Update app.py

app.py

@@ -23,6 +23,15 @@ from datetime import datetime
 from typing import Generator, List, Dict, Optional
 from xml.etree import ElementTree as ET
 
+# Groq 라이브러리 임포트
+try:
+    from groq import Groq
+    GROQ_AVAILABLE = True
+    print("✅ Groq library loaded")
+except ImportError:
+    GROQ_AVAILABLE = False
+    print("❌ Groq library not available - pip install groq")
+
 # ============== Comic Style CSS ==============
 COMIC_CSS = """
 @import url('https://fonts.googleapis.com/css2?family=Bangers&family=Comic+Neue:wght@400;700&display=swap');
@@ -872,45 +881,46 @@ def convert_hwp_to_markdown(input_path: str) -> tuple:
         return text, None
     return None, error
 
-# ============== LLM API ==============
-def call_groq_api_stream(messages: List[Dict], api_key: str) -> Generator[str, None, None]:
-    if not api_key:
-        yield "❌ Groq API 키가 설정되지 않았습니다. 환경변수 GROQ_API_KEY를 설정해주세요."
+# ============== LLM API (Groq 라이브러리 사용) ==============
+def call_groq_api_stream(messages: List[Dict]) -> Generator[str, None, None]:
+    """Groq API 스트리밍 호출 - openai/gpt-oss-120b 모델 사용"""
+    if not GROQ_AVAILABLE:
+        yield "❌ Groq 라이브러리가 설치되지 않았습니다. pip install groq"
+        return
+    
+    if not GROQ_API_KEY:
+        yield "❌ GROQ_API_KEY 환경변수가 설정되지 않았습니다."
         return
+    
     try:
-        response = requests.post(
-            "https://api.groq.com/openai/v1/chat/completions",
-            headers={"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"},
-            json={
-                "model": "mopenai/gpt-oss-120b",
-                "messages": messages,
-                "temperature": 0.7,
-                "max_tokens": 8192,
-                "stream": True
-            },
-            stream=True
+        client = Groq(api_key=GROQ_API_KEY)
+        
+        completion = client.chat.completions.create(
+            model="openai/gpt-oss-120b",
+            messages=messages,
+            temperature=1,
+            max_completion_tokens=8192,
+            top_p=1,
+            reasoning_effort="medium",
+            stream=True,
+            stop=None
         )
-        if response.status_code != 200:
-            yield f"❌ Groq API 오류: {response.status_code}"
-            return
-        for line in response.iter_lines():
-            if line:
-                line = line.decode('utf-8')
-                if line.startswith('data: ') and line[6:] != '[DONE]':
-                    try:
-                        data = json.loads(line[6:])
-                        content = data.get('choices', [{}])[0].get('delta', {}).get('content', '')
-                        if content:
-                            yield content
-                    except:
-                        continue
+        
+        for chunk in completion:
+            if chunk.choices[0].delta.content:
+                yield chunk.choices[0].delta.content
+                
     except Exception as e:
-        yield f"❌ API 오류: {str(e)}"
-
-def call_fireworks_api_stream(messages: List[Dict], image_base64: str, mime_type: str, api_key: str) -> Generator[str, None, None]:
-    if not api_key:
-        yield "❌ Fireworks API 키가 설정되지 않았습니다. 환경변수 FIREWORKS_API_KEY를 설정해주세요."
+        error_msg = str(e)
+        print(f"❌ Groq API 오류: {error_msg}")
+        yield f"❌ API 오류: {error_msg}"
+
+def call_fireworks_api_stream(messages: List[Dict], image_base64: str, mime_type: str) -> Generator[str, None, None]:
+    """Fireworks API 스트리밍 호출 (이미지 분석용)"""
+    if not FIREWORKS_API_KEY:
+        yield "❌ FIREWORKS_API_KEY 환경변수가 설정되지 않았습니다."
         return
+    
     try:
         formatted_messages = [{"role": m["role"], "content": m["content"]} for m in messages[:-1]]
         formatted_messages.append({
@@ -920,21 +930,24 @@ def call_fireworks_api_stream(messages: List[Dict], image_base64: str, mime_type
                 {"type": "text", "text": messages[-1]["content"]}
             ]
         })
+        
         response = requests.post(
             "https://api.fireworks.ai/inference/v1/chat/completions",
-            headers={"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"},
+            headers={"Authorization": f"Bearer {FIREWORKS_API_KEY}", "Content-Type": "application/json"},
             json={
                 "model": "accounts/fireworks/models/qwen3-vl-235b-a22b-thinking",
-                "max_tokens": 8000,
+                "max_tokens": 4096,
                 "temperature": 0.6,
                 "messages": formatted_messages,
                 "stream": True
             },
             stream=True
         )
+        
         if response.status_code != 200:
             yield f"❌ Fireworks API 오류: {response.status_code}"
             return
+        
         for line in response.iter_lines():
             if line:
                 line = line.decode('utf-8')
@@ -1093,6 +1106,7 @@ def chat_response(message: str, history: List[Dict], file: Optional[str],
     
     # 디버그 로그
     print(f"\n🤖 [API 요청]")
+    print(f"  - 모델: openai/gpt-oss-120b")
     print(f"  - 메시지 수: {len(api_messages)}")
     print(f"  - 파일 타입: {file_type}")
     print(f"  - 문서 길이: {len(file_content) if file_content else 0} 글자")
@@ -1102,12 +1116,12 @@ def chat_response(message: str, history: List[Dict], file: Optional[str],
     # 응답 생성
     full_response = ""
     if file_type == "image":
-        for chunk in call_fireworks_api_stream(api_messages, file_content, file_mime, FIREWORKS_API_KEY):
+        for chunk in call_fireworks_api_stream(api_messages, file_content, file_mime):
             full_response += chunk
             history[-1] = {"role": "assistant", "content": full_response}
             yield history, session_id
     else:
-        for chunk in call_groq_api_stream(api_messages, GROQ_API_KEY):
+        for chunk in call_groq_api_stream(api_messages):
             full_response += chunk
             history[-1] = {"role": "assistant", "content": full_response}
             yield history, session_id