
+      - local_path: saved file path (for binary files like .py the model runs directly)
+      - injected_text: pre-read text content to inject into the question (for Excel/CSV)
+    Returns (None, None) if no attachment exists.
     """
     url = f"{DEFAULT_API_URL}/files/{task_id}"
     try:
         r = requests.get(url, timeout=20)
         if r.status_code == 404:
-            return None
+            log("File     : no attachment (404)")
+            return None, None
         r.raise_for_status()
 
-        # Determine extension from Content-Disposition or Content-Type
+        # Determine extension
         ext = ""
         cd = r.headers.get("Content-Disposition", "")
         ct = r.headers.get("Content-Type", "")
 
         if "filename=" in cd:
-            fname = cd.split("filename=")[-1].strip().strip('"')
-            ext = os.path.splitext(fname)[-1]
-        elif "spreadsheet" in ct or "excel" in ct:
+            raw_name = cd.split("filename=")[-1].strip().strip('"').strip("'")
+            ext = os.path.splitext(raw_name)[-1].lower()
+        elif "spreadsheet" in ct or "excel" in ct or "openxmlformats" in ct:
             ext = ".xlsx"
         elif "csv" in ct:
             ext = ".csv"
@@ -135,8 +139,10 @@ def fetch_task_file(task_id: str) -> str | None:
             ext = ".png"
         elif "jpeg" in ct or "jpg" in ct:
             ext = ".jpg"
-        elif "python" in ct or "text/plain" in ct:
+        elif "python" in ct:
             ext = ".py"
+        elif "text/plain" in ct:
+            ext = ".txt"
         else:
             ext = ""
 
@@ -144,12 +150,45 @@ def fetch_task_file(task_id: str) -> str | None:
         with open(local_path, "wb") as f:
             f.write(r.content)
 
-        log(f"File     : downloaded → {local_path} ({len(r.content)} bytes)")
-        return local_path
+        log(f"File     : downloaded → {local_path} ({len(r.content)} bytes, type={ext})")
+
+        # For Excel/CSV: read with pandas and inject as text — don't rely on
+        # the model downloading it again inside run_python.
+        injected_text = None
+        if ext in (".xlsx", ".xls"):
+            try:
+                import pandas as pd
+                dfs = pd.read_excel(local_path, sheet_name=None)
+                parts = []
+                for sheet, df in dfs.items():
+                    parts.append(f"=== Sheet: {sheet} ===\n{df.to_csv(index=False)}")
+                injected_text = "\n\n".join(parts)[:6000]
+                log(f"File     : Excel read OK — {sum(len(d) for d in dfs.values())} rows total")
+            except Exception as e:
+                log(f"File     : Excel read failed — {e}")
+
+        elif ext == ".csv":
+            try:
+                import pandas as pd
+                df = pd.read_csv(local_path)
+                injected_text = df.to_csv(index=False)[:6000]
+                log(f"File     : CSV read OK — {len(df)} rows")
+            except Exception as e:
+                log(f"File     : CSV read failed — {e}")
+
+        elif ext in (".py", ".txt"):
+            try:
+                with open(local_path, "r", errors="replace") as f:
+                    injected_text = f.read()[:4000]
+                log(f"File     : text read OK")
+            except Exception as e:
+                log(f"File     : text read failed — {e}")
+
+        return local_path, injected_text
 
     except Exception as e:
         log(f"File     : download failed — {e}")
-        return None
+        return None, None
 
 # ==========================================================
 # MAIN RUNNER
@@ -204,9 +243,20 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
         log(f"Question : {question[:200]}")
         log("-" * 80)
 
-        # Download any attached file and append its path to the question
-        file_path = fetch_task_file(task_id)
-        if file_path:
+        # Download any attached file and inject content into the question
+        file_path, injected_text = fetch_task_file(task_id)
+        if injected_text:
+            # For Excel/CSV/text: embed the content directly so the model
+            # doesn't need to re-download anything
+            ext = os.path.splitext(file_path)[-1].lower() if file_path else ""
+            if ext in (".xlsx", ".xls", ".csv"):
+                question = question + f"\n\n[ATTACHED FILE DATA — use this for calculations, do NOT try to re-read the file]:\n{injected_text}"
+            elif ext in (".py", ".txt"):
+                question = question + f"\n\n[ATTACHED FILE CONTENT — this is the file to process]:\n{injected_text}"
+            else:
+                question = question + f"\n[ATTACHED FILE: {file_path}]"
+        elif file_path:
+            # Binary files (images, PDF) — just pass the path
             question = question + f"\n[ATTACHED FILE: {file_path}]"
 
         start = time.time()