from supabase import create_client
from openai import OpenAI
from dotenv import load_dotenv
import os

# 加载环境变量
load_dotenv()

# 初始化 Supabase 客户端
supabase = create_client(
    os.getenv("SUPABASE_URL"),
    os.getenv("SUPABASE_KEY")
)

# 初始化 DeepSeek AI 客户端
ai = OpenAI(
    api_key=os.getenv("DEEPSEEK_API_KEY"),
    base_url="https://api.deepseek.com"
)

def embed(text: str) -> list[float]:
    """生成文本向量"""
    res = ai.embeddings.create(
        model="text-embedding-ada-002",
        input=text
    )
    return res.data[0].embedding

def retrieve(query: str, top_k: int = 3) -> list[str]:
    """RAG:检索最相关的文档段落"""
    vec = embed(query)
    res = supabase.rpc("match_documents", {
        "query_embedding": vec,
        "match_count": top_k,
        "match_threshold": 0.5
    }).execute()
    return [r["content"] for r in res.data]

# 测试检索
if __name__ == "__main__":
    contexts = retrieve("E 配速跑心率应该控制在多少")
    for c in contexts:
        print(f"· {c[:60]}...")