Update app.py

app.py

@@ -10,9 +10,6 @@ import torch
 import numpy as np
 from qa_vector_store import build_qa_vector_store, retrieve_and_rerank, generate_response_from_local_llm
 
-# 建立 FastAPI 應用
-app = FastAPI()
-
 # 初始化模型和資料庫
 model_name = "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
 collection_name = model_name.split("/")[-1]
@@ -34,21 +31,15 @@ class SearchResult(BaseModel):
     score: float
 
 # 搜尋＋rerank API
-@app.post("/chat", response_model=List[SearchResult])
-def search_and_generate(input: QueryInput):
-    reranked = retrieve_and_rerank(input.query, model_name, collection_name, cross_encoder_model, score_threshold=0.5, search_top_k=20, rerank_top_k=input.top_k)
-    
-    # 如果沒有找到相關答案，則返回 404 錯誤
-    if not reranked:
-        raise HTTPException(status_code=404, detail="找不到相關答案，請嘗試換個問題或降低門檻。")
-    
+def search_and_generate:
+    reranked = retrieve_and_rerank(input.query, model_name, collection_name, cross_encoder_model, score_threshold=0.5, search_top_k=20, rerank_top_k=5)
     final_passages = [r[0] for r in reranked] 
 
     # 使用 LLM 生成回答
     answer = generate_response_from_local_llm(input.query, final_passages, tokenizer, llm_model, max_new_tokens=256)
     
     if not answer:
-        raise HTTPException(status_code=404, detail="無法生成回答，請檢查輸入或模型設定。")
+        return "沒有資料，請重問。"
     return answer
 
 # demo = gr.ChatInterface(
@@ -73,22 +64,16 @@ def search_and_generate(input: QueryInput):
 
 import gradio as gr
 
-# def respond(message, history, system_message, max_tokens, temperature, top_p):
-#     try:
-#         llm.temperature = temperature
-#         llm.max_output_tokens = max_tokens
-
-#         search_results = hybrid_search(message)
-
-#         rerank_response = rerank_chunks_with_llm(message, search_results, llm, top_n=3)
-#         reranked_indices = [int(i.strip()) - 1 for i in rerank_response.content.split(",") if i.strip().isdigit()]
-#         reranked_docs = [search_results[i] for i in reranked_indices]
+def respond(message, history, system_message, max_tokens, temperature, top_p):
+    try:
+        llm.temperature = temperature
+        llm.max_output_tokens = max_tokens
 
-#         answer = generate_answer_with_rag(message, reranked_docs, llm)
-#         return answer.content
+        answer = search_and_generate(message)
+        return answer.content
 
-#     except Exception as e:
-#         return f"[錯誤] {str(e)}"
+    except Exception as e:
+        return f"[錯誤] {str(e)}"
 
 chat_interface = gr.ChatInterface(
     fn=search_and_generate,