Update app.py

app.py

@@ -1,48 +1,56 @@
 import gradio as gr
 from huggingface_hub import InferenceClient
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+from fastapi import FastAPI, HTTPException, Request
+from pydantic import BaseModel
+from typing import List
+from sentence_transformers import CrossEncoder, util
+from transformers import AutoTokenizer, AutoModelForCausalLM
+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()
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
+# 初始化模型和資料庫
+model_name = "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
+collection_name = model_name.split("/")[-1]
+cross_encoder_model = "cross-encoder/mmarco-mMiniLMv2-L12-H384-v1"
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-0.5B-Instruct", trust_remote_code=True)
+llm_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-0.5B-Instruct", trust_remote_code=True)
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+# 構建向量資料庫
+build_qa_vector_store(model_name, collection_name)
 
-    messages.append({"role": "user", "content": message})
+# 輸入格式
+class QueryInput(BaseModel):
+    query: str
+    top_k: int = 5
 
-    response = ""
+# 輸出格式
+class SearchResult(BaseModel):
+    text: str
+    score: float
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+# 搜尋＋rerank API
+@app.post("/search", response_model=List[SearchResult])
+def search(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。")
+    
+    final_passages = [r[0] for r in reranked] 
 
-        response += token
-        yield response
+    # 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 answer
 
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[