Update app.py

app.py

@@ -58,45 +58,87 @@ def parse_think_blocks(text: str) -> tuple[str, str]:
     m = re.search(r"<think>(.*?)</think>\s*", text, re.DOTALL)
     return (m.group(1).strip(), text[m.end():].strip()) if m else ("", text)
 
+def _is_thinking_line(line: str) -> bool:
+    """한 줄이 reasoning/thinking인지 감지"""
+    l = line.strip()
+    if not l:
+        return True  # 빈 줄은 thinking 블록에 포함
+    # 영어 reasoning 패턴
+    think_starts = [
+        "The user", "the user", "This is", "this is", "I should", "I need to",
+        "Let me", "let me", "My task", "my task", "I'll ", "I will",
+        "Since ", "since ", "Now,", "now,", "So,", "so,", "First,", "first,",
+        "Okay", "okay", "Alright", "Hmm", "Wait", "Actually",
+        "The question", "the question", "The input", "the input",
+        "The request", "the request", "The prompt", "the prompt",
+        "Thinking Process", "Thinking process", "**Thinking",
+        "Step ", "step ", "Approach:", "Analysis:", "Reasoning:",
+        "1. **", "2. **", "3. **", "4. **", "5. **",
+    ]
+    for s in think_starts:
+        if l.startswith(s):
+            return True
+    # 글머리 기호 + 영어 reasoning
+    if l.startswith(("- ", "* ", "○ ")) and any(c.isascii() and c.isalpha() for c in l[:20]):
+        if not any(ord(c) > 0x1100 for c in l[:30]):  # 한글 없으면 thinking
+            return True
+    return False
+
+
+def _split_thinking_answer(raw: str) -> tuple:
+    """응답에서 thinking 부분과 실제 답변을 분리"""
+    lines = raw.split("\n")
+    answer_start = -1
+    
+    for i, line in enumerate(lines):
+        if not _is_thinking_line(line):
+            # 한글/비ASCII가 포함된 실제 답변 시작점
+            if any(ord(c) > 0x1100 for c in line.strip()[:10]):
+                answer_start = i
+                break
+            # 또는 빈줄 2개 이후 영어 답변
+            if i > 2 and not _is_thinking_line(line):
+                # 앞의 2줄이 빈줄이면 답변 시작
+                if all(not lines[j].strip() for j in range(max(0,i-2), i)):
+                    answer_start = i
+                    break
+    
+    if answer_start > 0:
+        thinking = "\n".join(lines[:answer_start]).strip()
+        answer = "\n".join(lines[answer_start:]).strip()
+        return thinking, answer
+    
+    return "", raw
+
+
 def format_response(raw: str) -> str:
-    # 1. <think>...</think> 완료된 경우
+    # 1. <think>...</think> 완료
     chain, answer = parse_think_blocks(raw)
     if chain:
-        lines  = chain.split("\n")
-        quoted = "\n".join(f"> {l}" for l in lines)
-        block  = (
+        return (
             "<details>\n"
             "<summary>🧠 Reasoning Chain — click to expand</summary>\n\n"
-            f"{quoted}\n\n"
+            f"{chain}\n\n"
             "</details>\n\n"
+            f"{answer}"
         )
-        return block + answer
-    # 2. <think> 열림 but 닫히지 않음 — 스트리밍 중
+    # 2. <think> 열림 but 닫히지 않음
     if "<think>" in raw and "</think>" not in raw:
         think_len = len(raw) - raw.index("<think>") - 7
         return f"🧠 Reasoning... ({think_len} chars)"
-    # 3. NIPA 엔드포인트: 평문 "Thinking Process:" 감지
-    think_markers = ["Thinking Process:", "Thinking process:", "**Thinking Process", "1. **Analyze"]
-    for marker in think_markers:
-        if marker in raw:
-            # 아직 최종 답변이 안 나왔으면 캐릭터 수만 표시
-            # 최종 답변 시작 패턴: 한국어 인사, 영어 인사, 번호 없는 일반 문장
-            answer_patterns = [
-                "\n\n안녕", "\n\n반갑", "\n\n네,", "\n\n Hello", "\n\nHi", "\n\nI ",
-                "\n\n---\n", "\n\n**Answer", "\n\n**Response",
-            ]
-            for ap in answer_patterns:
-                idx = raw.find(ap)
-                if idx > 0:
-                    think_part = raw[:idx].strip()
-                    answer_part = raw[idx:].strip()
-                    return (
-                        f"<details>\n"
-                        f"<summary>🧠 Reasoning Chain ({len(think_part)} chars)</summary>\n\n"
-                        f"{think_part}\n\n"
-                        f"</details>\n\n"
-                        f"{answer_part}"
-                    )
+    # 3. 평문 thinking 감지 (NIPA 엔드포인트)
+    first_line = raw.strip().split("\n")[0] if raw.strip() else ""
+    if _is_thinking_line(first_line) and len(raw) > 20:
+        thinking, answer = _split_thinking_answer(raw)
+        if thinking and answer:
+            return (
+                f"<details>\n"
+                f"<summary>🧠 Reasoning Chain ({len(thinking)} chars)</summary>\n\n"
+                f"{thinking}\n\n"
+                f"</details>\n\n"
+                f"{answer}"
+            )
+        elif thinking and not answer:
             # 아직 답변 안 나옴 — 캐릭터 수만
             return f"🧠 Reasoning... ({len(raw)} chars)"
     return raw