app.py

from fastapi import FastAPI
from pydantic import BaseModel
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

# 初始化 Qwen 模型與 tokenizer（加上 trust_remote_code）
model_id = "Qwen/Qwen-1_8B-Chat"
device = "cuda" if torch.cuda.is_available() else "cpu"

print(f"🚀 載入模型：{model_id} on {device}")

tokenizer = AutoTokenizer.from_pretrained(
    model_id,
    trust_remote_code=True
)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    trust_remote_code=True,
    torch_dtype=torch.float32
).to(device)

# 建立 FastAPI 應用
app = FastAPI()

chat_history = []

class Prompt(BaseModel):
    text: str
    reset: bool = False

@app.post("/chat")
async def chat(prompt: Prompt):
    global chat_history

    print(f"\n📝 使用者輸入：{prompt.text}")
    if prompt.reset:
        chat_history = []
        print("🔄 Chat history 已重置")

    chat_history.append({"role": "user", "content": prompt.text})

    # 組合 ChatML 格式
    chatml = ""
    for msg in chat_history:
        chatml += f"<|im_start|>{msg['role']}\n{msg['content']}\n<|im_end|>\n"
    chatml += "<|im_start|>assistant\n"

    try:
        inputs = tokenizer(chatml, return_tensors="pt").to(device)
        outputs = model.generate(
            **inputs,
            max_new_tokens=512,
            do_sample=True,
            temperature=0.7,
            top_p=0.9
        )
        response = tokenizer.decode(outputs[0], skip_special_tokens=True).strip()

        print("🧠 原始模型回覆：", response)

        # 擷取 assistant 回覆內容
        if "<|im_start|>assistant\n" in response:
            reply = response.split("<|im_end|>")[0].split("<|im_start|>assistant\n")[-1].strip()
        else:
            reply = response  # fallback

        if not reply:
            reply = "⚠️ 模型未產生回覆，請稍後再試。"
            print("⚠️ 回覆為空字串")

        chat_history.append({"role": "assistant", "content": reply})
        print("✅ 最終回覆：", reply)
        return {"reply": reply}

    except Exception as e:
        print("❌ 模型回應錯誤：", e)
        return {"reply": "目前無法取得模型回覆，請稍後再試。"}

if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=7860)