Update app.py

app.py

@@ -27,4 +27,152 @@ class BasicAgent:
         if not self.hf_token:
             raise RuntimeError("HF_TOKEN missing. Set it in Space Settings → Secrets.")
 
-        # 模型（可在 Variab
+        # 模型（可在 Variables 改）
+        self.model_id = os.getenv("MODEL_ID", "Qwen/Qwen2.5-7B-Instruct")
+
+        # ✅ 正確用法：不要給 base_url
+        self.client = InferenceClient(
+            model=self.model_id,
+            token=self.hf_token,
+            timeout=120,
+        )
+
+        self.system = (
+            "You answer questions with EXACT MATCH.\n"
+            "Return ONLY the final answer.\n"
+            "No explanation.\n"
+            "No extra words.\n"
+            "No punctuation unless required.\n"
+            "No quotes.\n"
+        )
+
+    def _sanitize(self, text: str) -> str:
+        if not text:
+            return ""
+
+        t = str(text).strip()
+
+        t = re.sub(r"(?i)final answer\s*[:\-]*", "", t)
+        t = re.sub(r"(?i)answer\s*[:\-]*", "", t)
+
+        lines = [ln.strip() for ln in t.splitlines() if ln.strip()]
+        if lines:
+            t = lines[-1]
+
+        t = t.strip().strip('"').strip("'")
+        t = re.sub(r"[.,;:!?]$", "", t)
+
+        return t
+
+    def __call__(self, question: str) -> str:
+        print(f"Q: {question[:60]}")
+
+        prompt = f"{self.system}\nQuestion: {question}\nAnswer:"
+
+        try:
+            out = self.client.text_generation(
+                prompt,
+                max_new_tokens=64,
+                temperature=0.0,
+                do_sample=False,
+                return_full_text=False,
+            )
+        except Exception:
+            out = self.client.chat_completion(
+                messages=[
+                    {"role": "system", "content": self.system},
+                    {"role": "user", "content": question},
+                ],
+                max_tokens=64,
+                temperature=0.0,
+            ).choices[0].message.content
+
+        ans = self._sanitize(out)
+        print(f"A: {ans}")
+        return ans
+
+
+# ===============================
+# Run & Submit
+# ===============================
+def run_and_submit_all(profile: gr.OAuthProfile | None):
+
+    space_id = os.getenv("SPACE_ID")
+
+    if not profile:
+        return "Please login with Hugging Face.", None
+
+    username = profile.username
+    print(f"User: {username}")
+
+    questions_url = f"{DEFAULT_API_URL}/questions"
+    submit_url = f"{DEFAULT_API_URL}/submit"
+
+    try:
+        agent = BasicAgent()
+    except Exception as e:
+        return f"Agent init error: {e}", None
+
+    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
+
+    resp = requests.get(questions_url, timeout=20)
+    resp.raise_for_status()
+    questions = resp.json()
+
+    answers_payload = []
+    log_rows = []
+
+    for q in questions:
+        task_id = q["task_id"]
+        question = q["question"]
+        try:
+            ans = agent(question)
+        except Exception:
+            ans = ""
+
+        answers_payload.append({
+            "task_id": task_id,
+            "submitted_answer": ans
+        })
+
+        log_rows.append({
+            "Task ID": task_id,
+            "Question": question,
+            "Submitted Answer": ans
+        })
+
+    submission = {
+        "username": username,
+        "agent_code": agent_code,
+        "answers": answers_payload
+    }
+
+    resp = requests.post(submit_url, json=submission, timeout=60)
+    resp.raise_for_status()
+    result = resp.json()
+
+    status = (
+        f"Submission Successful!\n"
+        f"User: {result.get('username')}\n"
+        f"Score: {result.get('score')}% "
+        f"({result.get('correct_count')}/{result.get('total_attempted')})\n"
+        f"{result.get('message')}"
+    )
+
+    return status, pd.DataFrame(log_rows)
+
+
+# ===============================
+# Gradio UI
+# ===============================
+with gr.Blocks() as demo:
+    gr.Markdown("# Basic Agent Evaluation Runner (PASS MODE)")
+    gr.LoginButton()
+    run_btn = gr.Button("Run Evaluation & Submit All Answers")
+    status = gr.Textbox(label="Result", lines=6)
+    table = gr.DataFrame(label="Answers", wrap=True)
+
+    run_btn.click(fn=run_and_submit_all, outputs=[status, table])
+
+if __name__ == "__main__":
+    demo.launch()