Update app.py

app.py

@@ -66,36 +66,56 @@ atexit.register(cleanup_temp_files)
 @tool
 def search_web(query: str, max_results: int = 3) -> str:
     """
-    Websuche via Tavily oder DuckDuckGo.
+    Search the web for the given query and return a summary of the top results.
+
+    Args:
+        query (str): The search query to look up online.
+        max_results (int): The maximum number of search results to retrieve.
+
+    Returns:
+        str: A concatenated string summarizing the URLs and snippets of the results.
     """
     if not search_client:
-        return "Search tool not configured."
+        return "Error: Search tool not configured."
     try:
         if USE_TAVILY and isinstance(search_client, TavilyClient):
             res = search_client.search(query=query, search_depth="basic", max_results=max_results)
             items = res.get('results', [])
-            return "\n".join([f"URL: {i['url']}\n{ i['content'][:200] }..." for i in items])
+            if not items:
+                return "No search results found."
+            return "\n".join([f"URL: {i['url']}\nContent: {i.get('content','')[:200]}..." for i in items])
         elif USE_DUCKDUCKGO and isinstance(search_client, DDGS):
             results = search_client.text(query, max_results=max_results)
-            return "\n".join([f"Title: {r['title']}\nURL: {r['href']}\n{r['body'][:200]}..." for r in results])
+            if not results:
+                return "No search results found."
+            return "\n".join([f"Title: {r['title']}\nURL: {r['href']}\nSnippet: {r['body'][:200]}..." for r in results])
+        else:
+            return "Error: No compatible search client available."
     except Exception as e:
         return f"Error during search: {e}"
-    return "No results."
 
 @tool
 def download_task_file(task_id: str) -> str:
-    """Download einer Datei zur Task ID vom Server."""
+    """
+    Download the file associated with a specific task ID from the evaluation server.
+
+    Args:
+        task_id (str): Unique identifier of the task whose file should be downloaded.
+
+    Returns:
+        str: Local filesystem path to the downloaded file, or an error message.
+    """
     url = f"{DEFAULT_API_URL}/files/{task_id}"
     try:
-        r = requests.get(url, stream=True, timeout=30)
-        r.raise_for_status()
-        ct = r.headers.get('content-type', '')
-        suffix = '.pdf' if 'pdf' in ct else '.tmp'
-        tmp = tempfile.gettempdir()
-        name = f"task_{task_id}_{datetime.now().strftime('%Y%m%d%H%M%S')}{suffix}"
-        path = os.path.join(tmp, name)
+        response = requests.get(url, stream=True, timeout=30)
+        response.raise_for_status()
+        content_type = response.headers.get('content-type', '')
+        suffix = '.pdf' if 'pdf' in content_type else '.tmp'
+        tmp_dir = tempfile.gettempdir()
+        filename = f"task_{task_id}_{datetime.now().strftime('%Y%m%d%H%M%S')}{suffix}"
+        path = os.path.join(tmp_dir, filename)
         with open(path, 'wb') as f:
-            for chunk in r.iter_content(8192): f.write(chunk)
+            for chunk in response.iter_content(8192): f.write(chunk)
         temp_files_to_clean.add(path)
         return path
     except Exception as e:
@@ -103,21 +123,35 @@ def download_task_file(task_id: str) -> str:
 
 @tool
 def read_file_content(file_path: str) -> str:
-    """Liest Text aus einer heruntergeladenen Datei."""
+    """
+    Read the text content of a previously downloaded file (PDF or plain text).
+
+    Args:
+        file_path (str): Absolute local path to the file to read (from download_task_file).
+
+    Returns:
+        str: Extracted text content truncated if necessary, or an error message.
+    """
     if not file_path.startswith(tempfile.gettempdir()):
-        return "Error: Invalid path."
+        return "Error: Invalid file path."
+    if not os.path.exists(file_path):
+        return "Error: File does not exist."
     try:
-        if file_path.endswith('.pdf'):
-            if not PDF_READER_AVAILABLE: return "Error: PyPDF2 fehlt."
-            txt = ''
+        if file_path.lower().endswith('.pdf'):
+            if not PDF_READER_AVAILABLE:
+                return "Error: PyPDF2 not installed."
+            text = ''
             with open(file_path, 'rb') as f:
-                rdr = PyPDF2.PdfReader(f)
-                for p in rdr.pages:
-                    txt += p.extract_text() or ''
-                    if len(txt) > 5000: break
-            return txt
+                reader = PyPDF2.PdfReader(f)
+                for page in reader.pages:
+                    text += page.extract_text() or ''
+                    if len(text) > 7000:
+                        text = text[:7000] + '\n... (truncated)'
+                        break
+            return text
         else:
-            return open(file_path, 'r', encoding='utf-8', errors='ignore').read(5000)
+            with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:
+                return f.read(7000)
     except Exception as e:
         return f"Error: {e}"
 
@@ -129,63 +163,79 @@ def initialize_agent():
             key = os.getenv('TAVILY_API_KEY')
             search_client = TavilyClient(api_key=key) if key else False
         elif USE_DUCKDUCKGO:
-            search_client = DDGS()
+            try:
+                search_client = DDGS()
+            except:
+                search_client = False
         else:
             search_client = False
     token = os.getenv('HUGGINGFACE_TOKEN')
     if not token:
-        raise ValueError("HUGGINGFACE_TOKEN fehlt.")
+        raise ValueError("HUGGINGFACE_TOKEN environment variable is required.")
     hf_model = HfApiModel()
     tools = [search_web, download_task_file, read_file_content]
-    if not search_client: tools = [t for t in tools if t != search_web]
+    if not search_client:
+        tools.remove(search_web)
     agent_instance = CodeAgent(tools=tools, model=hf_model)
 
-# --- Hauptfunktion ---
+# --- Main Logic ---
 def run_and_submit_all(profile, progress=gr.Progress(track_tqdm=True)):
-    # Profil parsen (evtl. JSON-String)
+    # Parse profile if passed as JSON string
     if isinstance(profile, str):
-        try: profile = json.loads(profile)
-        except: return "Ungültiges Profilformat.", None
+        try:
+            profile = json.loads(profile)
+        except json.JSONDecodeError:
+            return "Error: Invalid profile format.", None
     if not profile or 'username' not in profile:
         return "Bitte zuerst einloggen.", None
     username = profile['username']
+
     try:
         initialize_agent()
     except Exception as e:
-        return f"Init-Error: {e}", None
+        return f"Initialization Error: {e}", None
+
+    # Fetch questions
+    resp = requests.get(f"{DEFAULT_API_URL}/questions", timeout=30)
+    resp.raise_for_status()
+    questions = resp.json()
 
-    # Fragen holen
-    questions = requests.get(f"{DEFAULT_API_URL}/questions").json()
     logs, payload = [], []
-    for item in progress.tqdm(questions, desc="Bearbeite" ):
-        tid, q = item.get('task_id'), item.get('question')
-        if not tid or not q: continue
-        prompt = f"Task {tid}: {q}"
+    for item in progress.tqdm(questions, desc="Bearbeite Fragen"):
+        task_id = item.get('task_id')
+        question = item.get('question')
+        if not task_id or question is None:
+            continue
+        prompt = f"Task {task_id}: {question}"
         try:
-            res = agent_instance.run(prompt=prompt)
-            ans = re.sub(r"^(Answer:|Final Answer:)", "", res or "").strip()
+            result = agent_instance.run(prompt=prompt)
+            answer = re.sub(r"^(Answer:|Final Answer:)", "", result or "").strip()
         except Exception as e:
-            ans = f"ERROR: {e}"
-        logs.append({'Task ID': tid, 'Question': q, 'Submitted Answer': ans})
-        payload.append({'task_id': tid, 'submitted_answer': ans})
+            answer = f"ERROR: {e}"
+        logs.append({'Task ID': task_id, 'Question': question, 'Submitted Answer': answer})
+        payload.append({'task_id': task_id, 'submitted_answer': answer})
+
     df = pd.DataFrame(logs)
 
-    # Submission
-    sub = {'username': username, 'agent_code': '...', 'answers': payload}
+    # Submit answers
+    submission = {'username': username, 'agent_code': '...', 'answers': payload}
     try:
-        r = requests.post(f"{DEFAULT_API_URL}/submit", json=sub, timeout=180)
-        r.raise_for_status()
-        status = f"Erfolg: {r.json().get('score',0):.2f}%"
+        post = requests.post(f"{DEFAULT_API_URL}/submit", json=submission, timeout=180)
+        post.raise_for_status()
+        score = post.json().get('score', 0)
+        status = f"Erfolg! Score: {score:.2f}%"
     except Exception as e:
-        status = f"Submit-Error: {e}"
+        status = f"Submission Error: {e}"
+
     cleanup_temp_files()
     return status, df
 
 # --- Gradio UI ---
 with gr.Blocks() as demo:
     gr.Markdown("# Smol CodeAgent Evaluation Runner")
-    gr.Markdown("Bitte einloggen und dann auf Ausführen klicken.")
-    with gr.Row(): login_btn = gr.LoginButton()
+    gr.Markdown("Bitte einloggen und dann auf "Run Evaluation & Submit All Answers" klicken.")
+    with gr.Row():
+        login_btn = gr.LoginButton()
     run_btn = gr.Button("Run Evaluation & Submit All Answers")
     out_status = gr.Textbox(label="Status", lines=5)
     out_table = gr.DataFrame(label="Ergebnisse")