Update app.py

app.py

@@ -11,7 +11,7 @@ import gradio as gr
 from together import Together
 
 load_dotenv("../.env")
-print("Environment variables:", os.environ)
+# print("Environment variables:", os.environ)
 
 
 together_client = Together(
@@ -80,16 +80,8 @@ def get_naver_answers( message: str ):
     print( "Query: ", query)
 
     context = get_answers( query )
-
-    answer_count = 0
-    sorted_answers = []
-    for answer in context:
-        curr = '. '.join(answer['answers'])
-        answer_count += len(curr)
-        sorted_answers.append(curr)
-        if answer_count > 55000:
-            break
-    # sorted_answers = ['. '.join(answer['answers']) for answer in context]
+        
+    sorted_answers = ['. '.join(answer['answers']) for answer in context]
     naver_end_time = time.time()
     print(f"Time taken to extract from Naver: { naver_end_time - naver_start_time } seconds")
     document = '\n'.join(sorted_answers)
@@ -116,9 +108,10 @@ def get_qwen_small_answer( message: str ):
     response = together_client.chat.completions.create(
         model="Qwen/Qwen2.5-7B-Instruct-Turbo",
         messages=[
-            {"role": "system", "content": "You are a helpful question-answer, CONCISE conversation assistant that answers in Korean."},
+            {"role": "system", "content": "You are a helpful question-answer, conversation assistant that answers in Korean. Your responses should sound human-like."},
             {"role": "user", "content": message}
-        ]
+        ],
+        max_tokens = None
         #TODO: Change the messages option
     )
     qwen_end_time = time.time()
@@ -262,18 +255,29 @@ def compare_answers(message: str):
     return results
 
 def chatFunction( message, history ):
-    content = f'아래 문서를 바탕으로 질문에 답하세요. 답변에서 질문을 따라 출력 하지 마세요. 답변은 한국어로만 해주세요! 찾은 Naver 문서와 다른 문서에서 답변이 없는 내용은 절대 출력하지 마세요 \n 질문: {message}\n 문서: '
+    content = f'아래 문서를 바탕으로 질문에 답하세요. 답변에서 질문을 따라 출력 하지 마세요. 답변은 한국어로만 해주세요! 찾은 Naver 문서와 다른 문서에서 답변이 없는 내용은 절대 출력하지 마세요. 친절하고 인간답게 말하세요. \n 질문: {message}\n 문서: '
     naver_docs, naver_time_taken = get_naver_answers( message )
 
+    if len(naver_docs) > 55000:
+        overlap = 200
+        answers = []
+        split_len = len(naver_docs) // ( ( len(naver_docs) - 55000 ) // 55000 + 2 ) + 1
+        for i in range( len(naver_docs), split_len ):
+            if i == 0:
+                split = naver_docs[:split_len]
+            else:
+                split = naver_docs[i * split_len - overlap: (i + 1) * split_len]
+            answer, _ = get_qwen_small_answer(f"Summarize important points in a paragraph, given the information below, using only Korean language. Give me only the summary!!! \n {split}")
+            answers.append(answer)
+        naver_docs = '\n'.join(answers)
+
     start_time = time.time()
     content += "\n Naver 문서: " + naver_docs
 
-    print( ">>> Length: ", len(naver_docs) )
-
     completion = gpt_client.chat.completions.create(
         model="gpt-4o-mini",
         messages=[
-            {"role": "system", "content": "You are a helpful assistant that answers only in korean."},
+            {"role": "system", "content": "You are a helpful assistant that gives detailed answers only in korean."},
             {
                 "role": "user",
                 "content": message