opc_agent/agent-craft
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

优化代码，提高解耦

Browse Source
This commit is contained in:
Annyfee
2025-11-19 23:44:14 +08:00
parent 8265c2a7a9
commit a70517f6d8
25 changed files with 143 additions and 355 deletions
Download Patch File Download Diff File Expand all files Collapse all files
07_rag_advanced/03_rag_as_tool.py
+20 -48
View File
@@ -1,25 +1,21 @@
import os
from dotenv import load_dotenv
load_dotenv()
api_key = os.getenv("OPENAI_API_KEY")
from langchain_huggingface import HuggingFaceEmbeddings
from config import OPENAI_API_KEY
from embeddings import get_embeddings
from langchain_chroma import Chroma
from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough
from langchain.retrievers import ContextualCompressionRetriever
from langchain_classic.retrievers import ContextualCompressionRetriever
from langchain_community.cross_encoders import HuggingFaceCrossEncoder
from langchain.retrievers.document_compressors import CrossEncoderReranker
from langchain_classic.retrievers.document_compressors import CrossEncoderReranker
from langchain_core.tools import tool
# 全局 LLM (供Agent和Rag共用)
llm = ChatOpenAI(
model="deepseek-chat",
api_key=api_key,
api_key=OPENAI_API_KEY,
base_url="https://api.deepseek.com"
)
@@ -27,23 +23,26 @@ llm = ChatOpenAI(
def build_rag_chain(llm_instance):
print('---正在构建RAG链条...---\n')
Persist_directory = './chroma_db_war_and_peace_bge_small_en_v1.5'
Embedding_model = 'BAAI/bge-small-en-v1.5'
Encoder_model = "BAAI/bge-reranker-base"
persist_directory = './chroma_db_war_and_peace_bge_small_en_v1.5'
embedding_model_name = 'BAAI/bge-small-en-v1.5'
encoder_model_name = "BAAI/bge-reranker-base"
if not os.path.exists(Persist_directory):
raise FileNotFoundError(f'索引目录{Persist_directory}未找到，请先运行 build_index.py')
if not os.path.exists(persist_directory):
raise FileNotFoundError(f'索引目录{persist_directory}未找到，请先运行 build_index.py')
embeddings_model = HuggingFaceEmbeddings(model_name=Embedding_model)
print(f'正在加载/下载 Embedding模型：{embedding_model_name}')
embeddings_model = get_embeddings(model_name=embedding_model_name,device='cpu')
db = Chroma(
persist_directory=Persist_directory,
persist_directory=persist_directory,
embedding_function=embeddings_model
)
# 1. R-检索--强化版
base_retriever = db.as_retriever(search_kwargs={"k":5})
encoder = HuggingFaceCrossEncoder(model_name=Encoder_model)
reranker = CrossEncoderReranker(model=encoder,top_n=2)
base_retriever = db.as_retriever(search_kwargs={"k":50})
print(f'正在加载 Reranker模型:{encoder_model_name}...')
encoder = HuggingFaceCrossEncoder(model_name=encoder_model_name)
reranker = CrossEncoderReranker(model=encoder,top_n=6)
compression_retriever=ContextualCompressionRetriever(
base_retriever=base_retriever,
base_compressor=reranker
@@ -59,6 +58,7 @@ def build_rag_chain(llm_instance):
[上下文]: {context}
[问题]: {question}
"""
prompt = ChatPromptTemplate.from_messages([
('system',sys_prompt),
('human','{question}')
@@ -105,32 +105,4 @@ if __name__ == '__main__':
question = "皮埃尔是共济会成员吗？他在其中扮演什么角色？"
res = search_war_and_peace.invoke(question)
print(f'问题:{question}')
print(f'回答:{res}')
print(f'回答:{res}')