{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# RAG based on Qianfan and BES"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This notebook is an implementation of Retrieval augmented generation (RAG) using Baidu Qianfan Platform combined with Baidu ElasricSearch, where the original data is located on BOS.\n",
    "## Baidu Qianfan\n",
    "Baidu AI Cloud Qianfan Platform is a one-stop large model development and service operation platform for enterprise developers. Qianfan not only provides including the model of Wenxin Yiyan (ERNIE-Bot) and the third-party open-source models, but also provides various AI development tools and the whole set of development environment, which facilitates customers to use and develop large model applications easily.\n",
    "\n",
    "## Baidu ElasticSearch\n",
    "[Baidu Cloud VectorSearch](https://cloud.baidu.com/doc/BES/index.html?from=productToDoc) is a fully managed, enterprise-level distributed search and analysis service which is 100% compatible to open source. Baidu Cloud VectorSearch provides low-cost, high-performance, and reliable retrieval and analysis platform level product services for structured/unstructured data. As a vector database , it supports multiple index types and similarity distance methods. "
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Installation and Setup\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#!pip install qianfan\n",
    "#!pip install bce-python-sdk\n",
    "#!pip install elasticsearch == 7.11.0"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from baidubce.bce_client_configuration import BceClientConfiguration\n",
    "from baidubce.auth.bce_credentials import BceCredentials\n",
    "from langchain.document_loaders.baiducloud_bos_directory import BaiduBOSDirectoryLoader\n",
    "from langchain.text_splitter import RecursiveCharacterTextSplitter\n",
    "from langchain.embeddings.huggingface import HuggingFaceEmbeddings\n",
    "from langchain.vectorstores import BESVectorStore\n",
    "from langchain.llms.baidu_qianfan_endpoint import QianfanLLMEndpoint"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Document loading"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "bos_host = \"your bos eddpoint\"\n",
    "access_key_id = \"your bos access ak\"\n",
    "secret_access_key = \"your bos access sk\"\n",
    "\n",
    "# create BceClientConfiguration\n",
    "config = BceClientConfiguration(\n",
    "    credentials=BceCredentials(access_key_id, secret_access_key), endpoint=bos_host\n",
    ")\n",
    "\n",
    "loader = BaiduBOSDirectoryLoader(conf=config, bucket=\"llm-test\", prefix=\"llm/\")\n",
    "documents = loader.load()\n",
    "\n",
    "text_splitter = RecursiveCharacterTextSplitter(chunk_size=200, chunk_overlap=0)\n",
    "split_docs = text_splitter.split_documents(documents)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Embedding and VectorStore"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "embeddings = HuggingFaceEmbeddings(model_name=\"shibing624/text2vec-base-chinese\")\n",
    "embeddings.client = sentence_transformers.SentenceTransformer(embeddings.model_name)\n",
    "\n",
    "db = BESVectorStore.from_documents(\n",
    "    documents=split_docs,\n",
    "    embedding=embeddings,\n",
    "    bes_url=\"your bes url\",\n",
    "    index_name=\"test-index\",\n",
    "    vector_query_field=\"vector\",\n",
    ")\n",
    "\n",
    "db.client.indices.refresh(index=\"test-index\")\n",
    "retriever = db.as_retriever()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## QA Retriever"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "llm = QianfanLLMEndpoint(\n",
    "    model=\"ERNIE-Bot\",\n",
    "    qianfan_ak=\"your qianfan ak\",\n",
    "    qianfan_sk=\"your qianfan sk\",\n",
    "    streaming=True,\n",
    ")\n",
    "qa = RetrievalQA.from_chain_type(\n",
    "    llm=llm, chain_type=\"refine\", retriever=retriever, return_source_documents=True\n",
    ")\n",
    "\n",
    "query = \"什么是张量?\"\n",
    "print(qa.run(query))"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> 张量（Tensor）是一个数学概念，用于表示多维数据。它是一个可以表示多个数值的数组，可以是标量、向量、矩阵等。在深度学习和人工智能领域中，张量常用于表示神经网络的输入、输出和权重等。"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "name": "python",
   "version": "3.9.17"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "aee8b7b246df8f9039afb4144a1f6fd8d2ca17a180786b69acc140d282b71a49"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}