Chapter 8: Beyond Cosine — Matching Strategies Determine the Quality of Your Retrieval

In the previous tutorial, we introduced how to improve the recall rate of the retrieval module through query rewriting, various optimized retrieval strategies, and recall rearrangement strategies. One of the key components that affects the quality of retrieved and recalled documents is similarity, which is used to calculate the similarity between the retrieved documents and the query query. The similarity calculation functions provided by LazyLLM by default are bm25 (divided into English and Chinese) similarity calculation and cosine similarity calculation methods. Among them, the bm25 algorithm mainly calculates text, while the cosine similarity algorithm mainly calculates embedding. If the default similarity calculation method provided by LazyLLM cannot meet your needs, you can design and define a similarity calculation method that meets your own needs. Similar to the Similarity component, this tutorial also introduces the custom methods of Transform.

This tutorial mainly introduces how to use custom Similarity components and Transform methods. After reading this tutorial, you will learn how to customize Similarity and Transform methods in LazyLLM, and build a simple RAG application based on Similarity and Transform respectively.

Why customize Similarity

In the RAG service, the effect of the retrieval module directly affects the relevance and accuracy of the generated results. Different application scenarios have different understandings and requirements for "similarity". For example, in legal or medical documents, the emphasis is on precise matching at the keyword level; in general question and answer scenarios, more emphasis is placed on semantic-level understanding; and in multi-modal scenarios, it may involve alignment matching between text and images or structured data. In order to adapt to diverse needs, LazyLLM provides two general similarity calculation methods: BM25/BM25_chinese and cosine similarity. Although both perform well in general scenarios, they still have limitations in some specific tasks:

• BM25 cannot understand semantic approximation: it is based on word frequency and word surface matching, and cannot handle the normalization problem of synonyms, synonymous terms or industry-specific expressions;
• Cosine similarity has coarser granularity: In long texts or documents with complex structures, locally related content is easily averaged as a whole, thus diluting the matching effect.

Let's take a look at an example about the problem that BM25 cannot understand domain semantics (synonyms, term normalization).

① Sample background

• User query: "Dietary intervention methods for hyperglycemia"
• Query description: When users query medical-related information, they use synonymous expressions of professional terms.
• Candidate document: "People with diabetes should control carbohydrate intake to regulate blood sugar levels."

② Ideal Match Expectation

"Hyperglycemia" is one of the core manifestations of "diabetes". Although query and document formulations are different, they are highly related in the medical context. An ideal retrieval system should be able to understand the semantic equivalence between the two, so as to hit the document with high quality.

③Problem Analysis

When using BM25 or BM25_chinese, since they only focus on word-face overlap, "hyperglycemia" is different from "diabetes", so the score is lower and is most likely to be missed. This type of method based on surface matching is insufficient in semantic retrieval in professional fields. Therefore, in order to improve retrieval quality, it is often necessary to introduce more semantically sensitive and domain-customized similarity strategies in actual systems to adapt to complex and changing business needs.

Next, look at another example of the problem of cosine similarity being diluted in long documents.

① Sample background

• User query: "Who has excellent project management skills?"
• Query description: The user is looking for a piece of content describing "Zhang San's project management capabilities".
• Candidate document (long document summary): "Zhang San graduated from a key university and has many years of software development experience. He participated in the R&D and deployment of multiple large-scale projects, including a well-known Agent application development framework. His main responsibilities include code implementation, team collaboration and task allocation. In addition, he also assumes the role of project leader in multiple key nodes, demonstrating good organizational skills and propulsion."

② Ideal Match Expectation

User queries are highly correlated with the "project leader", "organizational capabilities" and other parts of the document. An ideal system should be able to identify local correlations and improve the ranking of the document.

③Problem Analysis

When using cosine similarity based on TF-IDF vectorization, all words of a document are included in the overall vector representation. In this case, although content such as "development experience" and "code implementation" have high word frequency, they have nothing to do with the query target. Instead, they dilute the weight of project management-related content, resulting in a low overall similarity and making it difficult for documents to be recalled correctly or ranked higher. Therefore, when processing long documents containing multiple topics, a single full-text vector matching is insufficient, and it is necessary to introduce local semantic perception mechanisms (such as fragment-level matching, summary guidance, attention mechanisms, etc.) to enhance the identification and expression of relevant fragments.

How to customize Similarity

Environment preparation

If Python is installed on your computer, please install lazyllm and necessary dependency packages through the following command. For more detailed preparations for the LazyLLM environment, please refer to the corresponding content in Chapter 2: Get started with a minimum usable RAG system in 10 minutes.

pip install lazyllm

Implement the function

Since LazyLLM only provides two similarity calculation methods, BM25 and cosine similarity, it may not meet the requirements in actual applications. This requires you to customize the similarity according to actual needs. Here we use the TF-IDF algorithm to implement an example of similarity calculation to illustrate how to define and apply it.

TF-IDF principles and examples

TF-IDF stands for Term Frequency - Inverse Document Frequency, which is a keyword extraction algorithm widely used in information retrieval and text mining. Its core idea is: if a word appears frequently in a document but is uncommon in all documents, then it is likely to be an important keyword in this document.

1. Term Frequency (TF)

Measure the importance of words in a document:

\[ TF(t, d) = \frac{f(t, d)}{\sum f(w, d)} \]

Among them, f(t,d) represents the number of times word t appears in document d, and the denominator is the sum of the number of occurrences of all words in the document.

2. Inverse Document Frequency (IDF)

Measure the rarity of a word in the corpus:

\[ IDF(t) = \log\left(\frac{N}{1 + df(t)}\right) \quad \text{or} \quad IDF(t) = \log\left(\frac{1 + N}{1 + df(t)}\right) + 1 \]

Among them, N is the total number of documents, and df(t) is the number of documents containing word t.

3. Final calculation formula: TF-IDF

\[ TF\text{-}IDF(t, d) = TF(t, d) \times IDF(t) \]

4. Example analysis

​Documentation Set​:

📄 Document 1: "Big models are changing the world"

📄 Document 2: "Large models are the key technology of artificial intelligence"

📄 Document 3: "I like programming"

​Query​: "Large model technology"

🔹 TF (take document 2 as an example): 7 words in total

"Large model" appears 1 time → TF = 1 / 7 ≈ 0.143

"Technology" appears 1 time → TF = 1 / 7 ≈ 0.143

🔹 ​IDF​（N = 3）

"Large model": df = 2 → IDF = log(3 / (1+2)) = log(1) = 0

"Technology": df = 1 → IDF = log(3 / 2) ≈ 0.405

🔹 TF-IDF (take document 2 as an example)

TF-IDF("Large Model") = 0.143 × 0 = 0

TF-IDF("Technology") = 0.143 × 0.405 ≈ 0.058

❗ Note: In order to avoid IDF being negative, Sklearn uses a smoothing version and does some processing to avoid division by 0 and ensure that IDF ≥ 1.

Although "large model" is a keyword of the query, since it appears frequently in multiple documents, the IDF is 0 and cannot help distinguish the documents. The word "technology" only appears in document 2, and its TF-IDF value is high and contributes to differentiating documents.

Code Example

Code GitHub link🔗

from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity
import numpy as np
from scipy.linalg import norm
import sys
import heapq
from typing import List, Tuple
from lazyllm.tools.rag import DocNode

def tfidf_similarity(query: str, nodes: List[DocNode], **kwargs) -> List[Tuple[DocNode, float]]:
    def add_space(s):
        return ' '.join(list(s))
    corpus = [add_space(node.get_text()) for node in nodes]
    query = add_space(query)
    topk = min(len(nodes), kwargs.get("topk", sys.maxsize))
    cv = TfidfVectorizer(tokenizer=lambda s: s.split())
    tfidf_matrix = cv.fit_transform(corpus+[query])
    query_vec = tfidf_matrix[-1]
    doc_vecs = tfidf_matrix[:-1]
    similairyties = cosine_similarity(query_vec, doc_vecs).flatten()

    indexes = heapq.nlargest(topk, range(len(similairyties)), similairyties.__getitem__)
    results = [(nodes[i], similairyties[i]) for i in indexes]
    return results

In the above code, we first insert spaces into the text in query and node for subsequent tokenizer processing. The tfidf algorithm is implemented by calling TfidfVectorizer in sklearn. After calculating the similarity between query and node, extract the result of topk as output. Now we use the following example for testing.

query = "How is the weather today?"

candidates = [
    DocNode(text="It's sunny today"),
    DocNode(text="The price of the shirt is 100 yuan"),
    DocNode(text="The weather is very good today"),
    DocNode(text="I like eating apples"),
    DocNode(text="The weather is really bad today")
]

results = tfidf_similarity(query, candidates)

Finally, the following output results are obtained.

Register function

In the previous section, we defined a tfidf similarity calculation method. If you want to use this calculation method in LazyLLM, you need to register the method into the LazyLLM framework first. Registration of the similarity calculation method is carried out through the decorator, that is, through the register_similarity decorator function.

Decorator: Decorator is an advanced feature in Python that allows you to add additional functions (such as logging, performance testing, permission verification, etc.) to a function without modifying the original function code. It is essentially a function that "accepts a function as a parameter and returns a new function". You can think of it as a "packaging box for functions" - put a function in it and it will automatically gain new capabilities. The implementation code is as follows.

The code is as follows:

import lazyllm

@lazyllm.tools.rag.register_similarity(mode='text', batch=True)
def tfidf_similarity(query: str, nodes: List[DocNode], **kwargs) -> List[Tuple[DocNode, float]]:
    ...

The register_similarity decorator function has four parameters. Here we introduce them in detail.

• func, which is an optional call object representing the function to be decorated. If not provided, a decorator function will be returned.
• mode, which is an optional literal type with a value of "text" or "embedding", indicating the mode of similarity calculation.
  • "text" means that the similarity calculation method is mainly calculated for text;
  • "embedding" means that the similarity calculation method is mainly calculated for embedding vectors.
• descend, which is a Boolean value that indicates whether the results are sorted in descending order. The default value is True. It should be noted that this parameter does not take effect when calculating the similarity output, but only takes effect when indexing.
• batch, which is also a Boolean value indicating whether batch processing is supported, the default value is False.

The following is a brief demonstration of the three ways to use decorators:

# Usage 1: Use as a function
func1 = lazyllm.tools.rag.register_similarity(tfidf)

# Usage 2: Use as a parameterless decorator
@lazyllm.tools.rag.register_similarity
def tfidf(query: str, nodes: List[DocNode], 
             **kwargs) -> List[Tuple[DocNode, float]]:
    ...

# Usage 3: Use as a decorator with parameters
@lazyllm.tools.rag.register_similarity(mode='text')
def tfidf(query: str, nodes: List[DocNode], 
             **kwargs) -> List[Tuple[DocNode, float]]:

Detailed explanation of parameters

Below we use some small examples to compare the role of each parameter:

1. Parameter func

Regarding the func parameter of the register_similarity function, there are two ways to use it. One is to pass the parameters in through a normal function call, and the other is to register it directly when the function is defined through a decorator. In order to illustrate the difference between passing this parameter and not passing it in, we use function calling to demonstrate below.

func1 = lazyllm.tools.rag.register_similarity(tfidf_similarity)
func2 = lazyllm.tools.rag.register_similarity()
print(f"func1: {func1.__name__}, func2: {func2.__name__}")

The output is:

func1: tfidf_similarity, func2: decorator

Func1 corresponds to the registration function passing in the similarity calculation function, so its return value is the decorated similarity calculation function, while func2 corresponds to the fact that no similarity calculation function was passed in, so its return value is the decorator function.

2. Parameter mode

Mode, an optional literal type, the value is "text" or "embedding", indicating the mode of similarity calculation.

• "text" means that the similarity calculation method is mainly calculated for text;
• "embedding" means that the similarity calculation method is mainly calculated for embedding vectors;

📌 As can be seen from the table, our input, mode and parameter types must be completely consistent before they can be used. But in actual use, you only need to ensure that the node parameter type of the function is consistent with the mode parameter during registration, and there is no need to care about the query type.

Here we first define a calculation function for Euclidean distance:

import lazyllm
from typing import List
import numpy as np

def euclidean_distance(query: List[float], node: List[float], **kwargs) -> float:
    point1 = np.array(query)
    point2 = np.array(node)
    return np.linalg.norm(point1 - point2)

Then register this similarity calculation function, using different modes here:

func1 = lazyllm.tools.rag.register_similarity(euclidean_distance, mode="text")
func2 = lazyllm.tools.rag.register_similarity(euclidean_distance, mode="embedding")

Then we define different queries and nodes to calculate the similarity using different modes of similarity calculation functions to see the effect:

query_t = "hello world."
node_t = [DocNode(text="hello lazyllm.")]
query_e = {"key": [1.0, 0.4, 2.1]}
node_e = [DocNode(embedding={"key": [4.2, 2.1, 3.9]})]

2.1 Use func1 to calculate text similarity

ret = func1(query_t, node_t)
print(f"ret: {ret}")

The output result reported an error:

TypeError: unsupported operand type(s) for -: 'str' and 'DocNode'

2.2 Use func1 to calculate vector similarity

ret = func1(query_e, node_e)
print(f"ret: {ret}")

The output result reported an error:

TypeError: unsupported operand type(s) for -: 'dict' and 'DocNode'

2.3 Use func2 to calculate text similarity

ret = func2(query_t, node_t)
print(f"ret: {ret}")

The output result reported an error:

AssertionError: query must be of dict type, used for similarity calculation.

2.4 Use func2 to calculate vector similarity

ret_2e = func2(query_e, node_e)
print(f"ret_2e: {ret_2e}")

The output is:

ret: {'key': [(<Node id=2865d5c9-730b-4fda-8077-57b706944ad9>, 4.045985664828782)]}

The output results are in line with expectations, because LazyLLM supports multiple embeddings, so the previous key represents which type of embedding, and then the first item Node in the tuple represents which node and query are calculated, and the second item represents the similarity score.

By comparing the above four results, we can clearly understand the role of the mode parameter.

3. Parameter descend

descend, a Boolean value indicating whether the results are sorted in descending order, the default value is True. It should be noted that this parameter does not take effect when calculating the similarity output, but only takes effect during indexing. Because duplication will be removed during output, only the topK nodes are returned, not necessarily in order. Here we can see the effect of this parameter by checking whether the results retrieved by Retriever are similar to query.

We continue to use the Euclidean distance calculation function defined above and register this similarity calculation function. We first use descend as the default value True:

import lazyllm
from typing import List
import numpy as np

@lazyllm.tools.rag.register_similarity(mode="embedding")
def euclidean_distance(query: List[float], node: List[float], **kwargs) -> float:
    point1 = np.array(query)
    point2 = np.array(node)
    return np.linalg.norm(point1 - point2)

Then we use the previous html document for testing. We first define a Retriever object:

import lazyllm
import os
from lazyllm import OnlineEmbeddingModule, SentenceSplitter, Retriever
from lazyllm.tools.rag import Document

prompt = 'You will play the role of an AI Q&A assistant and complete a dialogue task. In this task, you need to provide your answer based on the given context and question.'

documents = Document(dataset_path=os.path.join(os.getcwd(), "rag_data"), embed=OnlineEmbeddingModule(source="glm", embed_model_name="embedding-2"), manager=False)
documents.create_node_group(name="sentences", transform=SentenceSplitter, chunk_size=1024, chunk_overlap=100)
ppl = Retriever(documents, group_name="CoarseChunk", similarity="euclidean_distance", similarity_cut_off=0.003, topk=3)

nodes = ppl("Main contents of the National Housing and Urban-Rural Development Work Conference")
for node in nodes:
    print(f"node: {node.text}")

Here we just pass in the previously registered similarity function name when defining Retriever, and then we select the top 3 results.

The output is:

node: cn/release/blogv2/dist/pc/themesSkin/skin-1024/skin-1024-ecd36efea2.min.css">
    <script src="https://g.csdnimg.cn/lib/jquery/1.12.4/jquery.min.js" type="text/javascript"></script>
    <script src="https://g.csdnimg.cn/lib/jquery-migrate/1.4.1/jquery-migrate.js" type="text/javascript"></script>
    <script type="text/javascript">
        var isCorporate = false;
        var username =  "star_nwe";
        var skinImg = "white";

        var blog_address = "https://blog.csdn.net/star_nwe";
        var currentUserName = "wangjian052163";
        var isOwner = false;
        var loginUrl = "http://passport.csdn.net/account/login?from=https://blog.csdn.net/star_nwe/article/details/141174167";
        var blogUrl = "https://blog.csdn.net/";
        var avatar = "https://profile-avatar.csdnimg.cn/9d615c5dd45743bea262227ce1ce205e_star_nwe.jpg!1";
var articleTitle = "Large model entry to advanced: What is RAG? Why is RAG needed? RAG process";
var articleDesc = "The article has been viewed and read 5k times, liked 22 times, and collected 25 times. Learning large AI models is a systematic process, which requires starting from the basics and gradually deepening into more advanced technologies. Here we have carefully compiled a comprehensive set of AI large model learning resources, including: a full set of AI large model learning roadmaps (from entry to practical), high-quality AI large model learning books and manuals, video tutorials, practical learning and other recorded videos, free to share! _big model rag";
var articleTitles = "Large model entry to advanced: What is RAG? Why do you need RAG? RAG process_Large model rag-CSDN blog";
var nickName = "AI old monkey";
        var articleDetailUrl = "https://blog.csdn.net/star_nwe/article/details/141174167";
        var vipUrlV = "https://mall.csdn.net/vip?vipSource=learningVip";
        if(window.location.host.split('.').length == 3) {
            blog_address = blogUrl + username;
        }
        var skinStatus = "White";
        var blogStaticHost = "https://csdnimg.cn/release/blogv2/"
          var payColumn = false
    </script>
        <meta name="toolbar" content='{"type":"0","fixModel":"1"}'>
    <script src="https://g.csdnimg.cn/??fixed-sidebar/1.1.7/fixed-sidebar.js" type="text/javascript"></script>
      <script src="https://cdn-static-devbit.csdn.net/ai100/chat/chat-search.js?v=2" type="text/javascript"></script>
    <script src='//g.csdnimg.cn/common/csdn-report/report.js' type='text/javascript'></script>
    <link rel="stylesheet" type="text/css" href="https://csdnimg.cn/public/sandalstrap/1.4/css/sandalstrap.min.css">
    <style>
        .MathJax, .MathJax_Message, .MathJax_Preview{
            display: none
        }
    </style>
    <script src="https://dup.baidustatic.com/js/ds.js"></script>
      <script type="text/javascript">
        (function(c,l,a,r,i,t,y){
            c[a]=c[a]||function(){(c[a].q=c[a].q||[]).push(arguments)};
            t=l.createElement(r);t.async=1;t.src="https://www.clarity.ms/tag/"+i;
            y=l.getElementsByTagName(r)[0];y.parentNode.insertBefore(t,y);
        })(window, document, "clarity", "script", "lgtpix6r85");
      </script>
    <script src="/cdn_cgi_bs_captcha/static/js/waf_captcha_embedded_bs.
node: opener.location.href;
                    }
                } catch (e) {}
            }
            if (referrer) {
                $.post(web_root_url + "/api/waplog", {
                    uuid: $("#uuid").val(),
                    refer: referrer,
                    ua: navigator.userAgent
                }, function(e) {
                    //console.log('waplog:' + e);
                });
            } else {
                //$.post(web_root_url + "/api/waplog", {uuid:'1234567890123', refer:'https://www.baidu.com', ua:'user-agent!'}, function(e){
                //console.log('waplog:' + e);
                //});
            }
            //console.log(navigator.userAgent);



//Record sharing behavior
            var urlSearch = window.location.href.substring(window.location.href.indexOf("?")+1) || "";
            let shareuser = "";
            let shareUrl = ""
            let searchArr = urlSearch.split("&")
            searchArr.forEach((item,index) => {
                let ab = item.split("=");
                if(ab[0] === 'shareuser'){
                    shareuser = ab[1];
                    shareUrl = item;
                }
            })
            // console.log('shareUrl1 ',shareUrl)

            if(shareuser){
                $.ajax({
                    type: 'GET',
                    url: web_root_url + '/japi/pv?article_uuid=' + uuid + '&user_uuid=' + shareuser,
                    dataType: 'JSON',
                    success: function(st) {
                        if (st.suc == 1 && st.info == 'success') {
                        }else{
                        }
                    },
                    error: function() {}
                })
            }
        });
    </script>
<!-- Common to the preview page -->
    <script src="https://static.bjnews.com.cn/wap/js/article.js"></script>

    <script src="https://res.wx.qq.com/open/js/jweixin-1.6.0.js"></script>

<!--Share-->
    <script type="text/javascript">
        $.ajax({
            type: "GET",
            url: "https://m.bjnews.com.cn/api/wxshare?app_id=wxbad48ed25d961bb9",
            // url: "https://m.bjnews.com.cn/api/wxshare?app_id=wxbad48ed25d961bb9",
            dataType: 'json',
            success: function(data) {
                var signpackage = data.signpackage;
                wx.config({
                    // debug: true,
appId: "wxbad48ed25d961bb9", // Required, the unique identifier of the official account
// appId: "wxbad48ed25d961bb9", // Required, the unique identifier of the official account
timestamp: signpackage.timestamp, // Required, timestamp to generate signature
nonceStr: signpackage.nonceStr, // Required, generate a random string of signature
signature: signpackage.signature, // Required, signature, see Appendix 1
                    jsApiList: [
                        'checkJsApi',
                        'onMenuShareTimeline',
                        'onMenuShareAppMessage'
                    ],
                    openTagList: ['wx-open-launch-app']
                });
            }
        });
        wx.ready(function() {

            var share_uuid = "1735095611129204";
// Get the url sharing suffix
            var wxurlSearch = window.location.href.substring(window.location.href.indexOf("?")+1) || "";
let wxshareuser = "";//url share information
            if(wxurlSearch){
                let wxsearchArr = wxurlSearch.split("&")
                wxsearchArr.forEach((item,index) => {
                    let ab = item.split("=");
                    if(ab[0] === 'shareuser'){
                        wxshareuser = item;
                    }
                })
            }

// Handle sharing link
            var shareLinkUrl;
            shareLinkUrl = web_root_url + "/detail/" + share_uuid + ".html";
            if (wxshareuser) {
                shareLinkUrl = shareLinkUrl + "?"
node: </span>llm<span class="token punctuation">.</span>llm <span class="token keyword">import</span> ChatSparkLLM<span class="token punctuation">,</span> ChunkPrintHandler
<span class="token keyword">from</span> sparkai<span class="token punctuation">.</span>core<span class="token punctuation">.</span>messages <span class="token keyword">import</span> ChatMessage
<span class="token keyword">from</span> dotenv <span class="token keyword">import</span> load_dotenv


load_dotenv<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token keyword">if</span> __name__ <span class="token operator">&#61;&#61;</span> <span class="token string">&#39;__main__&#39;</span><span class="token punctuation">:</span>
    spark <span class="token operator">&#61;</span> ChatSparkLLM<span class="token punctuation">(</span>
        spark_api_url<span class="token operator">&#61;</span>os<span class="token punctuation">.</span>environ<span class="token punctuation">[</span><span class="token string">&#34;SPARKAI_URL&#34;</span><span class="token punctuation">]</span><span class="token punctuation">,</span>
        spark_app_id<span class="token operator">&#61;</span>os<span class="token punctuation">.</span>environ<span class="token punctuation">[</span><span class="token string">&#34;SPARKAI_APP_ID&#34;</span><span class="token punctuation">]</span><span class="token punctuation">,</span>
        spark_api_key<span class="token operator">&#61;</span>os<span class="token punctuation">.</span>environ<span class="token punctuation">[</span><span class="token string">&#34;SPARKAI_API_KEY&#34;</span><span class="token punctuation">]</span><span class="token punctuation">,</span>
        spark_api_secret<span class="token operator">&#61;</span>os<span class="token punctuation">.</span>environ<span class="token punctuation">[</span><span class="token string">&#34;SPARKAI_API_SECRET&#34;</span><span class="token punctuation">]</span><span class="token punctuation">,</span>
        spark_llm_domain<span class="token operator">&#61;</span>os<span class="token punctuation">.</span>environ<span class="token punctuation">[</span><span class="token string">&#34;SPARKAI_DOMAIN&#34;</span><span class="token punctuation">]</span><span class="token punctuation">,</span>
        streaming<span class="token operator">&#61;</span><span class="token boolean">False</span><span class="token punctuation">,</span>
    <span class="token punctuation">)</span>

    prompt <span class="token operator">&#61;</span> <span class="token string-interpolation"><span class="token string">f&#34;&#34;&#34;
You are an assistant who is good at answering questions. Please use the search content provided below and your own knowledge to answer the question. If you don't know the answer, just say you don't know and don't make up the answer. Please answer in three sentences or less and keep it concise.

Question: </span><span class="token interpolation"><span class="token punctuation">{<!-- --></span>query<span class="token punctuation">}</span></span><span class="token string">

Retrieve content: </span><span class="token interpolation"><span class="token punctuation">{<!-- --></span>retrieval_content<span class="token punctuation">}</span></span><span class="token string">
              &#34;&#34;&#34;</span></span>

    messages <span class="token operator">&#61;</span> <span class="token punctuation">[</span>ChatMessage<span class="token punctuation">(</span>
        role <span class="token operator">&#61;</span> <span class="token string">&#34;user&#34;</span><span class="token punctuation">,</span>
        content <span class="token operator">&#61;</span> prompt
    <span class="token punctuation">)</span><span class="token punctuation">]</span>
    handler <span class="token operator">&#61;

You can see from the output results that the retrieved text has almost nothing to do with the requested query, because the similarity calculation method we use is Euclidean distance. The closer the distance, the more similar the text is. Since it is arranged in descending order, top3 corresponds to the three pieces of text with the largest distance. It is expected that these three pieces of text are not similar to the query.

Next, we set descend to False when registering the similarity calculation function:

import lazyllm
from typing import List
import numpy as np

@lazyllm.tools.rag.register_similarity(mode="embedding", descend=False)
def euclidean_distance(query: List[float], node: List[float], **kwargs) -> float:
    point1 = np.array(query)
    point2 = np.array(node)
    return np.linalg.norm(point1 - point2)

Then we run the above retrieval code again and find that the output result is:

node: <!DOCTYPE html>
    <html data-log-pv='{"mpc":39}'>
    <head>

<title>National Working Conference on Housing and Urban-Rural Development: Vigorously promote the reform of the commercial housing sales system and vigorously and orderly promote the sales_development of existing homes</title>
<meta http-equiv="Cache-Control" content="no-transform" />
<meta http-equiv="Cache-Control" content="no-siteapp" />
<meta name="copyright" content="Copyright © 2017 Sohu.com Inc.All Rights Reserved."/>

<meta name="mediaid" content="Financial Associated Press"/>
<meta property="og:type" content="news"/>
<meta property="og:image" content=""/>
<meta property="og:url" content="www.sohu.com/a/841580226_222256"/>
<meta property="og:release_date" content="2024-12-25 10:27"/>
<meta itemprop="dateUpdate" content="2024-12-25 10:27" />
<meta itemprop="datePublished" content="2024-12-25 10:27" />
<link rel="canonical" href="https://www.sohu.com/a/841580226_222256"/>
<link rel="alternate" media="only screen and(max-width: 640px)" href="m.sohu.com/a/841580226_222256"/>
<meta name="keywords" content="commodity, existing house, housing, system, work, reform, sales, conference, development, vigorously, existing house, housing, work meeting, real estate, financial association" />
<meta name="description" content="Associated Press, December 25, the National Housing and Urban-Rural Construction Work Conference was held in Beijing from the 24th to the 25th. The meeting pointed out that in 2025, we will promote the construction of a new model of real estate development. First, we will focus on optimizing and improving the housing supply system, accelerating the development of affordable housing, and meeting the rigid housing needs of working-class groups with urban housing difficulties..." />
<meta property="og:description" content="China Finance News, December 25, the National Housing and Urban-Rural Construction Work Conference was held in Beijing from the 24th to the 25th. The meeting pointed out that in 2025, we will promote the construction of a new model of real estate development. First, we will focus on optimizing and improving the housing supply system, accelerating the development of affordable housing, and meeting the rigid housing needs of working-class groups with housing difficulties in cities and towns..."/>
<meta property="og:title" content="National Housing and Urban-Rural Construction Work Conference: Vigorously promote the reform of the commercial housing sales system and vigorously and orderly promote the sales_development of existing homes"/>
        <meta charset="utf-8"/><meta name="data-spm" content="smpc"><meta name="renderer" content="webkit"><meta http-equiv="X-UA-Compatible" content="IE=Edge,chrome=1"/><meta name="viewport" content="width=device-width,initial-scale=1,maximum-scale=1"/><link rel="dns-prefetch" href="//statics.itc.cn"><link rel="dns-prefetch" href="//g1.itc.cn"><link rel="dns-prefetch" href="//js.sohu.com"><link rel="icon" href="//statics.itc.cn/web/static/images/pic/sohu-logo/favicon.ico" type="image/x-icon"/><link rel="shortcut icon" href="//statics.itc.cn/web/static/images/pic/sohu-logo/favicon.ico" type="image/x-icon"/><link rel="apple-touch-icon" sizes="57x57" href="//statics.itc.cn/web/static/images/pic/sohu-logo/logo-57.png"/><link rel="apple-touch-icon" sizes="72x72" href="//statics.itc.cn/web/static/images/pic/sohu-logo/logo-72.png"/><link rel="apple-touch-icon" sizes="114x114" href="//statics.itc.cn/web/static/images/pic/sohu-logo/logo-114.png"/><link rel="apple-touch-icon" sizes="144x144" href="//statics.itc.
node: js"></script>        <script>
try {
        var cfgs = {
        channel_id: "39",
        news_id: "841580226",
        cms_id: "$mpNews.cmsId",
        media_id: "222256",
        passport: "pj6PliNNPJ@virtual-author.com",
        weboUrl: "https://mp.sohu.com/profile?xpt=c29odXptdG5hdnVnMDhAc29odS5jb20=",
title: "National Working Conference on Housing and Urban-Rural Development: Vigorously promote the reform of the commercial housing sales system and vigorously and orderly promote the sales of existing homes",
        channel_url: "//house.sohu.com",
        integralLevel: "7",
        categoryId: "-1",
//abData_fd is used for abtest
        abData: "",
// abData_discuss:"4", // Discussion
abData_discuss: "", // Discussion
        abData_fd: "",
        abData_tw: "",
        originalId: "$mpNews.originalId",
        originalStatus: "10",
        isBaiDuAd: "",
        isPure: "${pure}",
        reprint: false,
        reprintSign: "",
        secureScore: '100',
        sGrade: '0',
        editor: '',
        hideAd: '',
keywords: "[commodity, existing house, housing, system, work, reform, sales, conference, development, vigorously, existing house, housing, work meeting, real estate, financial association]",
        mpNewsExt: {
            "modelId": ""
        },
        imgsList: [
                                ],
        topNavigation: [
                                     {
                "url": "http://news.sohu.com/",
"name": "news",
            }
                                    , {
                "url": "http://sports.sohu.com/",
"name": "Sports",
            }
                                    , {
                "url": "http://auto.sohu.com/",
"name": "car",
            }
                                    , {
                "url": "http://www.focus.cn/",
"name": "property",
            }
                                    , {
                "url": "http://travel.sohu.com/",
"name": "tourism",
            }
                                    , {
                "url": "http://learning.sohu.com/",
"name": "education",
            }
                                    , {
                "url": "http://fashion.sohu.com/",
"name": "fashion",
            }
                                    , {
                "url": "http://it.sohu.com/",
"name": "Technology",
            }
                                    , {
                "url": "http://business.sohu.com/",
"name": "Finance",
            }
                                    , {
                "url": "http://yule.sohu.com/",
"name": "Entertainment",
            }
                                    , {
                "url": "http://baobao.sohu.com/",
"name": "Mother and Baby",
            }
                                    , {
                "url": "https://healthnews.sohu.com/",
"name": "Health",
            }
                                    , {
                "url": "http://history.sohu.com/",
"name": "history",
            }
                                    , {
                "url": "http://mil.sohu.com/",
"name": "military",
            }
                                    , {
                "url": "http://chihe.sohu.com/",
"name": "Gourmet",
            }
                                    , {
                "url": "http://cul.sohu.com/",
"name": "culture",
            }
                                    , {
                "url": "http://astro.sohu.com/",
"name": "zodiac sign",
            }
                                    , {
                "url": "https://www.sohu.com/xchannel/TURBd01EQXhPVGt5",
"name": "Topic",
            }
                                    , {
                "url": "http://game.sohu.com/",
"name": "game",
            }
                                    ,
node: The third is to vigorously promote the reform of the commercial housing sales system, vigorously and orderly promote the sales of existing homes, and optimize pre-sales <span>fund supervision</span>. The fourth is to accelerate the establishment of a full life cycle safety management system for houses to provide a strong guarantee for house safety. The fifth is to improve the supervision of the entire real estate process, rectify the order of the real estate market, and effectively safeguard the legitimate rights and interests of the people. <a href="//www.sohu.com/?strategyid=00001 " target="_blank" title="Click to enter Sohu homepage" id="backsohucom" style="white-space: nowrap;"><span class="backword"><i class="backsohu"></i>Return to Sohu to see more</span></a></p> <!-- Government account adding source marking processing -->
<!-- Processing of adding source tags to government accounts -->
<p data-role="editor-name">Editor in charge: <span></span></p>
    </article>
<div id="articleTransfer"><transfer/></div>
<!--Comment ban notification -->
<div id="bannedNotice"><banned-notice/></div>
<div class="statement">Platform statement: The opinions in this article represent only the author's own. Sohu is an information publishing platform, and Sohu only provides information storage space services. </div><div class="bottom-relate-wrap clear type-3">
            <div id="article-like" data-like-type="type-3">
            <article-like/>
        </div>
        <div class="read-wrap">
<span class="read-num">Read (<em data-role="pv" data-val="$articleStat.pv"></em>)</span>
    </div>
    <div id="articleReport">
        <report/>
    </div>
</div>

                            <div id="sohu-play-content"></div>
                        </div>
                    </div>
                    <div data-spm="middle-banner-ad">
                                            </div>
                                            <div id="articleAllsee" style='height:629px'><all-see-list/></div>
                                        <div class="_0u4o3bh76zbp"></div>
                    <div class="god-article-bottom" id="god_bottom_banner" data-spm="ad-text-bottom" style="display:block">
</div>
<div class="user-god clear" id="user-post" style="display:none">

</div> <!-- Comments -->
                    <div id="meComment" style='min-height: 100px;'>
                        <me-comment/>
                    </div>
                    <div id="commentList">
                        <comment-list></comment-list>
                    </div>
                    <div id="discuss"></div>
<!-- Recommended reading -->
                                        <div style='min-height:1500px' id="groomRead">
                        <groom-read/>
                    </div>
                                    </div>
<!-- Right sidebar -->
                <div class="sidebar right" id="right-side-bar" data-a="${isBaiDuAd}">
                    <right-side-bar/>
                </div>
             </div>
            <div id="float-btn"> <float-btn/> </div>
            <div class="left-bottom-float-fullScreenSleepContainer" style="display:none;">
    <div class="left-bottom-float-fullScreenSleep" style="display:none;" data-spm="ad-fullScreenSleep">
        <div class="close-tag"></div>
    </div>
</div>
<div class="left-bottom-float" id="left-bottom-god" data-spm="ad-ss">
</div>        </div>
        <script type="text/javascript">
                        window.deployEnv = "prod"
                    </script>
        <script src="//js.sohu.com/pv.js"></script><script src="https://g1.itc.cn/msfe-pcarti-prod/300000000000/assets/js/vendors-d407b7.js"></script><script src="https://g1.itc.cn/msfe-pcarti-prod/300000000000/assets/js/main_article-64e941.js"></script>        <script>
try {
        var cfgs = {
        channel_id: "39",
        news_id: "841580226",
        cms_id: "$mpNews.cmsId",
        media_id: "222256",
        passport: "pj6PliNNPJ@virtual-author.com",
        weboUrl: "https://mp.sohu.

You can see that the results returned this time are all texts that are very similar to the request query, which is in line with expectations.

4. Parameter batch

Batch, a Boolean value, indicates whether to submit all documents (Node) to the similarity calculation formula at once. The default value is False. Batch is mainly needed when statistical features such as mean and variance need to be calculated, not for parallel computing. LazyLLM has considered parallel computing outside of similarity.

4.1 Euclidean distance calculation function

Here we first take the Euclidean distance calculation function defined above as an example, and register this similarity calculation function. Different batches are used here:

func1 = lazyllm.tools.rag.register_similarity(euclidean_distance, mode="embedding", batch = True)
func2 = lazyllm.tools.rag.register_similarity(euclidean_distance, mode="embedding")

We first calculate the similarity calculation function with batch set to True:

ret = func1(query_e, node_e)
print(f"ret: {ret}")

The output result reported an error:

TypeError: 'numpy.float64' object is not iterable

We then calculate the similarity calculation function with batch set to False:

ret = func2(query_e, node_e)
print(f"ret: {ret}")

The output is:

ret: {'key': [(<Node id=07cce4d6-6d4f-466b-9895-b0d77f607dbd>, 4.045985664828782)]}

Because this function supports non-batch calculations, the results are as expected.

4.2 tfidf similarity calculation function

Next, we use the previously customized tfidf_similarity to see the effect. We also register it first. We use different batches here:

func3 = lazyllm.tools.rag.register_similarity(tfidf_similarity, mode="text", batch=True)
func4 = lazyllm.tools.rag.register_similarity(tfidf_similarity, mode="text")

We first calculate the similarity calculation function with batch set to True:

ret = func3(query_t, node_t)
print(f"ret: {ret}")

The output is:

ret: [(<Node id=4a04a93d-28b4-4dbc-83d5-91150f60a648>, 0.7031197927433698)]

Because the tfidf similarity calculation function defined earlier supports batch calculation, the results are in line with expectations.

We then calculate the similarity calculation function when batch is False:

ret = func4(query_t, node_t)
print(f"ret: {ret}")

The output result reported an error:

TypeError: 'DocNode' object is not iterable

Because this function does not support non-batch calculations, the results are as expected.

Use custom Similarity to build RAG applications

Let’s first build a RAG application using LazyLLM’s default similarity function. Here we still use the html data used previously as an example to demonstrate:

import os
import lazyllm
from lazyllm import pipeline, bind, OnlineEmbeddingModule, SentenceSplitter, Retriever, Reranker
from lazyllm.tools.rag import Document

prompt = 'You will play the role of an AI Q&A assistant and complete a dialogue task. In this task, you need to provide your answer based on the given context and question.'

documents = Document(dataset_path=os.path.join(os.getcwd(), "rag_data"), embed=OnlineEmbeddingModule(source="glm", embed_model_name="embedding-2"), manager=False)
documents.create_node_group(name="sentences", transform=SentenceSplitter, chunk_size=1024, chunk_overlap=100)

with pipeline() as ppl:
    ppl.prl = Retriever(documents, group_name="CoarseChunk", similarity="bm25_chinese", similarity_cut_off=0.003, topk=3)
    ppl.reranker = Reranker("ModuleReranker", model=OnlineEmbeddingModule(type="rerank", source="glm", embed_model_name="rerank"), topk=1, output_format='content', join=True) | bind(query=ppl.input)
    ppl.formatter = (lambda nodes, query: dict(context_str=nodes, query=query)) | bind(query=ppl.input)
    ppl.llm = lazyllm.OnlineChatModule(source="glm", model="glm-4", stream=False).prompt(lazyllm.ChatPrompter(prompt, extra_keys=["context_str"]))
print(ppl("Why RAG is needed"))

The output is:

The reason why RAG (Retrieval Augmented Generation) is needed is that large language models have some inherent limitations. These limitations mainly include:

1. **Timeliness**: The training data set of the large language model is as of a certain point in time, which means that subsequent events, new discoveries, new trends, or data updates will not be included in the model's knowledge base. Since model training requires huge computing resources and time, it is unrealistic to frequently update the model to include the latest information.

2. **Coverage**: Although the training data set of a large model is very large, it may still not cover all domain knowledge or in-depth information in a specific domain. For example, some specialized medical, legal, or technical issues may only be discussed in detail in specific literature that is not included in the model's training data.

When a large model is faced with a problem that is not covered by its training data set, it can produce inaccurate or misleading answers, known as "illusions."

To solve these problems, RAG enhances the performance of large language models through the following process:

- **Retrieval**: Retrieve relevant information from the external knowledge base based on the user's query content.
- **Enhancement**: Integrate the user's query content and retrieved relevant knowledge into a prompt word template.
- **Generate**: Input this enhanced prompt word into a large language model to generate more accurate and targeted answers.

In this way, RAG can make up for the shortcomings of large language models, reduce the generation of misleading information, and improve the credibility and accuracy of answers.

Next, we modify this RAG application to use the previously customized tfidf_similarity method. Because the previous tfidf_similarity function has been registered, it is very simple to use here. We only need to specify the parameter similarity as the name of our customized similarity calculation function when initializing Retriever. The code is as follows:

import lazyllm
import os
from lazyllm import pipeline, bind, OnlineEmbeddingModule, SentenceSplitter, Retriever, Reranker
from lazyllm.tools.rag import Document

prompt = 'You will play the role of an AI Q&A assistant and complete a dialogue task. In this task, you need to provide your answer based on the given context and question.'

documents = Document(dataset_path=os.path.join(os.getcwd(), "rag_data"), embed=OnlineEmbeddingModule(source="glm", embed_model_name="embedding-2"), manager=False)
documents.create_node_group(name="sentences", transform=SentenceSplitter, chunk_size=1024, chunk_overlap=100)

with pipeline() as ppl:
    ppl.prl = Retriever(documents, group_name="CoarseChunk", similarity="tfidf_similarity", similarity_cut_off=0.003, topk=3)
    ppl.reranker = Reranker("ModuleReranker", model=OnlineEmbeddingModule(type="rerank", source="glm", embed_model_name="rerank"), topk=1, output_format='content', join=True) | bind(query=ppl.input)
    ppl.formatter = (lambda nodes, query: dict(context_str=nodes, query=query)) | bind(query=ppl.input)
    ppl.llm = lazyllm.OnlineChatModule(source="glm", model="glm-4", stream=False).prompt(lazyllm.ChatPrompter(prompt, extra_keys=["context_str"]))
print(ppl("Why RAG is needed"))

The output content is:

The reason why RAG (Retrieval Augmented Generation) is needed is that large-scale language models, although powerful, still have some limitations when it comes to generating answers or text. Here are a few specific reasons:

1. **Knowledge update limitations**: The knowledge of a large language model is based on the training data set, which means that its knowledge is limited and may not contain the latest information. By searching external knowledge bases, RAG can introduce the latest information to enhance the generated content.

2. **Context Sensitivity**: Without additional information, large models may not be able to generate the most accurate answers for a specific context. RAG is able to retrieve information related to the current context, making the generated text more accurate and relevant.

3. **Reduce misleading information**: Large models can sometimes generate inaccurate or misleading information. RAG helps reduce the generation of this misleading information by integrating retrieved accurate information.

4. **Improve generation quality**: By combining the retrieved external information, RAG can provide richer context for large models, thereby optimizing the generation results and improving the quality and relevance of the text.

5. **Coping with unknown or sparse data**: For some specific domains or sparse data sets, large models may lack enough training data to generate high-quality output. RAG can help the model handle these situations better by retrieving knowledge in related fields.

To sum up, RAG aims to use retrieval technology to make up for the shortcomings of large language models, improve the accuracy and relevance of generated text, while reducing the risk of misleading information.

Because the tfidf algorithm is not as good as the BM25 algorithm, the final output result is not as good as the above, which is in line with expectations.

Check the Similarity Score

Full GitHub code link🔗

One more thing to add here, how to check the similarity score of the retrieved node? We can actually view it through the similarity_score attribute of node. Here we still take tfidf_similarity as an example, the code is as follows:

import lazyllm
import os
from lazyllm import OnlineEmbeddingModule, SentenceSplitter, Retriever
from lazyllm.tools.rag import Document

prompt = 'You will play the role of an AI Q&A assistant and complete a dialogue task. In this task, you need to provide your answer based on the given context and question.'

documents = Document(dataset_path=os.path.join(os.getcwd(), "rag_data"), embed=OnlineEmbeddingModule(source="glm", embed_model_name="embedding-2"), manager=False)
documents.create_node_group(name="sentences", transform=SentenceSplitter, chunk_size=1024, chunk_overlap=100)

ppl = Retriever(documents, group_name="CoarseChunk", similarity="tfidf_similarity", similarity_cut_off=0.003, topk=3)

nodes = ppl("Main contents of the National Housing and Urban-Rural Development Work Conference")
print(f"nodes: {nodes}")
for node in nodes:
    print(f"node: {node.similarity_score}")

The output is:

nodes: [<Node id=8076e2ef-61c5-46b4-b622-91e9fddf1294>, <Node id=c68319a3-8f34-4cd4-948b-2cb4de22cf0c>, <Node id=7e6c3032-c9b8-4929-bce4-8fbc42417df2>]
node: 0.24727356514169233
node: 0.20530975832300122
node: 0.0970966448783235

You can see that there are three retrieved nodes, the highest score is 0.247, and the other two are 0.205 and 0.097. ** Note: The nodes returned by Retriever are deduplicated, so the similarity scores of the nodes are not necessarily in order. **How ​​can we filter out nodes with lower scores? This requires setting the similarity_cut_off parameter in Retriever. The function of this parameter is to filter out nodes with similarity scores lower than this threshold. This time we set similarity_cut_off in Retriever to 0.21 and take a look at the effect:

nodes: [<Node id=f2c8f5eb-4250-4aa1-ba32-81f0fa4c1017>]
node: 0.24727356514169233

You can see that there is only one node in the output this time, indicating that the similarity_cut_off parameter configuration has taken effect.

[Supplement] In the RAG code above, we can see that the two parameters output_format and join are passed in when Reranker is instantiated. What are they used for? In fact, they can not only be set in Reranker, but also set in Retriever. They are used for post-processing.

Detailed explanation of Retriever parameters

Full GitHub code link🔗

Let’s first look at the values ​​of these two parameters in Retriever. The valid values ​​of output_format are content, dict and None. content means that Retriever retrieves and outputs the content of Node, that is, str type. dict means that the Retriever retrieval output is a dictionary type, that is, the content in Node is converted into a dictionary for output. None means no post-processing is performed and output is directly in the Node type. Valid values ​​for join are Boolean values ​​and string values. When join is False, it will only affect output_format to content, that is, the output content will not be spliced, but will be output in List[str] format. When join is True, join will be assigned an empty string. When join is a string and output_format is content, use join to splice and output the text in nodes, that is, output it in the form of str.

Let's demonstrate it with code below. We take the above code as an example to illustrate, and use the bm25_chinese similarity calculation method provided by LazyLLM, and then use the Chinese studies dataset, [dataset download method].

from lazyllm import OnlineEmbeddingModule, SentenceSplitter, Retriever
from lazyllm.tools.rag import Document

prompt = 'You will play the role of an AI Q&A assistant and complete a dialogue task. In this task, you need to provide your answer based on the given context and question.'

documents = Document(dataset_path="rag_master", embed=OnlineEmbeddingModule(source="glm", embed_model_name="embedding-2"), manager=False)
documents.create_node_group(name="sentences", transform=SentenceSplitter, chunk_size=1024, chunk_overlap=100)

ppl = Retriever(documents, group_name="sentences", similarity="cosine", similarity_cut_off=0.003, topk=3)

nodes = ppl("What is the way of heaven")
print(f"nodes: {nodes}")

The resulting output is:

nodes: [<Node id=e25dbd6c-fb37-4181-88e5-4dc734afe2bf>, <Node id=3c92362f-f553-4325-9eb9-080d4ceb8757>, <Node id=e863a9b9-9962-4592-9c31-62bbab55052f>]

You can see that if nothing is set, the retrieved results are output in Node form.

Next we set output_format to dict:

ppl = Retriever(documents, group_name="sentences", similarity="cosine", similarity_cut_off=0.003, topk=3, output_format="dict")

Running the above code, the output is:

nodes: [{'content': 'Chapter 1\nThe Tao is Tao, but it is very Tao. \nThe name can be named, but it is very famous. \nNothing is called the beginning of heaven and earth; Being is called the mother of all things. \nTherefore, if there is always nothing, I want to observe its wonderfulness; if there is always something, I want to observe its beauty. \nThe two have the same origin but different names. They are both called mysteries, and mysteries are mysteries, and they are the door to all mysteries. \nChapter 2\nEveryone in the world knows that beauty is beautiful, but it is evil; \nEveryone knows that good is good, but it is not good. \nTherefore, presence and absence complement each other, difficulty and ease complement each other, long and short complement each other, high and low complement each other, sounds harmonize with each other, front and rear follow each other, and they are eternal. \nThis is because the sage does nothing and teaches without saying anything. \nEverything is created without beginning, born without being, done without relying on it, and accomplished without living. \nI feel like I'm not living here, so I won't go. \nChapter 3\nIf you don't respect the virtuous, the people will not fight. \nInexpensive and hard-to-find goods prevent people from stealing. \nThere is no desire to be seen, so that the people's hearts are not in chaos. \nThis is the rule of the sage: \nIf the heart is empty, the belly is strong; if the ambition is weak, the bones are strong. \nIt often makes the people ignorant and desireless, and makes people who know it dare not do anything. \nIf you do nothing, everything will be cured. \nChapter 4\nIf you use it when the Tao rushes, it may not make a profit. \nThe abyss is like the origin of all things, and the deep is like Ruexun. \nI don’t know whose son is the first of Emperor Xiang. \nChapter 5\nHeaven and earth are unkind and regard all things as stupid dogs. \nThe saint is unkind and treats the common people as stupid dogs. \nBetween heaven and earth, is it still a mess? \nIf it is empty but not yielding, it will get better when it moves. \nIt is better to stay focused than to talk too much. \nChapter 6\nThe God of Grain is immortal and is called the mysterious female. \nThe gate of Xuan Ni is called the root of heaven and earth. \nIf it persists, it will not be used diligently. \nChapter 7\n forever. \nThe reason why heaven and earth can last forever is because they do not generate themselves, so they can live forever. \nThis means that the saint comes first after his body, and exists outside his body. \nIt’s not because of his selflessness! \nSo it can be done privately. \nChapter 8\nThe best is like water. \nWater is good for all things without fighting.\nIt is what everyone dislikes,\nso it is close to the Tao. \nLiving in a good place, having a good heart, being kind and benevolent, speaking good words and trustworthy, doing good government, doing good things, and doing good things at the right time. \nThe husband does not fight, so there is no need for you. \nChapter 9\nIf you hold on to it and gain it, it will not be as good as it is; if you hold on to it and keep it sharp, you will not be able to maintain it for a long time. \nIf a house is full of gold and jade, no one can guard it; if you are arrogant and rich, you will be blamed for it. \nThe merit is then withdrawn, which is the way of heaven. \nChapter 10\nZai Ying Po hugs one, can we be inseparable. \nSpecialized Qi makes it as soft as a baby. \nIs it possible to eliminate all mysteries and make them flawless? \nIf you love the people and govern the country, you can do nothing. \nThe gate of heaven is open and closed, can it be a female? \nIf you understand the Sida, you can be ignorant. \nChapter 11\nThirty spokes make up one hub. If it is not there, it can be used as a car. \nThe clay is regarded as a utensil, but it should be used as a utensil if it is not there. \nCut out a door to make a room. If it doesn't exist, use it as a room. \nTherefore, if you have it, you will benefit from it, and if you don’t have it, you will use it. ', 'embedding': {'__default__': [0.02805577, 0.023282783, -0.010409373, 0.053345673, 0.054548774, -0.021444315, -0.03477331, 0.009788121,-0.00026711644, -0.019736782, 0.011505052, -0.0037637316, -0.005963491, 0.041383855, 0.040610127, 0.000985921, -0.0166217, 0.025156088, 0.0017116133, -0.004511433, -0.026657365, 0.027576007, 0.01825779, -0.019136382, 0.011662057, 0.009423636, 0.0024921207, 0.0025204297, 0.0037775566, -0.022618348, 0.0013726661, -0.035238307, -0.013865939, 0.001558638, 0.022386922, 0.008083175, 0.07320647, 0.08896838, -0.0036581368, -0.0025714708, -0.021176474, 0.03489965, -0.040128466, -0.038380902,-0.015130355, -0.013884746, -0.015268907, 0.025365353, -0.03942374, 0.019842712, -0.004921817, 0.037063666, 0.049696382, 0.0022446876, -0.021358738, -0.006035082, -0.03891001, -0.0012354126, -0.004749335, 0.04006749, -0.015485865, 0.01656961, -0.0069618057, 0.02573444, -0.01582435, -0.0034912361, -0.025114788, -0.054328755, 0.0043460196, -0.027710024, 0.032266844, 0.033886887, -0.05694017, 0.024670593, -0.042889997, -0.07643343, -0.027408268, -0.0037645432, 0.049862724, 0.034287237,0.0019841655, 0.01078216, -0.0059950864, -0.010718801, 0.035949897, 0.043880112, -0.010869555, 0.0025348524, 0.06355861, -0.012848395, 0.022585338, 0.032618817, 0.044830754, 0.022901794, -0.0006210798, 0.023177462, 0.005330909, -0.046414472, -0.0045889826, -0.014616943, -0.033123583, -0.0512881, -0.0020927102, -0.007836456, -0.048342444, -0.0070576523, -0.013562943, 0.02388182, -0.026922677, -0.026092608, 0.012044963, 0.01646408, 0.04047805, 0.06462115, 0.023438517, -0.016478423, 0.02324674,0.031171313, 0.0362842, 0.018051483, -0.01053648, 0.021465031, 0.023477495, -0.00013451152, -0.011715929, 0.03454124, 0.013441222, -0.046513334, 0.0061332174, -0.0035214778, -0.021454038, 0.05006143, 0.016918018, 0.022030752, 0.007946085, 0.045935158, 0.034973834, 0.00629696, 0.0073370235, 0.025189713, 0.05712071, 0.009955108, 0.050782707, 0.021099582, 0.016486313, -0.050984617, -0.04715113, -0.029727397, 0.04059149, 0.033788107, 0.034945745, 0.05894489, 4.011729e-14, -0.011751112,-0.050978955, 0.05046298, -0.071110204, -0.0019882438, -0.018479101, -0.051272694, -0.010007738, 0.006884054, -0.034583736, 1.1196757e-17, 0.003905397, 0.02047898, 0.035552427, 0.007619795, 0.05529369, -0.020593314, -0.048601326, 0.008128059, 0.0289784, 0.017017765, -0.019920338, -0.028185569, -0.025367916, -0.026009448, -0.027133418, -0.036126122, 0.014420889, -0.0030427186, -0.002751916, -0.022126038, -0.026038717, -0.034733053, -0.0075950925, -0.020967247, -0.020053484, -0.0027370602,-0.037891444, 0.0049866363, -0.04834721, 0.034292456, 0.027533008, 0.018361798, 0.027508227, -0.02045841, -0.017287603, 0.0034947034, -0.043515395, -0.0053127245, -0.043165147, 0.08445005, -0.027922887, -0.016033549, 0.03531948, 0.020594265, 0.03838893, -0.06151098, -0.020862881, -0.0145209795, 2.2020384e-05, -0.0194264, 0.06360542, -0.004401274, 0.017582126, -0.058675535, 0.039020978, 0.05540472, 0.014137382, 0.018975371, -0.027120763, -0.0006975507, -0.0052835485, -0.039460376, 0.0057850983,-0.018254379, 0.056609638, -0.060044613, 0.020291893, 0.006809394, -0.018061927, -0.020031512, -0.022705525, -0.027026203, -0.018345924, 0.016909966, 0.021607758, -0.0012275928, -0.013516523, -0.05039473, -0.051422264, -0.026746254, 0.070365995, -0.0057425452, -0.042636707, 0.04210949, -0.01575858, 0.01585634, -0.014072475, 0.061945125, -0.061828148, -0.09160385, -0.019648878, 0.005205405, 0.04710268, -0.023101546, 0.009841605, 0.046674192, 0.038244978, 0.041187655, 0.06351536, 0.0065300423,-0.018576019, 0.0005083096, 0.0081122685, -0.008605757, -0.0022208246, 0.028158564, 0.02641683, 0.001838783, 0.02509545, 0.0027108253, 0.023963299, -0.041949738, -0.013529197, -0.028406734, 0.012666118, -0.04136234, 0.025941981, 0.029364305, 0.003928735, 0.013956154, -0.029576076, 0.005434449, -0.02822633, 0.033941615, -0.0492418, 0.0077956063, 0.04823387, 0.03536443, 0.0069991727, 0.02820339, 0.016064752, -0.042256713, -0.10030166, 0.008323236, 0.0006843557, 0.022676226, -0.00090373657,0.0070267376, 0.030184641, -0.01695967, -0.026980516, 0.0632163, -0.010148003, 0.019774279, -0.017960021, 0.026519297, -0.045076273, -0.022787727, -0.009274319, 0.02229007, -0.012502117, 0.030040009, -0.02335761, 0.009685071, 0.05651463, -0.024839891, -0.019800588, 0.03620323, -0.013136933, 0.00065234746, -0.010354659, -0.054040577, 0.0154526085, -0.0008120195, 0.0016255731, 0.027276509, 0.08151536, 0.053237963, 0.040713258, 0.030182822, -0.02873101, -0.063819155, -0.01401389, -0.03257143,-0.0130671365, -0.049955063, -0.020377934, -0.015753347, 0.009992964, 0.013277232, 0.010839926, 0.059939686, 0.028921248, -0.037305593, 0.023923561, -0.01083718, 0.041034095, -0.013648627, 0.039294437, -0.0056047714, 0.03269078, 0.030684272, -0.01967473, 0.029130636, -0.06409252, 0.009602843, 0.013003515, -0.0014473207, -0.00059630856, 0.0007928729, 0.0394453, 0.026099138, 0.0013514259, -0.04909594, 0.009835781, 0.0013566319, 0.042203527, -0.028487109, -0.045480847, -0.028581986, 0.023174047,0.026821401, 0.08383023, -0.06158942, -0.028400997, -0.007825939, 0.003234013, -0.057725567, 0.0064291023, 0.012853949, 0.020342965, -0.005471113, 0.032541998, -0.0010940416, -0.008109905, 0.01918718, 0.021351868, -0.014318267, -0.012899077, -0.02170698, -0.0069873175, 0.009628323, -0.024537625, -0.015283444, -0.033228494, -0.0014736207, -0.056704417, -0.0048376825, -0.00083587086, -0.013281995, 0.020264463, -0.049195595, -0.024637522, -0.011718843, -0.021825653, 0.004700579, 0.015963696,0.032105163, 0.0017268145, 0.009395046, 0.03965735, 0.010019313, 0.03593677, 0.0035611999, 0.010699732, 0.052226007, 0.020346584, 0.038028255, -0.0076610018, -0.040621884, 0.011577985, -0.005221753, 0.018597549, 0.031076666, -0.004861098, 0.017033093, 0.039025906, 0.046661437, -0.006155237, -0.012757879, 0.03810979, -0.009047799, -0.030244311, -0.010257482, 0.045990936, 0.02887998, -0.019625623, -0.031448547, -0.01273922, 0.013901697, 0.02527813, -0.03144724, 0.022464182, -0.027144372,-0.02345889, -0.04950639, -0.018502356, 0.017688517, 0.010369425, 0.02123409, -0.028073085, -0.011926072, -0.042963635, -0.058450896, 0.055865563, -0.0059462283, 0.045135874, 0.0077739223, 0.002659368, -0.016431257, -0.02149337, -0.03735157, -0.06444941, -0.0014105599, 0.005779389, 0.078287676, -0.0042932495, -0.021970661, -0.03339254, -0.031541187, 0.016650775, -0.006059932, -0.04497521, 0.02074348, -0.018448783, -0.0050848476, -0.028253563, 0.014440935, 0.0100940475, 0.053356335,-0.083560206, -0.011833814, -0.042367645, 0.036735408, 0.028277006, 0.06732231, 0.014818252, -0.029717922, 0.0019335257, -0.044133835, 0.08157332, -0.037129093, -0.023763556, -0.021009404, 0.024365442, -0.022566712, 0.03615989, 0.019918706, -0.0045654653, -0.0147744175, -0.08768809, -0.027434569, 0.008535524, -0.014643252, 0.04218006, -0.04418376, -0.0063099093, -0.0553463, 0.0038829402, 0.06492348, -0.010223985, -0.02583102, 0.04767369, -0.0031157823, -0.037336517, -0.010527561, 0.012848088,0.012727778, -0.073188215, 0.019639697, -0.0035746093, 0.011461775, -0.011808149, -0.024037987, -0.01767288, -0.017726846, -0.05001298, -0.015326746, -0.015128984, -0.039607592, -0.041593075, -0.016392805, 0.027638663, -0.007096099, 0.012386369, -0.008532201, 0.021859642, 0.008219009, 0.009499561, -0.013660722, -0.011760268, -0.01918902, 0.030194886, -0.032402657, 0.020689184, 0.038485542, 0.040394958, -0.042311437, -0.005354799, -0.0053760037, 0.013059683, -0.011825115, 0.042120803,0.010707755, 0.010666628, 0.064930595, 0.022026783, -0.045309015, -0.013215051, -0.0035855586, -0.018754315, -0.04237147, 0.009189613, -0.001861292, 0.01305605, 0.016130973, 0.016797947, 0.04227604, 0.0010169245, 0.00045087113, 0.039859597, 0.010306915, -0.021312863, 0.013667852, 0.0045401216, 0.041180596, -0.0008014906, -0.016648965, -0.047358204, 0.024641568, 0.063572116, -0.05323623, 0.027804086, -0.0231327, -0.00818245, 0.008933412, -0.02742569, -0.024679849, -0.0070232884, 0.07939336,0.011911154, 0.038045548, 0.01569419, 0.0052608084, 0.01345627, -0.032691434, -0.006347253, 0.04491138, 0.012246293, -0.014546515, 0.0090763355, 0.022790618, -0.036779568, 0.04574143, -0.023112146, 0.039602417, 0.024539718, 0.029363053, -0.057744678, -0.00423374, 0.041941483, -0.018338485, -0.006981724, 0.036857042, -0.0051898616, -0.02998108, -0.025315555, 0.011249301, 0.013763156, -0.0014266107, -0.003557593, 0.029689435, -0.008416172, 0.0055658235, 0.041912172, -0.0066596386, 0.014426449,0.027964776, 0.041610844, 0.030743865, -0.019479757, -0.015880436, 0.013414679, 0.033685517, 0.025084984, 0.0119731445, -0.032747474, -0.016790409, -0.010821385, 0.020628633, -0.019503403, -0.00031055597, -0.038631473, -0.005416922, -0.044260085, 0.009548929, -0.054613594, -0.027896475, -0.028240051, 0.013305551, 0.0026884757, 0.0029414417, -0.04595156, -0.06401382, -0.0026240156, 0.018469386, -0.022476513, 0.011497057, 0.087008506, 0.026063265, -0.069855325, -0.00500086, -0.01826556,-0.011640849, 0.015342398, -0.032531984, 0.033423632, 0.037746735, -0.063191675, 0.009925365, 0.032167457, 0.011919864, -0.024339493, 0.03451287, 0.0016408766, -0.027204601, 0.022536965, -0.0034854359, -0.05221978, -0.002954332, 0.00039458516, -0.037060868, 0.04824446, -0.038343217, -0.008858709, -0.020064302, -0.034424454, -0.0008146635, -0.03229093, 0.020989783, 0.038041, 0.04462357, 0.022627661, 0.021487724, 0.016195275, 0.058926657, 0.0016035914, -0.0781729, -0.01440017, -0.065230064,-0.04473913, -0.05347233, 0.0056564305, 0.026773501, 0.0093747275, -0.0023014166, 0.034559432, -0.014295065, 0.059227344, 0.016397202, -0.010564607, -0.016085293, 0.027223522, -0.039766714, -0.013913507, -0.021724187, 0.019855114, 0.046961125, -0.050933924, 0.028491676, -0.025045706, -0.05075212, 0.034840867, 0.030845476, 0.037544426, -0.04044042, 0.06494801, 0.02402745, 0.027269753, -0.02984972, 0.032292705, -0.0031673112, -0.013952973, -0.024952147, -0.0055473237, 0.020091854, -0.0025514988,-0.030173488, -0.017058272, -0.027787264, 0.025821265, 0.035203192, -0.004649104, -0.0053550736, 0.00918115, -0.033489943, 0.025983026, 0.037477348, -0.01963841, 0.03444502, 0.03194153, -0.06143998, 0.045849923, -0.034311283, 0.042799775, 0.023144597, -0.01235297, -0.0723703, -0.017136108, 0.007224321, -0.0083739795, -0.01820453, -0.011421597, 0.008455473, -0.016129691, 0.014105691, -0.004665744, -0.010970606, 0.032578386, -0.02076877, 0.0136699425, -0.0101013975, -0.032376077, 0.07448448,-0.03354995, 0.035129014, -0.0034162682, -0.019508207, 0.009046866, 0.0062084193, -0.11542149, 0.0007030675, 0.015893739, 0.009404671, 0.003927252, 0.02014505, 0.009974044, -0.028185304, -0.0127715375, 0.04931239, 0.03772544, 0.027806891, 0.030862588, -0.0037065197, -0.04931788, -0.007824991, 3.4293641e-09, -0.0051761563, 0.025381971, -0.008663922, 0.015905118, -0.04079911, -0.019575298, 0.06621208, -0.0035683208, 0.00025738007, 0.04481607, 0.054077428, -0.0072247866, 0.009435948,-0.023140049, -0.05412917, 0.008451867, -0.010966122, -0.034531895, 0.004752035, -0.04602555, -0.0023236184, 0.009561207, -0.023378637, 0.02203002, 0.017096436, 0.03657697, -0.00084620155, 0.004841395, 0.05531498, 0.016193043, 0.012097276, 0.053189192, 0.025982248, -0.03255773, 0.009982884, -0.036482878, -0.0074970718, 0.091324344, 0.0009932049, -0.006143996, 0.0119158905, -0.03613227, 0.047933377, -0.008942387, -0.026436765, 0.0150227435, 0.020003684, 0.07058418, 0.066080704, 0.028907618,-0.044925384, 0.01017655, -0.02693322, -0.03381552, 0.003237652, 0.012430394, -0.014999726, 0.026932973, -0.03142256, -0.023699204, -0.019175638, 0.034598034, 0.0006931077, -0.014283599, -0.008699737, 0.06428559, -0.011042515, 0.038128216, 0.022812221, -0.02596328, 0.0021123243, -0.026327409, -0.0008933892, -0.025974091, 0.03291074, -0.0022326256, 0.008192097, 0.013533811, -0.02715469, 0.0126990145, 0.024964565, 0.05334236, -0.023978073, 0.018099934, -0.01582323, -0.06513108, 0.009603815,-0.042491097, -0.0066699567, -0.007812225, -0.009894219, 0.00089794875, -0.0269737, 0.079909176, -0.043051008, -0.040161613, -0.012302384, -0.009370962, 0.00968951, -0.003516922, 0.02560884, -0.0031629042, -1.8169522e-27, 0.019083612, 0.05595246, 0.08110245, 0.009914496, 0.045291685, 0.077853344, -0.05095552, -0.023066273, -0.013523827, -0.030841304, -0.045739565, 0.03553436, -0.00127252, 0.011472, -0.0025980568, 0.03063463, 0.0018815541, 0.036709815, 0.0418239, 0.0048987665, 0.051705733,0.027615868, 0.024299819, 0.0338089, -0.034141615, 0.0012112331, 0.045954734, -0.050197713, 0.0018632635, 0.038185284, 0.043542426, 0.003177657, 0.02623213, 0.047126044, 0.032685865, 0.018318154, 0.0059990953, 0.014030563, 0.06493271, 0.06270943, -0.034145653, -0.024160989, 0.01995237, -0.082519315, -0.026380774, -0.0008480955, 0.006037592, -0.027352571, 0.00014761026, -0.02092509, 0.024393355, 0.010887946, 0.014051744, -0.050117142, -0.026565975, 0.02288321, 0.0013072105, -0.051732846,0.029837305, 0.008642002, 0.0039808676, 0.03630373, -0.0036453824, 0.04533731, 0.04966945, 0.010970786, 0.010568572, -0.009174333, -0.030876184, 0.02845327, 0.0035904122, -0.004267927, 0.020078829, -0.052125897, -0.047840167, 0.044615194, -0.012507692, -0.03702641, 0.020036895, 0.0070881178, -0.017886484, 0.023605479, -0.01983722, -0.016965086, -0.016553123, 0.057427112, 0.02306789, -0.028244715, -0.00089197716, 0.0067641498, -0.01275369, -0.0065226373, -0.009160864, -0.0067051817,0.021619145, -0.034704223, -0.007742388, -0.019153567, -0.0031101278, 0.023993352, -6.7284605e-24, 0.019229813, 0.036678217, 0.026083732, 0.007746939, 0.005390164, -0.02422693, 0.013378925, 0.0064707967, 0.022271916, 0.0047905026, 0.027713606, -0.059018284, 0.06695722, 0.0046320846, 0.020561788, -0.054861706, -0.013244152, 0.024165925, 0.016604742]}, 'metadata': {}}, {'content':'The way of observing the sky and the journey of holding on to the sky have been completed. \nTherefore, there are five thieves in heaven. Those who see them will prosper. \nThe five thieves are in the heart and execute them in heaven. \nThe universe depends on the hands, and everything depends on the body. \nIt is human nature and scheming. The way of establishing heaven is used to determine people. \nThe sky sends murderous intentions, and the stars move to their places; the earth sends murderous intentions, and dragons and snakes rise from the ground; humans send murderous intentions, and heaven and earth turn over and over again; heaven and man unite, and everything is established. \nSex is cunning and can be hidden. The evil of the nine orifices depends on the three essentials, which can cause movement and stillness. \nFire is born in wood, and disaster will be overcome; treachery is born in a country, and it will collapse in time. Those who know how to practice are called saints. \nHe is born to be killed by nature, which is the principle of Tao. The thief of all things in heaven and earth, the thief of all things by man, the thief of man and all things. The three thieves are both suitable and the three talents are safe. \nTherefore, it is said that if you eat at the right time, everything will be in order; if you use its mechanism, everything will be at peace. People know the gods of their gods, but they don’t know the reason why they are gods. \nThe sun and the moon are numbered, the size is determined, the divine power is born, and the gods are born. \nNo one in the world can see or know about his stolen machine. A gentleman will have a firm bow, while a villain will have a light life. \nThe blind person is good at hearing, and the deaf person is good at seeing. One source of absolute profit, use the teacher ten times. Return to day and night three times, and use your troops ten thousand times. \nThe heart is born from things and dies from things. The machine is in the eyes. \nHeaven is unkind but has great kindness. Thunder and strong wind are all stupid. \nThe most joyful nature, the most tranquil nature and incorruptibility. Heaven is supremely private, and its use is supremely public. \nThe control of birds lies in Qi. The living is the root of death, and the dead is the root of life. Grace is born from harm, harm is born from grace. \nFools regard heaven and earth, culture and science as sages, I regard time, objects, culture and philosophy as sages. ', 'embedding': {'__default__': [0.0041554114,0.0075854566, -0.007057654, 0.045291275, 0.04896368, 0.028331915, -0.05941855, 0.023905145, 0.041470468, -0.048525453, 0.041954085, -0.009402919, -0.026280737, -0.0059798467, 0.03535848, -0.01118226, -0.034457922, 0.05115644, -0.0008446922, -0.026550785, -0.007863325, 0.027618073, -0.011360486, -0.039742425, -0.004911823, -0.04183471, -0.005898696, 0.044848423, -0.030583072, -0.01361686, 0.038467087, -0.04660302, -0.02357292, 0.04048558, 0.03306541, -0.027309675, 0.048068833, 0.088752195,-0.009470059, 0.014505637, 0.0019608818, 0.024321485, -0.0624449, -0.050921675, -0.009048667, -0.014897456, -0.02090078, 0.0161172, -0.031017896, -0.006250185, 0.0034054928, 0.017325701, 0.05056032, 0.007071267, -0.008573053, 0.0057591763, -0.032224294, 0.015077854, -0.021500586, 0.050998785, -0.012885331, -0.0015166812, 0.016822474, 0.010382974, -0.01868565, -0.029895348, -0.035598777, -0.017912406, 0.003902434, -0.019268971, 0.025545066, -0.0016119897, -0.07398274, -0.0090994835,-0.015249873, -0.05385994, 0.041456234, 0.025098544, 0.011212095, 0.023621807, 0.02871689, 0.0155634, -0.017000291, -0.0057249004, 0.041292667, 0.014244264, -0.020487353, 0.009796285, 0.0017768699, 0.0029904752, -0.021676986, 0.050575286, 0.05815488, 0.041090943, 0.03852607, 0.0007469177, 0.0146834655, -0.019205064, -0.02501764, -0.014589938, -0.025496105, 0.0029830346, 0.009922189, -0.014940715, -0.05994815, 0.018909076, -0.0024642693, 0.04501117, -0.047053687, -0.014784821, -0.0020920835,0.035193928, 0.042950764, 0.048697557, -0.015930817, -0.018142601, -0.021126166, 0.0013488104, 0.02504943, -0.007454921, -0.026205262, 0.03224436, 0.06320423, 0.0017354549, -0.013907578, 0.053008184, 0.013806328, -0.040066086, -0.011262496, -0.00783606, -0.032809734, 0.042265926, 0.003031213, -0.011611754, 0.0038381496, 0.026966704, 0.042800177, 0.032915533, 0.016108219, 0.019568527, 0.06046945, -0.010420891, 0.03388435, -0.0009712575, 0.032708555, -0.039891865, -0.029775867, -0.014499864,0.022318674, 0.029443137, 0.010519894, 0.057602942, -2.1037191e-13, -0.040184353, -0.023729598, 0.059133578, -0.060146023, -0.009997531, 0.0071468763, 0.02307546, -0.008445198, 0.025550805, -0.062879466, 1.1355897e-17, 0.005254596, 0.039959684, 0.008519935, 0.028987143, 0.069844805, -0.014054391, -0.035702214, 0.03448102, 0.035850395, -0.026240302, -0.060955, -0.015616291, 0.005105854, -0.03699429, 0.013898516, -0.011044874, -0.008250217, 0.017763525, 0.014568678, 0.011624834, -0.016273065,-0.016237382, -0.030793475, -0.014173828, -0.004470327, -0.00031035003, -0.047902904, -0.033007123, -0.037031222, 0.0365982, 0.030188967, 0.030725434, 0.023961129, -0.02410339, -0.029226728, 0.03464168, -0.024372779, -0.021495208, -0.057038497, 0.073897764, -0.0141333, 0.019894436, 0.05053698, -0.012459984, 0.030188244, -0.0424957, -0.009643966, -0.001958911, 0.020065648, 0.014195304, 0.0558965, 0.01836145, -0.031815484, -0.03150872, 0.0695297, 0.031345744, 0.008421123, 0.0014973548,-0.033643615, -0.014629425, 0.022442328, -0.028167598, 0.015380408, -0.014378052, 0.012282112, -0.053193033, 0.019917091, 0.019215647, -0.033822346, 0.0027193974, -0.014981052, 0.00055290776, -0.004483305, -0.006755365, 0.028678952, 0.023475507, 0.020434262, -0.03490839, -0.027436005, 0.010972408, 0.07965436, 0.0011874165, -0.045413267, 0.028245784, -0.0063281492, 0.009295957, -0.03730253, 0.06491051, -0.045857605, -0.06690101, -0.0045730066, -0.024991188, 0.034862872, -0.015739491, 0.0529046,0.03173905, 0.0703093, -0.0026089184, 0.054319624, 0.007398687, -5.9605067e-05, 0.017072119, -0.008852119, 0.0026614992, 0.038429067, 0.016159521, 0.026893796, -0.035562247, 0.038095932, 0.0065683676, 0.03324841, 0.016334219, -0.023613118, -0.040397674, -0.009194677, -0.008144042, -0.016446576, -0.01151405, -0.010928354, 0.038340032, -0.027145639, -0.046593655, 0.015169281, 0.050768223, -0.03294848, 0.01951992, 0.013797078, 0.035349254, 0.008481241, 0.031466227, -0.003075734, -0.029220875,-0.07521815, -0.0043719853, 0.048672307, 0.0058878683, -0.019860264, -0.03880463, 0.016394598, -0.0122021, -1.4963505e-05, 0.044320922, 0.007890018, 0.0054832306, -0.0123292245, 0.03318079, -0.015232731, -0.013402485, -0.02389836, 0.0055185272, -0.028106516, 0.022771437, -0.030289527, 0.02677829, 0.083992176, -0.012977583, -0.020257147, 0.07025292, -0.0054380586, 0.00113687, -0.003930704, -0.059119448, 0.0005393099, 0.006615847, -0.008725304, -0.011702107, 0.07533191, 0.028773967, 0.029288677,0.048617944, -0.012453482, -0.02434141, -0.026455658, -0.038755085, 0.010074069, -0.043186586, -0.008087229, -0.030695459, 0.0028579491, 0.010711575, -0.027886447, 0.024541127, 0.011004065, -0.06523756, 0.043040413, 0.025965191, 0.013071675, 0.023030335, 0.029349305, 0.021783385, 0.018128468, 0.006819187, 0.006515866, 0.029666856, -0.06356491, 0.039198253, -0.011071719, 0.023561964, -0.03198062, 0.023516942, 0.059817895, 0.004185616, -0.008066767, -0.009041066, 0.015803682, 0.0067058923,0.041337356, 0.0052649537, -0.027125185, -0.038213175, 0.014553458, 0.04226332, 0.070982724, -0.11455366, -0.032585356, 0.0102267265, 0.011878967, -0.033385586, 0.0010560753, -0.015395409, -0.00441643, -0.01337264, -0.011390325, 0.03452036, -0.0043478706, 0.021369405, 0.022401562, -0.022205178, 0.01836901, -0.040310632, -0.014155401, 0.022052731, -0.033351872, 0.023539973, 0.0066144154, -0.02636437, -0.049085867, 0.021261329, 0.025078895, -0.03848087, 0.031188102, -0.048261605, 0.019504659,-0.005102396, -0.01836231, 0.02110646, 0.025233712, 0.031062676, 0.0010754078, -0.017802356, -0.013914974, 0.014522387, 0.056926217, 0.0053230077, 0.028501919, 0.048797324, 0.023192475, 0.027942462, 0.0073998393, -0.03499081, 0.022158591, -0.019583017, 0.005857129, 0.046547387, -0.038932085, 0.040460087, 0.042364065, 0.0429083, 0.009138595, -0.012183861, 0.058576614, -0.030963775, -0.034653503, -0.025639888, -0.011284233, 0.047425315, -0.029514484, -0.036847457, -0.019771382, 0.008259031,0.022998115, -0.017095372, 0.0066413945, -0.05487991, 0.002815856, -0.033681728, -0.014198309, 0.026634654, -0.00822199, -0.00530929, -0.011223847, -0.011880681, -0.034348324, -0.04212451, 0.067977875, -0.034103774, 0.05572993, -0.0062887007, 0.010488381, -0.020459691, 0.012848595, -0.025240324, -0.06849307, 0.025805129, -0.011156502, 0.05313631, -0.024320655, -0.028916016, 0.010576907, -0.0015999046, 0.024717461, -0.0144396275, -0.055715647, 0.030035136, 0.05483442, 0.010892865,-0.0075423573, 0.03750252, -0.005360984, 0.061075144, -0.05138248, -0.018550297, -0.06849128, 0.03832335, 0.051121823, 0.07625229, -0.02237382, -0.02532701, 0.005352677, -0.008111206, 0.073260285, -0.050452314, -0.03776045, -0.049333394, 0.044611394, -0.026005713, 0.039077275, 0.0220915, -0.029445095, -0.0064717205, -0.027871978, -0.035875883, 0.0024223006, 0.018516455, 0.05704868, -0.03898351, 0.006290837, -0.06967245, -0.021957148, 0.030887723, 0.019454552, -0.016694501, 0.013290308,-0.0031141401, -0.013080929, -0.016133493, -0.014472164, -0.01712892, -0.037303273, 0.027202567, -0.020992804, 0.042868, 0.013083941, -0.0220378, 0.004116772, -0.014082884, -0.03672434, -0.025468608, -0.004967429, -0.043549094, -0.059921604, -0.0267599, 0.020867042, -0.024328588, -0.009694849, -0.008663098, 0.029830154, 0.015654892, -0.01019365, -0.005135016, -0.005788498, 0.005652381, 0.018197127, -0.048258357, 0.018697621, 0.009251598, 0.013933649, -0.05764918, 0.017524255, -0.055997822,0.00016845911, -0.02751447, 0.032793835, 0.0019559327, -0.0075224875, -0.01907267, -0.008904949, -0.030548248, 0.013870509, 0.0069685555, -0.028905436, -0.02268364, -0.020575576, -0.0030170542, 0.015130529, -0.00084239285, -0.029587327, 0.037500374, -0.027868606, -0.006660683, 0.031205097, 0.012189993, -0.000101473976, 0.012176301, -0.012304269, 0.016419522, -0.031622086, -0.031915437, -0.07447143, 0.044259857, 0.034456633, -0.030144012, 0.006155522, -0.047044933, -0.053129446, -0.003883295,-0.02238673, -0.036794078, 0.018963076, 0.038487703, 0.043547265, 0.020530019, 0.013525153, 0.03025931, -0.012935889, 0.0006987865, 0.001170972, 0.049213756, -0.021247994, -0.010136263, -0.008933942, 0.040884793, -0.014472714, 0.04323616, -0.024579003, 0.044722065, -0.00044178023, 0.022410542, -0.043617256, 0.022241985, 0.04859234, -0.034833476, 0.013860875, 0.023002412, 0.00644085, -0.01903299, -0.030982777, -0.02251995, 0.025486361, 0.013646234, -0.025401177, 0.028737532, -0.0021397755,0.005839515, 0.031907808, -0.017933525, 0.0054767667, 0.011915933, 0.04648308, -0.0016207868, -0.02563286, -0.044512413, 0.03895756, 0.026000049, 0.028605616, 0.0051885676, 0.0020558056, 0.0013344906, 0.018722871, 0.01694392, -0.048954893, -0.0022045183, -0.04172797, -0.021244593, 0.013639434, 0.0023175476, -0.057096064, -0.017740823, -0.041350298, -0.0043968866, -0.03799386, -0.0042910026, -0.09530829, -0.038248345, -0.030624647, -0.0110857375, -0.010244377, 0.046464287, 0.11650956,0.015119723, -0.069281384, 0.01305812, 0.012854464, -0.0055132634, 0.018360458, -0.0165773, -0.0032716265, 0.031324998, -0.047736414, 0.014519401, 0.048135165, -0.035682708, -0.0365996, 0.03389063, -0.0043802285, -0.01877874, 0.04836236, 0.020456195, -0.056787007, 0.010132782, 0.0004068694, -0.047474485, 0.03445135, -0.02724392, -0.03328944, -0.0029203526, -0.026272407, -0.00029890105, 0.013430369, 0.013834825, 0.060670216, 0.012882629, 0.06432347, 0.015768427, -0.010093849, 0.024928765,0.012602784, -0.064383924, -0.0050611207, -0.044316787, -0.041941397, -0.04322314, 0.006486478, 0.0071116807, -0.012460606, -0.0074164756, 0.029274957, -0.024530469, 0.018598046, 0.03159384, -0.001951544, 0.00523427, 0.010126147, -0.017179206, -0.018108787, -0.007169589, 0.009502527, 0.008932048, -0.069397494, 0.044141974, -0.005750729, -0.045750074, 0.023508383, 0.03368248, 0.027255893, -0.015447886, 0.056365486, 0.008277988, 0.008696923, -0.059466016, -0.0071224286, 0.020547476, 0.028926386,-0.031088999, 0.0034343835, -0.009711087, 0.0011468836, -0.013077483, -0.025331382, -0.016229326, 0.02545839, 0.011994319, 0.008591205, -0.0015080342, 0.02231805, -0.04337707, 0.040335864, 0.026538009, -0.0062174527, 0.050592676, 0.029670224, -0.013418461, 0.0727519, -0.037769295, 0.04368287, 0.03542585, -0.020641856, -0.05900921, -0.027732523, 0.0072757383, -0.019184208, -0.013169488, 0.022744527, -0.010796951, 0.013560294, -0.035595182, 0.023965288, 0.010551866, 0.024181299, -0.03136934,-0.044932388, -0.020812577, -0.054782774, 0.04152298, -0.011101999, 0.024159305, 0.019186165, -0.062283333, 0.042109832, -0.0031904012, -0.056608956, 0.030812602, 0.020276703, 0.03021552, -0.039611798, 0.03263928, -0.008093682, -0.029934445, -0.003328774, 0.04017525, 0.05742037, 0.065777995, 0.011823178, -0.022649916, -0.030095868, -0.029602733, 3.6471364e-09, 0.019888746, 0.042886823, -0.013049726, 0.037731685, -0.036016267, -0.046768334, 0.060499705, 0.00012572506, 0.016042653, 0.024610577,0.033402827, 0.004141988, -0.0006018001, -0.055084515, -0.039775707, 0.008617055, -0.01653752, -0.013155907, 0.016346993, -0.023589248, 0.024627982, -0.012017998, -0.053784538, 0.037820525, 0.03518804, 0.046451963, 0.016566351, 0.019617809, 0.03032722, 0.051768657, 0.003973617, 0.011890001, 0.030245746, -0.009497901, -0.001179198, -0.026856216, 0.015133692, 0.04234234, 0.007812807, -0.02503269, 0.017157003, -0.061224688, 0.04974642, 0.033100616, -0.0151869515, 0.024570305, 0.018643305,0.06040034, 0.038901594, 0.04535261, -0.04548045, 0.0020181928, -0.034227967, -0.0013509532, -0.030452356, -0.018487962, -0.023855759, 0.05625316, -0.00090770074, -0.00640004, -0.014777776, 0.027941104, 0.016304482, -0.038740773, 0.011337658, 0.07173977, -0.03550307, 0.023911772, -0.0016779553, -0.033899304, -0.022336006, -0.015419301, 0.001863339, 0.0008760349, 0.009762301, -0.028453615, -0.011362094, 0.04275605, -0.043521807, 0.029829519, -0.02416117, 0.0335993, -0.006421268, 0.04060249,-0.032474253, -0.07600941, 0.004471598, -0.04126713, -0.02453735, 0.011764305, 0.015372352, 0.0028955622, -0.0291478, 0.05877516, -0.009823992, -0.054267995, 0.0067844344, 0.03217804, -0.011592191, -0.00035714993, 0.028778687, 0.0087011475, -1.7158678e-27, 0.0017932786, 0.03569676, 0.054138917, -0.008628211, 0.043432187, 0.023903232, -0.0067582014, -0.03050816, -0.028248005, -0.02057966, -0.057666212, 0.031576764, -0.0151130445, -0.027907273, 0.019604214, 0.019528171, -0.030740337,0.013142259, 0.04407093, -0.0029533103, 0.0063910317, 0.017176246, 0.032132074, 0.022815445, -0.050552312, -0.020800956, 0.018182818, -0.037349436, 0.014202859, -0.00087027677, 0.058879256, 0.0056527536, -0.005539146, 0.032238558, 0.022045877, 0.03943093, 0.0077616903, 0.02564012, 0.06469182, 0.020434259, -0.028960383, -0.01953917, 0.016642196, -0.04865382, -0.009785435, 0.00070976716, 0.005543073, -0.00053107564, 0.019826816, -0.046817478, 0.02328427, 0.0038257786, 0.041383106, -0.049317617,-0.012191691, 0.040143386, -0.020945452, -0.015941637, 0.014165448, -0.014123277, 0.0048813364, 0.009677354, -0.001581121, 0.03251474, 0.042997185, -0.015337093, -0.021302974, -0.03482498, -0.02814613, 0.014423916, 0.030641912, -0.0033474728, 0.019615268, -0.046341933, -0.007550764, 0.014528103, -0.024804426, -0.04388465, -0.00038302864, 0.018112512, 0.016945953, 0.022862429, -0.06407195, 0.02999205, -0.029377965, 0.038923685, -0.003280456, -0.0140948305, -0.016372854, -0.0024543288,-0.019321902, 0.029282054, -0.0075866296, -0.018350568, 0.014755779, -0.050756894, -0.038549382, -0.026098477, 0.0016459071, 0.023259882, -6.1448135e-24, -0.022952693, 0.022025546, 0.014744688, 0.009926959, 0.044778008, 0.018253125, 0.04506746, -0.019753184, 0.046412323, 0.00015234029, 0.060078166, -0.060511682, 0.055275027, 0.02107507, 0.0074826907, -0.026740791, -0.012698739, 0.017056648, 0.050418142]}, 'metadata': {}}, {'content':'Destiny is called nature, nature is called Tao, and cultivating the Tao is called teaching. \nThe Tao cannot be separated in a moment (yú); it can be separated, but it is not the Tao. Therefore, a gentleman should be cautious about what he does not see and be afraid of what he does not hear. Don't see (xian) what is hidden, don't show what is subtle, so a gentleman should be cautious about being alone. \nWhen joy, anger, sorrow, and joy have not arisen, it is called the middle; when they arise, they are all in the middle, it is called the harmony. The one who is in the middle is the foundation of the world; the one who is in harmony is the way of the world. To achieve neutrality, the heaven will have its place and all things will be nurtured. \n\nZhongni said: "A gentleman is the mean, and a villain is against the mean. A gentleman is in the mean, and a gentleman is in the middle at all times; a villain is against the mean, and a villain has no scruples."\n\nConfucius said: "The mean is the end! People rarely last long. ! "\n\nConfucius said: "I know that the way is not good: those who know it can't do it, and the fool can't do it. I know that the way can't be done; the wise person can't do it, and the unworthy person can't do it."', 'embedding': {'__default__': [0.025426213, 0.034025285, -0.031320818, 0.016555943, 0.06301127, 0.0053874906, -0.03851282, 0.05617112, -0.026647208, -0.04409193, 0.07942905, -0.016034704, 0.0020301498,0.009465688, 0.022152705, -0.02026594, -0.05420044, 0.048457634, 0.029878676, -0.023549352, 0.016439304, 0.034842428, 0.020412622, -0.0025755076, 0.0027205113, -0.022589676, 0.020291883, 0.03922448, 0.009154288, -0.007037486, 0.018945666, -0.038242333, -0.0059177317, 0.035270065, 0.023194453, -0.024663404, 0.045518205, 0.078890085, -0.0049626785, -0.0013358546, -0.020847179, 0.008833306, -0.057070132, -0.046780605, 0.0070816474, 0.009134139, -0.019501934, -0.034442246, -0.03516072,0.006713291, -0.006998874, 0.032105204, 0.052466325, 0.02138813, -0.009691537, -0.009268791, -0.04773306, -0.01108876, -0.03728646, 0.06281455, -0.015532677, 0.012268486, 0.005829601, -0.0013577868, 0.006930408, -0.05296581, -0.023193905, -0.046643063, 0.011809876, 0.00023975057, 0.050946303, -0.03136553, -0.06390407, 0.03504418, -0.01805315, -0.04344443, 0.00677673, 0.013735469, 0.045064267, 0.029002164, 0.03938319, 0.034683738, 0.006368947, -0.00906264, 0.035292383, 0.058860265,-0.028741615, -0.0036960396, 0.012889149, -0.008738161, -0.012072681, 0.044848487, 0.047913834, 0.0025441165, 0.033541866, 0.0077599497, 0.025643757, -0.016167128, 0.02875357, -0.012898535, -0.03224147, -0.021762878, -0.015384127, -0.022707786, -0.0061772997, -0.020073174, 0.02529043, 0.014178167, -0.018244589, -0.008239636, 0.02126484, 0.0034560761, 0.051880326, 0.06363023, 0.01969195, -0.011307083, 0.005372883, -0.019629491, 0.0017945681, 0.011739278, -0.023668867, 0.009739918, 0.059829958,0.034342274, -0.0019169106, 0.03975126, 0.0023158423, -0.044209726, -0.0027910431, 0.00031879256, -0.040849816, 0.04776571, 0.0020381147, 0.01910589, 0.0027710563, 0.011146467, 0.020007843, 0.0042484123, 0.005176938, 0.021434778, 0.0777557, -0.0038533013, 0.047306042, 0.0034837353, 0.042881086, -0.06315707, -0.052171838, -0.04829033, 0.019041482, 0.031125337, 0.045959365, 0.04621173, -1.540408e-13, -0.006997954, -0.020921359, 0.037427846, -0.07051564, 0.00041891588, 0.026310083, 0.02499563,0.0017974637, 0.002034413, -0.023606593, 9.611427e-18, -0.009685626, -0.019306758, 0.033706345, 0.0356712, 0.039736345, -0.012066647, -0.0059178593, 0.031503696, 0.023679575, -0.011107136, -0.048489127, 0.0053288713, -0.0123187, -0.049243163, 0.00041808805, 0.0003394062, 0.015456655, -0.0028465304, 0.014629242, -0.011859338, -0.029523129, -0.026531836, -0.004875002, -0.009635127, 0.007637854, 0.016192164, -0.027906612, 0.0043651653, -0.011299037, 0.009702859, 0.020838758, 0.0027266808,0.013553497, -0.020599183, -0.052582372, 0.047934312, -0.0211804, -0.014517992, -0.06661431, 0.057032656, -0.048424397, 0.011499752, 0.0056307977, -0.016409324, 0.044111975, -0.04115124, -0.026549118, 0.01797299, 0.022250125, -0.020612499, 0.060236566, -0.006571854, -0.00393553, -0.031184353, 0.064215384, 0.046786495, -0.00046514318, -0.00701734, -0.02017527, 0.008758273, 0.0016055728, -0.02660426, 0.009220425, -0.014724887, 0.010843108, -0.044277404, 0.010815744, 0.034662224, -0.026707007,-0.005213057, -0.011781672, -0.019246181, -0.019320516, -0.0016598016, 0.005209462, 0.028500574, 0.013622352, -0.026014416, -0.03194754, -0.010626347, 0.070779, -0.02930651, -0.049735054, 0.019265305, -0.017812809, -0.03864494, -0.057057705, 0.047812644, -0.021030407, -0.06017914, -0.008436314, -0.00565766, 0.040196173, -0.0015437013, 0.03359702, 0.07130042, 0.019466452, 0.030842118, 0.030507375, -0.007607868, -0.014626105, -0.011076136, -2.4094108e-05, -0.0024054602, 0.007530281, 0.029631741,0.031458937, -0.006896458, 0.014952638, -0.0036152329, 0.027877389, 0.016777638, -0.021922763, -0.03586302, -0.0010009537, -0.016428495, 0.008924784, 0.023603037, -0.020115128, 0.00865001, -0.026730474, -0.033528563, 0.013149746, 0.030888822, -0.005040427, -0.0014151632, 0.020947043, 0.028981369, 0.00765713, 0.032147344, -0.023203162, -0.0044152536, -0.045422826, 0.016876237, 0.036611527, 0.013162975, 0.026166268, -0.027787978, 0.034219008, -0.013293831, -0.019005, 0.055061992, 0.00763016,0.032072656, 0.014239542, 0.032104347, -0.037449952, -0.02217896, -0.032320846, 0.017871961, -0.03970623, 0.026835883, -0.054085482, -0.0027009265, 0.07634273, -0.024354596, -0.0029545398, 0.059289258, -0.027541274, 0.019158483, -0.02926486, -0.059198033, -0.00045453524, 0.03309172, -0.058500547, 0.03847318, 0.06686244, 0.0359078, 0.037715927, 0.026004314, -0.010441379, -0.044468235, -0.04735834, -0.054638885, -0.003099882, -0.016880812, 0.00814928, -0.010210958, -0.022301093, 0.0029216122,-0.012461757, 0.030258488, 0.00953299, -0.03189832, 0.015131103, 0.0012498442, 0.014835826, 0.05411683, 0.020760238, 0.017031388, 0.020908138, 0.059172362, 0.033446126, -0.0024910853, -0.0605189, -0.0029194264, -0.001088384, -0.042831995, 0.008137215, 0.020813528, 0.037613075, -0.005095976, -0.01884459, -0.005923714, 0.023792582, -0.028474923, 0.03757395, 0.024344636, -0.035215013, -0.023457121, 0.052875917, 0.026036683, 0.10764514, -0.051340614, -0.018821908, 0.016108429, 0.024424875,-0.053446505, -0.0054197996, -0.0060087196, -0.0041229264, -0.006034636, 0.007538079, -0.018776668, -0.013380194, 0.0028277775, -0.0048381737, 0.004624616, -0.030349467, -0.04140086, 0.009394284, -0.008468681, -0.036008462, -0.011696614, -0.020296633, -0.030620858, -0.063259475, 0.038780067, -0.010312165, -0.053780414, -0.011514602, -0.04333725, -0.010768698, 0.0016200787, -0.029752402, 0.007386246, 0.018586209, 0.009053951, 0.00036854058, 0.022906242, 0.028303033, 0.019804642, 0.036445,0.005007112, 0.014926994, 0.029018547, -0.015011582, 0.0065884907, 0.029475305, -0.016382167, 0.007041045, -0.024008943, 0.046547353, 0.027779775, -0.0071733906, 0.012406345, 0.011313683, 0.03918696, -0.056599338, -0.034151986, 0.047533363, -0.01459687, -0.0679629, -0.011175669, 0.030152632, 0.0382269, 0.0041881935, -0.007858538, -0.018000403, 0.022335991, -0.003267302, -0.048622385, 0.023882171, -0.034625404, -0.035592694, -0.042968493, -0.004153031, 0.019842803, -0.0016533697, -0.004881279,-0.045086186, -0.0011175375, 0.0021369208, -0.027217858, 0.042248532, -0.030293925, 0.064989135, -0.029612714, 0.024552464, -0.033408698, -0.0101805655, -0.0029412394, -0.0915347, 0.04138679, 0.013030276, 0.06705817, -0.004188915, -0.0062264856, -0.051195238, -0.02774683, 0.00731135, -0.013503835, -0.052448288, 0.021496458, 0.019841682, -0.023670819, 0.009778803, -0.0014320388, 0.014062996, 0.07120368, -0.04988873, -0.00012608718, -0.04031885, 0.018206028, 0.0640813, 0.070187114,-0.0015529695, 0.020645538, 0.0052610333, -0.043595728, 0.061265957, -0.042493634, -0.04516448, -0.005089449, 0.010595472, -0.015604816, 0.03951871, 0.03183545, -0.016843012, 0.0235672, -0.040382978, -0.03078044, 0.014075101, -0.0014767948, 0.042445626, 0.026724413, 0.017799338, -0.08596669, -0.018500224, 0.051851522, -0.009828736, -0.029204156, 0.011474476, 0.0055204323, -0.004064649, -0.033348784, 0.00523749, 0.022435933, -0.046652254, 0.024559079, -0.016480578, 0.040265355, -0.03305009,-0.031315356, -0.010046829, -0.01731933, -0.020532912, -0.0029909958, 0.011842756, -0.048824314, -0.012646131, -0.055367026, 0.0031566594, 0.014152518, -0.01904883, -0.024933103, 0.04189374, 0.028184457, 0.025812266, 0.02407379, 0.022633042, -0.007829883, -0.014562866, -0.032446302, 0.008651778, 0.026188472, 0.04299915, -0.02544324, 0.021127068, -0.018766753, -0.0063632224, 0.025388861, 0.0063838237, -0.02366371, -0.017125644, 0.0014911906, -0.022412458, 0.009111594, -0.008172784,-0.008471129, -0.023930583, -0.05943837, 0.00966703, -0.006407053, 0.012494917, -0.012754937, -0.026357513, 0.0407052, -0.024343831, -0.0037548298, 0.029267209, -0.008125634, -0.010630878, 0.02786125, -0.007368123, 0.0062804264, -0.032700405, -0.020568395, -0.043276288, 0.015865225, 0.0311623, -0.043471303, 0.055800557, -0.011932295, -0.045096148, -0.0050109215, -0.011723784, -0.005000813, 0.003097428, 0.0811478, 0.040164344, 0.0036458236, 0.012483547, 0.010060236, 0.0023018436, 0.006584019,0.024982, 0.03525057, 0.0041822596, -0.015556074, 0.0034614399, 0.038330406, -0.00889707, 0.03221703, 0.03959702, 6.4116706e-05, 0.0036850327, 0.038901724, -0.028086498, 0.010405911, 0.021698542, -0.015949449, 0.005926045, 0.033439852, 0.0038825811, -0.025638554, -0.023800563, 0.0133479, 0.0346594, -0.00953549, -0.047149554, 0.024543112, 0.017636258, -0.013193999, 0.017017592, 0.028411236, -0.0025103707, 0.018269507, 0.051845845, 0.062203504, -0.025815979, -0.051945943, 0.0033721116,0.07998744, 0.025422625, 0.009253838, -0.03604736, -0.042337134, 0.014561826, 0.04507746, -0.020416573, -0.015352677, -0.036153104, -0.027990274, 0.01924466, -0.045471456, -0.008539538, -0.040453684, -0.05723972, -0.01498072, -0.030530391, -0.007754218, -0.039093077, -0.03979653, -0.02228624, 0.0008455001, 0.0071356776, 0.025943242, 0.11981686, 0.014268126, -0.05869411, 0.008343074, -0.040158387, -0.016552536, 0.018982463, -0.03898772, -0.007980556, 0.022687193, -0.051658224, 0.025367606,0.01998329, 0.021392373, -0.0434203, 0.02121578, 0.018556163, 0.003245211, 0.018774172, 0.0005500793, -0.03479759, 0.035857406, 0.00040770217, -0.035999756, 0.028827438, -0.029466469, -0.047930814, 0.0265652, -0.05008721, -0.0010059318, -0.01251698, 0.003447119, 0.05898002, 0.05744354, 0.06840657, 0.007024658, 0.006566852, 0.030507237, -0.041550875, -0.04153647, 0.0010598813, -0.0794358, -0.041427277, -0.06114106, -0.019902522, -0.012520727, -0.01163048, 0.0007943832, 0.028543016,-0.009956491, 0.025655229, 0.043709233, 0.0011159213, -0.030077763, 0.039286032, -0.03297622, 0.0095170215, -0.039622717, 0.002401525, 0.021634275, -0.057686023, 0.04041061, -0.03395288, -0.04493925, 0.028587993, 0.046589497, 0.03245823, -0.02444273, 0.036353294, 0.008618879, 0.0035412086, -0.046254568, 0.0148467375, 0.016634878, 0.0306935, 0.011313652, -0.031087654, 0.009768224, 0.007366226, -0.024521042, 0.013069051, -0.036803946, 0.045954864, 0.00825386, 0.013556772, -0.028380545,0.01906237, -0.017373545, 0.02040689, -0.004991811, -0.011323219, 0.038479466, -0.010787252, -0.019826626, 0.06550503, -0.021512486, 0.04311003, 0.06399184, 0.01899978, -0.06393204, -0.052857753, 0.0026758595, -0.023297222, 0.0063398667, 0.029147884, 0.0076334844, 0.029746521, -0.016748957, -0.0021633625, 0.021256661, 0.022264387, -0.022884088, -0.016388448, -0.033683803, -0.02238929, 0.055876125, -0.0052925227, 0.013481841, -0.0209584, -0.03506728, 0.016223963, 0.0012435996, -0.046723302,-0.0005387966, 0.012555997, 0.04083436, -0.012232332, -0.016081795, 0.0062284134, -0.027968781, -0.022376344, 0.02119133, 0.034631502, 0.02318447, 0.044668313, 0.0023679107, -0.04467385, -0.01789442, 2.5628348e-09, -0.010842855, 0.026404202, 0.018929183, -0.008009062, -0.03902315, -0.03539446, 0.045142844, 0.0016768258, -0.0059571, 0.048917703, 0.05116004, 0.0005371061, 0.004794815, -0.016625853, -0.06221956, 0.0022883292, -0.037917495, -0.026748385, -0.01792442, -0.0045090592, 0.034487702,0.029278198, -0.035341892, 0.019185508, 0.023990307, 0.040781856, 0.033953443, -0.012868148, 0.052692164, 0.005785138, -0.005816314, -0.017770205, 0.052668773, -0.011497479, -0.0042478456, -0.041163344, 0.017727418, 0.019524269, 0.0028672623, 0.027206969, 0.0098760445, -0.01839258, 0.064081036, -0.018610574, -0.03167949, 0.028112203, 0.038009446, 0.027018046, 0.034964353, 0.06139204, -0.000934609, -0.013590096, -0.0112873, -0.016028268, -0.06521579, 0.009233209, -0.010975224, 0.014953974,-0.013712477, -0.035961337, -0.01200976, 0.016888669, -0.0254347, -0.03294505, -0.011085823, 0.074079536, -0.041230906, 0.008066781, 0.016872587, -0.009588438, 0.015129394, 0.004597649, -0.012082546, -0.0205048, -0.011996115, -0.04258334, 0.01090112, 0.025189904, -0.026545173, 0.011398977, -0.035342906, 0.011842086, -0.046653755, 0.0023188903, -0.027082337, -0.053023692, 0.009047406, -0.04184604, -0.0027126037, 0.022172572, -0.0019665314, -0.020220164, -0.0055029565, 0.059257552, -0.038126208,-0.03557876, 0.01125294, 0.024858668, -0.0056422777, 0.0049045635, 0.011553989, 0.026200697, -1.432319e-27, -0.0010860946, 0.09268514, 0.06090386, -0.005796361, 0.025179828, 0.06804877, -0.050707188, 0.02777679, -0.031054588, -0.019260377, -0.04869623, 0.04939678, -0.024753766, -0.0072507886, 0.06201284, 0.038341094, -0.023940781, -0.0044804853, 0.0602865, 0.012955638, 0.041328467, 0.012215266, 0.041857783, 0.05578706, -0.04423121, -0.030089064, 0.039828192, -0.025488378, 0.044566587,0.01636707, 0.05779351, -0.029481513, 0.0053500645, 0.051323455, 0.03458597, 0.013696377, -0.035192043, 0.04330355, 0.058219656, 0.049582005, -0.02016202, -0.036638554, 0.020224864, -0.039496087, -0.015062504, 0.0016461068, 0.015035624, -0.03048053, -0.027442321, -0.019795708, 0.061282612, -0.0035766829, 0.049931448, -0.044324502, -0.034937955, -0.055052124, -0.002334567, 0.0011809984, -0.0005358032, 0.014579384, -0.006727375, 0.01961919, -0.024230124, 0.022260979, 0.041638635, 0.015936956,-0.0129846865, -0.028698431, -0.023311313, 0.020681119, -0.022004813, 0.0045342017, 0.10454673, -0.00907232, 0.018019272, 0.027693322, -0.045678418, 0.008820441, -0.018247915, 0.0022897322, -0.0045168954, 0.0330645, -0.046881303, -0.02397729, -0.03653085, 0.03674319, -0.027837344, -0.02493074, -0.025111457, -0.02385346, -0.026238898, 0.019991593, -0.00067811715, -0.014781632, 0.024634153, -0.016004443, -0.0052194847, 0.024627017, 0.0032142564, 0.013624659, -5.867911e-24, -9.8801356e-05,0.07649344, 0.02002069, 0.029181154, 0.03094678, 0.021331448, 0.029769164, -0.01106459, 0.035867818, 0.0059047374, 0.044848107, -0.034437712, 0.056731243, 0.0032158946, 0.018998314, -0.070217706, -0.029742291, 0.034644052, 0.024905458]}, 'metadata': {}}]

The result is output in the form of dict. If it was not configured before, it was output in the form of Node. At this time, join is either not set, or set to False, which has no effect on the result. It cannot be set to True or a string, otherwise an error will be reported.

ppl = Retriever(documents, group_name="sentences", similarity="cosine", similarity_cut_off=0.003, topk=3, output_format="dict", join=True)

The result looks like this:

AssertionError: Only content output can be joined

Then when we set output_format to content:

ppl = Retriever(documents, group_name="sentences", similarity="cosine", similarity_cut_off=0.003, topk=3, output_format="content")

Running the above code, the result output is as follows:

nodes: ['The way of observing the sky and the journey of adhering to the sky have been completed. \nTherefore, there are five thieves in heaven. Those who see them will prosper. \nThe five thieves are in the heart and execute them in heaven. \nThe universe depends on the hands, and everything depends on the body. \nIt is human nature and scheming. The way of establishing heaven is used to determine people. \nThe sky sends murderous intentions, and the stars move to their places; the earth sends murderous intentions, and dragons and snakes rise from the ground; humans send murderous intentions, and heaven and earth turn over and over again; heaven and man unite, and everything is established. \nSex is cunning and can be hidden. The evil of the nine orifices depends on the three essentials, which can cause movement and stillness. \nFire is born in wood, and disaster will be overcome; treachery is born in a country, and it will collapse in time. Those who know how to practice are called saints. \nHe is born to be killed by nature, which is the principle of Tao. The thief of all things in heaven and earth, the thief of all things by man, the thief of man and all things. The three thieves are both suitable and the three talents are safe. \nTherefore, it is said that if you eat at the right time, everything will be in order; if you use its mechanism, everything will be at peace. People know the gods of their gods, but they don’t know the reason why they are gods. \nThe sun and the moon are numbered, the size is determined, the divine power is born, and the gods are born. \nNo one in the world can see or know about his stolen machine. A gentleman will have a firm bow, while a villain will have a light life. \nThe blind person is good at hearing, and the deaf person is good at seeing. One source of absolute profit, use the teacher ten times. Return to day and night three times, and use your troops ten thousand times. \nThe heart is born from things and dies from things. The machine is in the eyes. \nHeaven is unkind but has great kindness. Thunder and strong wind are all stupid. \nThe most joyful nature, the most tranquil nature and incorruptibility. Heaven is supremely private, and its use is supremely public. \nThe control of birds lies in Qi. The living is the root of death, and the dead is the root of life. Grace is born from harm, harm is born from grace. \nFools regard heaven and earth, culture and science as sages, I regard time, objects, culture and philosophy as sages. ', 'Destiny is called nature, nature is called Tao, and cultivating Tao is called teaching. \nThe Tao cannot be separated in a moment (yú); it can be separated, but it is not the Tao. Therefore, a gentleman should be cautious about what he does not see and be afraid of what he does not hear. Don't see (xian) what is hidden, don't show what is subtle, so a gentleman should be cautious about being alone. \nWhen joy, anger, sorrow, and joy have not arisen, it is called the middle; when they arise, they are all in the middle, it is called the harmony. The one who is in the middle is the foundation of the world; the one who is in harmony is the way of the world. To achieve neutrality, the heaven will have its place and all things will be nurtured. \n\nZhongni said: "A gentleman is the mean, and a villain is against the mean. A gentleman is in the mean, and a gentleman is in the middle at all times; a villain is against the mean, and a villain has no scruples."\n\nConfucius said: "The mean is the end! People rarely last long. ! "\n\nConfucius said: "I know that the way is not good: those who know it can't do it, and the fool can't do it. I know that the way can't be done; the wise person can't do it, and the unworthy person can't do it."', 'Chapter 1\nThe Tao can be Tao, but it is very Tao. \nThe name can be named, but it is very famous. \nNothing is called the beginning of heaven and earth; Being is called the mother of all things. \nTherefore, if there is always nothing, I want to observe its wonderfulness; if there is always something, I want to observe its beauty. \nThe two have the same origin but different names. They are both called mysteries, and mysteries are mysteries, and they are the door to all mysteries. \nChapter 2\nEveryone in the world knows that beauty is beautiful, but it is evil; \nEveryone knows that good is good, but it is not good. \nTherefore, presence and absence complement each other, difficulty and ease complement each other, long and short complement each other, high and low complement each other, sounds harmonize with each other, front and rear follow each other, and they are eternal. \nThis is because the sage does nothing and teaches without saying anything. \nEverything is created without beginning, born without being, done without relying on it, and accomplished without living. \nI feel like I'm not living here, so I won't go. \nChapter 3\nIf you don't respect the virtuous, the people will not fight. \nInexpensive and hard-to-find goods prevent people from stealing. \nThere is no desire to be seen, so that the people's hearts are not in chaos. \nThis is the rule of the sage: \nIf the heart is empty, the belly is strong; if the ambition is weak, the bones are strong. \nIt often makes the people ignorant and desireless, and makes people who know it dare not do anything. \nIf you do nothing, everything will be cured. \nChapter 4\nIf you use it when the Tao rushes, it may not make a profit. \nThe abyss is like the origin of all things, and the deep is like Ruexun. \nI don’t know whose son is the first of Emperor Xiang. \nChapter 5\nHeaven and earth are unkind and regard all things as stupid dogs. \nThe saint is unkind and treats the common people as stupid dogs. \nBetween heaven and earth, is it still a mess? \nIf it is empty but not yielding, it will get better when it moves. \nIt is better to stay focused than to talk too much. \nChapter 6\nThe God of Grain is immortal and is called the mysterious female. \nThe gate of Xuan Ni is called the root of heaven and earth. \nIf it persists, it will not be used diligently. \nChapter 7\nForever. \nThe reason why heaven and earth can last forever is because they do not generate themselves, so they can live forever. \nThis means that the saint comes first after his body, and exists outside his body. \nIt’s not because of his selflessness! \nSo it can be done privately. \nChapter 8\nThe best is like water. \nWater is good for all things without fighting.\nIt is what everyone dislikes,\nso it is close to the Tao. \nLiving in a good place, having a good heart, being kind and benevolent, speaking good words and trustworthy, doing good government, doing good things, and doing good things at the right time. \nThe husband does not fight, so there is no need for you. \nChapter 9\nIf you hold on to it and gain it, it will not be as good as it is; if you hold on to it and keep it sharp, you will not be able to maintain it for a long time. \nIf a house is full of gold and jade, no one can guard it; if you are arrogant and rich, you will be blamed for it. \nThe merit is then withdrawn, which is the way of heaven. \nChapter 10\nZai Ying Po hugs one, can we be inseparable. \nSpecialized Qi makes it as soft as a baby. \nIs it possible to eliminate all mysteries and make them flawless? \nIf you love the people and govern the country, you can do nothing. \nThe gate of heaven is open and closed, can it be a female? \nIf you understand the Sida, you can be ignorant. \nChapter 11\nThirty spokes make up one hub. If it is not there, it can be used as a car. \nThe clay is regarded as a utensil, but it should be used as a utensil if it is not there. \nCut out a door to make a room. If it doesn't exist, use it as a room. \nTherefore, if you have it, you will benefit from it, and if you don’t have it, you will use it. ']

Because join is not configured, its default value is False, so it is output in List[str] format.

What if join is set to True at this time?

ppl = Retriever(documents, group_name="sentences", similarity="cosine", similarity_cut_off=0.003, topk=3, output_format="content", join=True)

The result output is as follows:

nodes: Destiny is called nature, nature is called Tao, and cultivating Tao is called teaching.
The Tao cannot be separated for a moment; the Tao can be separated, but it is not the Tao. Therefore, a gentleman should be cautious about what he does not see and be afraid of what he does not hear. Don't see (xian) what is hidden, don't show what is subtle, so a gentleman should be cautious about being alone.
When joy, anger, sorrow, and joy are not expressed, it is called the middle; when they are expressed, they are all in the middle, it is called the harmony. The one who is in the middle is the foundation of the world; the one who is in harmony is the way of the world. To achieve neutrality, the heaven will have its place and all things will be nurtured.

Zhongni said: "A gentleman is moderate, and a villain is against the mean. A gentleman is moderate, and a gentleman is in the right time; a villain is against the mean, and a villain has no scruples."

Confucius said: "The golden mean is the best! The people can survive for a long time!"

Confucius said: "I know what the Tao is not good at. Those who know it are better than it, but fools are not as good as it is. I know it is not clear about the Tao: the sages are beyond it, but the unworthy man is not as good as it. People all eat and drink, and they rarely know the taste." Chapter 1
The Tao can be Tao, but it is very Tao.
Famous, very famous.
Nothing is called the beginning of heaven and earth; Being is called the mother of all things.
Therefore, if there is always nothing, I would like to observe its wonderfulness; if there is always something, I would like to observe its beauty.
The two have the same origin but different names. They are both called xuan, and they are both mysterious and mysterious. They are the door to all mysteries.
Chapter 2
Everyone in the world knows that beauty is beauty, but this is evil;
Everyone knows that good is good, but this is not good.
Therefore, existence and non-being arise from each other, difficulty and ease complement each other, long and short complement each other, high and low complement each other, sound and sound harmonize each other, front and rear follow each other, and they are eternal.
This is why the sage does nothing but teaches without saying anything.
Everything is created without beginning, born without being, done without relying on it, and accomplished without living.
I feel like I have no place to live, so I won’t go.
Chapter 3
If you do not respect the virtuous, the people will not fight.
Goods that are cheap and hard to find will prevent people from stealing.
There is no desire to be seen, so that the people's minds are not in chaos.
This is the rule of saints:
If your heart is empty, your belly will be strong; if your will is weak, your bones will be strong.
It often makes the people ignorant and desireless, and makes people who know it dare not do anything.
If you do nothing, everything will be cured.
Chapter 4
Tao rushes and uses it or it may not be full.
The abyss is like the ancestor of all things, and the deep is like Ruexun.
I don’t know whose son is the first of Emperor Xiang.
Chapter 5
Heaven and earth are unkind and regard all things as stupid dogs.
The saint is unkind and treats the common people as stupid dogs.
Between heaven and earth, is it still like a quarrel?
If it is empty but not yielding, it will get better when it moves.
It is better to keep the center than to talk too much.
Chapter 6
The God of Grain is immortal and is called the mysterious female.
The gate of Xuanfei is called the root of heaven and earth.
If it persists, it will not be used diligently.
Chapter 7
Forever.
The reason why heaven and earth can be long-lasting is because they do not generate themselves, so they can live forever.
Therefore, the sage follows his body and comes first, and his body exists outside his body.
It’s not because of his selflessness!
Therefore, it can be done privately.
Chapter 8
The best is like water.
Water is good for all things without fighting for it.
Where everyone hates you,
Therefore, it is closer to Tao.
Live in a good place, have a good heart, be kind and benevolent, speak good words and have good faith, govern well, do good things and be able to do good things at the right time.
The husband does not fight, so there is no need for you.
Chapter 9
If you hold on to it and make it full, it won't be as good as it is; if you hold on to it and sharpen it, it won't last long.
A house full of gold and jade cannot be guarded by anyone; if you are arrogant and rich, you will be blamed for it.
When one's meritorious deeds are fulfilled, the body retreats, which is the way of heaven.
Chapter 10
If you hold the soul of Zai Ying in your arms, there will be no separation.
Focus on the Qi and become as soft as a baby.
Can you eliminate all mysteries and be flawless?
If you love the people and govern the country, you can do nothing.
The gate of heaven is open and closed, can it be a female?
If you understand the four elements, you can be ignorant.
Chapter 11
Thirty spokes make up one hub, which can be used as a vehicle if it is not there.
The clay is regarded as a utensil, but it should be used as a utensil if it is not there.
Cut out the door to make a room, pretend it doesn't exist, and use it as a room if there is one.
Therefore, if you have it, you will benefit from it, and if you don’t have it, you will use it. The way of observing the sky and the journey of holding on to the sky have been completed.
Therefore, there are five thieves in heaven, and those who see them will prosper.
The five thieves are in the heart and execute them in heaven.
The universe depends on the hands, and all transformations occur on the body.
Human beings are born with human nature and are scheming. The way of establishing heaven is used to determine people.
The sky sends murderous intentions, and the stars move to their places; the earth sends murderous intentions, and dragons and snakes rise from the ground; humans send murderous intentions, and heaven and earth turn over and over again; the combination of heaven and man creates the basis for all transformations.
Nature is clever and can be hidden. The evil of the nine orifices depends on the three essentials, which can cause movement and stillness.
If fire is born in wood, disaster will be overcome; if treachery is born in the country, it will collapse in time. Those who know how to practice are called saints.
God is born to kill, this is the principle of Tao. The thief of all things in heaven and earth, the thief of all things by man, the thief of man and all things. The three thieves are both suitable and the three talents are safe.
Therefore, it is said that if you eat at the right time, everything will be in order; if you move its mechanism, everything will be in peace. People know the gods of their gods, but they don’t know the reason why they are gods.
The sun and the moon are numbered, the size is determined, the divine power is generated, and the gods are revealed.
No one in the world can see or know how to steal a machine. A gentleman will have a firm bow, while a villain will have a light life.
The blind person is good at hearing, and the deaf person is good at seeing. One source of absolute profit, use the teacher ten times. Return to day and night three times, and use your troops ten thousand times.
The heart is born from things and dies from things. The mechanism lies in the eyes.
God is unfaithful but has great kindness. Thunder and strong wind are all stupid.
The most joyful nature, the most tranquil nature and incorruptibility. Heaven is supremely private, and its use is supremely public.
The control of birds lies in Qi. The living is the root of death, and the dead is the root of life. Grace is born from harm, harm is born from grace.
Fools regard heaven and earth, culture and science as sages, I regard time, objects, culture and philosophy as sages.

At this time, the result is text concatenated with empty strings. What if join is set to '11111111111111111111111111111'?

ppl = Retriever(documents, group_name="sentences", similarity="cosine", similarity_cut_off=0.003, topk=3, output_format="content", join='11111111111111111111111111111')

The result output is as follows:

nodes: Chapter 1
The Tao can be Tao, but it is very Tao.
Famous, very famous.
Nothing is called the beginning of heaven and earth; Being is called the mother of all things.
Therefore, if there is always nothing, I would like to observe its wonderfulness; if there is always something, I would like to observe its beauty.
The two have the same origin but different names. They are both called xuan, and they are both mysterious and mysterious. They are the door to all mysteries.
Chapter 2
Everyone in the world knows that beauty is beauty, but this is evil;
Everyone knows that good is good, but this is not good.
Therefore, existence and non-being arise from each other, difficulty and ease complement each other, long and short complement each other, high and low complement each other, sound and sound harmonize each other, front and rear follow each other, and they are eternal.
This is why the sage does nothing but teaches without saying anything.
Everything is created without beginning, born without being, done without relying on it, and accomplished without living.
I feel like I have no place to live, so I won’t go.
Chapter 3
If you do not respect the virtuous, the people will not fight.
Goods that are cheap and hard to find will prevent people from stealing.
There is no desire to be seen, so that the people's minds are not in chaos.
This is the rule of saints:
If your heart is empty, your belly will be strong; if your will is weak, your bones will be strong.
It often makes the people ignorant and desireless, and makes people who know it dare not do anything.
If you do nothing, everything will be cured.
Chapter 4
Tao rushes and uses it or it may not be full.
The abyss is like the ancestor of all things, and the deep is like Ruexun.
I don’t know whose son is the first of Emperor Xiang.
Chapter 5
Heaven and earth are unkind and regard all things as stupid dogs.
The saint is unkind and treats the common people as stupid dogs.
Between heaven and earth, is it still like a quarrel?
If it is empty but not yielding, it will get better when it moves.
It is better to keep the center than to talk too much.
Chapter 6
The God of Grain is immortal and is called the mysterious female.
The gate of Xuanfei is called the root of heaven and earth.
If it persists, it will not be used diligently.
Chapter 7
Forever.
The reason why heaven and earth can be long-lasting is because they do not generate themselves, so they can live forever.
Therefore, the sage follows his body and comes first, and his body exists outside his body.
It’s not because of his selflessness!
Therefore, it can be done privately.
Chapter 8
The best is like water.
Water is good for all things without fighting for it.
Where everyone hates you,
Therefore, it is closer to Tao.
Live in a good place, have a good heart, be kind and benevolent, speak good words and have good faith, govern well, do good things and be able to do good things at the right time.
The husband does not fight, so there is no need for you.
Chapter 9
If you hold on to it and make it full, it won't be as good as it is; if you hold on to it and sharpen it, it won't last long.
A house full of gold and jade cannot be guarded by anyone; if you are arrogant and rich, you will be blamed for it.
When one's meritorious deeds are fulfilled, the body retreats, which is the way of heaven.
Chapter 10
If you hold the soul of Zai Ying in your arms, there will be no separation.
Focus on the Qi and become as soft as a baby.
Can you eliminate all mysteries and be flawless?
If you love the people and govern the country, you can do nothing.
The gate of heaven is open and closed, can it be a female?
If you understand the four elements, you can be ignorant.
Chapter 11
Thirty spokes make up one hub, which can be used as a vehicle if it is not there.
The clay is regarded as a utensil, but it should be used as a utensil if it is not there.
Cut out the door to make a room, pretend it doesn't exist, and use it as a room if there is one.
Therefore, if you have it, you will benefit from it, and if you don’t have it, you will use it. 11111111111111111111111111111 The way of observing the sky and the journey of holding on to the sky are complete.
Therefore, there are five thieves in heaven, and those who see them will prosper.
The five thieves are in the heart and execute them in heaven.
The universe depends on the hands, and all transformations occur on the body.
Human beings are born with human nature and are scheming. The way of establishing heaven is used to determine people.
The sky sends murderous intentions, and the stars move to their places; the earth sends murderous intentions, and dragons and snakes rise from the ground; humans send murderous intentions, and heaven and earth turn over and over again; the combination of heaven and man creates the basis for all transformations.
Nature is clever and can be hidden. The evil of the nine orifices depends on the three essentials, which can cause movement and stillness.
If fire is born in wood, disaster will be overcome; if treachery is born in the country, it will collapse in time. Those who know how to practice are called saints.
God is born to kill, this is the principle of Tao. The thief of all things in heaven and earth, the thief of all things by man, the thief of man and all things. The three thieves are both suitable and the three talents are safe.
Therefore, it is said that if you eat at the right time, everything will be in order; if you move its mechanism, everything will be in peace. People know the gods of their gods, but they don’t know the reason why they are gods.
The sun and the moon are numbered, the size is determined, the divine power is generated, and the gods are revealed.
No one in the world can see or know how to steal a machine. A gentleman will have a firm bow, while a villain will have a light life.
The blind person is good at hearing, and the deaf person is good at seeing. One source of absolute profit, use the teacher ten times. Return to day and night three times, and use your troops ten thousand times.
The heart is born from things and dies from things. The mechanism lies in the eyes.
God is unfaithful but has great kindness. Thunder and strong wind are all stupid.
The most joyful nature, the most tranquil nature and incorruptibility. Heaven is supremely private, and its use is supremely public.
The control of birds lies in Qi. The living is the root of death, and the dead is the root of life. Grace is born from harm, harm is born from grace.
Fools regard heaven and earth, culture and science as sages, I regard time, objects, culture and philosophy as sages. 111111111111111111111111111111 Destiny is called nature, nature is called Tao, and cultivating Tao is called teaching.
The Tao cannot be separated for a moment; the Tao can be separated, but it is not the Tao. Therefore, a gentleman should be cautious about what he does not see and be afraid of what he does not hear. Don't see (xian) what is hidden, don't show what is subtle, so a gentleman should be cautious about being alone.
When joy, anger, sorrow, and joy are not expressed, it is called the middle; when they are expressed, they are all in the middle, it is called the harmony. The one who is in the middle is the foundation of the world; the one who is in harmony is the way of the world. To achieve neutrality, the heaven will have its place and all things will be nurtured.

Zhongni said: "A gentleman is moderate, and a villain is against the mean. A gentleman is moderate, and a gentleman is in the right time; a villain is against the mean, and a villain has no scruples."

Confucius said: "The golden mean is the best! The people can survive for a long time!"

Confucius said: "I know what the Tao is not good at: those who know it go beyond it, and those who are ignorant fail to do it. I know what the Tao is not clear about: the sage goes beyond it, and the unworthy man falls short of it. People all eat and drink, and they rarely know the taste."

It can be seen that the result is a string composed of three pieces of text.

Detailed explanation of Reranker parameters

Full GitHub code link🔗

The two parameter values ​​in Reranker are the same as above. We won’t go into details here and just look at the code. We let Reranker output the two results with the highest scores.

import lazyllm
from lazyllm import pipeline, bind, OnlineEmbeddingModule, SentenceSplitter, Retriever, Reranker
from lazyllm.tools.rag import Document

prompt = 'You will play the role of an AI Q&A assistant and complete a dialogue task. In this task, you need to provide your answer based on the given context and question.'

documents = Document(dataset_path="rag_master", embed=OnlineEmbeddingModule(source="glm", embed_model_name="embedding-2"), manager=False)
documents.create_node_group(name="sentences", transform=SentenceSplitter, chunk_size=1024, chunk_overlap=100)

with pipeline() as ppl:
    ppl.retriever = Retriever(documents, group_name="sentences", similarity="cosine", similarity_cut_off=0.003, topk=3)
    ppl.reranker = Reranker(name="ModuleReranker", model=lazyllm.OnlineEmbeddingModule(type="rerank", source="glm", embed_model_name="rerank"), topk=2) | bind(query=ppl.input)

nodes = ppl("What is the way of heaven")
print(f"nodes: {nodes}")

The result output is as follows:

nodes: [<Node id=a8713a16-fc5b-4904-a06f-9e8c55d11988>, <Node id=058767c8-de0e-4d17-addc-23816ae8e2d2>]

You can see that if nothing is set, the result is output in the form of Node.

Next set output_format to dict:

ppl.reranker = Reranker(name="ModuleReranker", model=lazyllm.OnlineEmbeddingModule(type="rerank", source="glm", embed_model_name="rerank"), topk=2, output_format="dict") | bind(query=ppl.input)

The result output is as follows:

nodes: [{'content': 'The way to observe the sky and the journey to the sky are complete. \nTherefore, there are five thieves in heaven. Those who see them will prosper. \nThe five thieves are in the heart and execute them in heaven. \nThe universe depends on the hands, and everything depends on the body. \nIt is human nature and scheming. The way of establishing heaven is used to determine people. \nThe sky sends murderous intentions, and the stars move to their places; the earth sends murderous intentions, and dragons and snakes rise from the ground; humans send murderous intentions, and heaven and earth turn over and over again; heaven and man unite, and everything is established. \nSex is cunning and can be hidden. The evil of the nine orifices depends on the three essentials, which can cause movement and stillness. \nFire is born in wood, and disaster will be overcome; treachery is born in a country, and it will collapse in time. Those who know how to practice are called saints. \nHe is born to be killed by nature, which is the principle of Tao. The thief of all things in heaven and earth, the thief of all things by man, the thief of man and all things. The three thieves are both suitable and the three talents are safe. \nTherefore, it is said that if you eat at the right time, everything will be in order; if you use its mechanism, everything will be at peace. People know the gods of their gods, but they don’t know the reason why they are gods. \nThe sun and the moon are numbered, the size is determined, the divine power is born, and the gods are born. \nNo one in the world can see or know about his stolen machine. A gentleman will have a firm bow, while a villain will have a light life. \nThe blind person is good at hearing, and the deaf person is good at seeing. One source of absolute profit, use the teacher ten times. Return to day and night three times, and use your troops ten thousand times. \nThe heart is born from things and dies from things. The machine is in the eyes. \nHeaven is unkind but has great kindness. Thunder and strong wind are all stupid. \nThe most joyful nature, the most tranquil nature and incorruptibility. Heaven is supremely private, and its use is supremely public. \nThe control of birds lies in Qi. The living is the root of death, and the dead is the root of life. Grace is born from harm, harm is born from grace. \nFools regard heaven and earth, culture and science as sages, I regard time, objects, culture and philosophy as sages. ', 'embedding': {'__default__':[0.0041554114, 0.0075854566, -0.007057654, 0.045291275, 0.04896368, 0.028331915, -0.05941855, 0.023905145, 0.041470468, -0.048525453, 0.041954085, -0.009402919, -0.026280737, -0.0059798467, 0.03535848, -0.01118226, -0.034457922, 0.05115644, -0.0008446922, -0.026550785, -0.007863325, 0.027618073, -0.011360486, -0.039742425, -0.004911823, -0.04183471, -0.005898696, 0.044848423, -0.030583072, -0.01361686, 0.038467087, -0.04660302, -0.02357292, 0.04048558, 0.03306541, -0.027309675, 0.048068833,0.088752195, -0.009470059, 0.014505637, 0.0019608818, 0.024321485, -0.0624449, -0.050921675, -0.009048667, -0.014897456, -0.02090078, 0.0161172, -0.031017896, -0.006250185, 0.0034054928, 0.017325701, 0.05056032, 0.007071267, -0.008573053, 0.0057591763, -0.032224294, 0.015077854, -0.021500586, 0.050998785, -0.012885331, -0.0015166812, 0.016822474, 0.010382974, -0.01868565, -0.029895348, -0.035598777, -0.017912406, 0.003902434, -0.019268971, 0.025545066, -0.0016119897, -0.07398274,-0.0090994835, -0.015249873, -0.05385994, 0.041456234, 0.025098544, 0.011212095, 0.023621807, 0.02871689, 0.0155634, -0.017000291, -0.0057249004, 0.041292667, 0.014244264, -0.020487353, 0.009796285, 0.0017768699, 0.0029904752, -0.021676986, 0.050575286, 0.05815488, 0.041090943, 0.03852607, 0.0007469177, 0.0146834655, -0.019205064, -0.02501764, -0.014589938, -0.025496105, 0.0029830346, 0.009922189, -0.014940715, -0.05994815, 0.018909076, -0.0024642693, 0.04501117, -0.047053687, -0.014784821,-0.0020920835, 0.035193928, 0.042950764, 0.048697557, -0.015930817, -0.018142601, -0.021126166, 0.0013488104, 0.02504943, -0.007454921, -0.026205262, 0.03224436, 0.06320423, 0.0017354549, -0.013907578, 0.053008184, 0.013806328, -0.040066086, -0.011262496, -0.00783606, -0.032809734, 0.042265926, 0.003031213, -0.011611754, 0.0038381496, 0.026966704, 0.042800177, 0.032915533, 0.016108219, 0.019568527, 0.06046945, -0.010420891, 0.03388435, -0.0009712575, 0.032708555, -0.039891865, -0.029775867,-0.014499864, 0.022318674, 0.029443137, 0.010519894, 0.057602942, -2.1037191e-13, -0.040184353, -0.023729598, 0.059133578, -0.060146023, -0.009997531, 0.0071468763, 0.02307546, -0.008445198, 0.025550805, -0.062879466, 1.1355897e-17, 0.005254596, 0.039959684, 0.008519935, 0.028987143, 0.069844805, -0.014054391, -0.035702214, 0.03448102, 0.035850395, -0.026240302, -0.060955, -0.015616291, 0.005105854, -0.03699429, 0.013898516, -0.011044874, -0.008250217, 0.017763525, 0.014568678, 0.011624834,-0.016273065, -0.016237382, -0.030793475, -0.014173828, -0.004470327, -0.00031035003, -0.047902904, -0.033007123, -0.037031222, 0.0365982, 0.030188967, 0.030725434, 0.023961129, -0.02410339, -0.029226728, 0.03464168, -0.024372779, -0.021495208, -0.057038497, 0.073897764, -0.0141333, 0.019894436, 0.05053698, -0.012459984, 0.030188244, -0.0424957, -0.009643966, -0.001958911, 0.020065648, 0.014195304, 0.0558965, 0.01836145, -0.031815484, -0.03150872, 0.0695297, 0.031345744, 0.008421123,0.0014973548, -0.033643615, -0.014629425, 0.022442328, -0.028167598, 0.015380408, -0.014378052, 0.012282112, -0.053193033, 0.019917091, 0.019215647, -0.033822346, 0.0027193974, -0.014981052, 0.00055290776, -0.004483305, -0.006755365, 0.028678952, 0.023475507, 0.020434262, -0.03490839, -0.027436005, 0.010972408, 0.07965436, 0.0011874165, -0.045413267, 0.028245784, -0.0063281492, 0.009295957, -0.03730253, 0.06491051, -0.045857605, -0.06690101, -0.0045730066, -0.024991188, 0.034862872,-0.015739491, 0.0529046, 0.03173905, 0.0703093, -0.0026089184, 0.054319624, 0.007398687, -5.9605067e-05, 0.017072119, -0.008852119, 0.0026614992, 0.038429067, 0.016159521, 0.026893796, -0.035562247, 0.038095932, 0.0065683676, 0.03324841, 0.016334219, -0.023613118, -0.040397674, -0.009194677, -0.008144042, -0.016446576, -0.01151405, -0.010928354, 0.038340032, -0.027145639, -0.046593655, 0.015169281, 0.050768223, -0.03294848, 0.01951992, 0.013797078, 0.035349254, 0.008481241, 0.031466227,-0.003075734, -0.029220875, -0.07521815, -0.0043719853, 0.048672307, 0.0058878683, -0.019860264, -0.03880463, 0.016394598, -0.0122021, -1.4963505e-05, 0.044320922, 0.007890018, 0.0054832306, -0.0123292245, 0.03318079, -0.015232731, -0.013402485, -0.02389836, 0.0055185272, -0.028106516, 0.022771437, -0.030289527, 0.02677829, 0.083992176, -0.012977583, -0.020257147, 0.07025292, -0.0054380586, 0.00113687, -0.003930704, -0.059119448, 0.0005393099, 0.006615847, -0.008725304, -0.011702107,0.07533191, 0.028773967, 0.029288677, 0.048617944, -0.012453482, -0.02434141, -0.026455658, -0.038755085, 0.010074069, -0.043186586, -0.008087229, -0.030695459, 0.0028579491, 0.010711575, -0.027886447, 0.024541127, 0.011004065, -0.06523756, 0.043040413, 0.025965191, 0.013071675, 0.023030335, 0.029349305, 0.021783385, 0.018128468, 0.006819187, 0.006515866, 0.029666856, -0.06356491, 0.039198253, -0.011071719, 0.023561964, -0.03198062, 0.023516942, 0.059817895, 0.004185616, -0.008066767,-0.009041066, 0.015803682, 0.0067058923, 0.041337356, 0.0052649537, -0.027125185, -0.038213175, 0.014553458, 0.04226332, 0.070982724, -0.11455366, -0.032585356, 0.0102267265, 0.011878967, -0.033385586, 0.0010560753, -0.015395409, -0.00441643, -0.01337264, -0.011390325, 0.03452036, -0.0043478706, 0.021369405, 0.022401562, -0.022205178, 0.01836901, -0.040310632, -0.014155401, 0.022052731, -0.033351872, 0.023539973, 0.0066144154, -0.02636437, -0.049085867, 0.021261329, 0.025078895, -0.03848087,0.031188102, -0.048261605, 0.019504659, -0.005102396, -0.01836231, 0.02110646, 0.025233712, 0.031062676, 0.0010754078, -0.017802356, -0.013914974, 0.014522387, 0.056926217, 0.0053230077, 0.028501919, 0.048797324, 0.023192475, 0.027942462, 0.0073998393, -0.03499081, 0.022158591, -0.019583017, 0.005857129, 0.046547387, -0.038932085, 0.040460087, 0.042364065, 0.0429083, 0.009138595, -0.012183861, 0.058576614, -0.030963775, -0.034653503, -0.025639888, -0.011284233, 0.047425315, -0.029514484,-0.036847457, -0.019771382, 0.008259031, 0.022998115, -0.017095372, 0.0066413945, -0.05487991, 0.002815856, -0.033681728, -0.014198309, 0.026634654, -0.00822199, -0.00530929, -0.011223847, -0.011880681, -0.034348324, -0.04212451, 0.067977875, -0.034103774, 0.05572993, -0.0062887007, 0.010488381, -0.020459691, 0.012848595, -0.025240324, -0.06849307, 0.025805129, -0.011156502, 0.05313631, -0.024320655, -0.028916016, 0.010576907, -0.0015999046, 0.024717461, -0.0144396275, -0.055715647,0.030035136, 0.05483442, 0.010892865, -0.0075423573, 0.03750252, -0.005360984, 0.061075144, -0.05138248, -0.018550297, -0.06849128, 0.03832335, 0.051121823, 0.07625229, -0.02237382, -0.02532701, 0.005352677, -0.008111206, 0.073260285, -0.050452314, -0.03776045, -0.049333394, 0.044611394, -0.026005713, 0.039077275, 0.0220915, -0.029445095, -0.0064717205, -0.027871978, -0.035875883, 0.0024223006, 0.018516455, 0.05704868, -0.03898351, 0.006290837, -0.06967245, -0.021957148, 0.030887723,0.019454552, -0.016694501, 0.013290308, -0.0031141401, -0.013080929, -0.016133493, -0.014472164, -0.01712892, -0.037303273, 0.027202567, -0.020992804, 0.042868, 0.013083941, -0.0220378, 0.004116772, -0.014082884, -0.03672434, -0.025468608, -0.004967429, -0.043549094, -0.059921604, -0.0267599, 0.020867042, -0.024328588, -0.009694849, -0.008663098, 0.029830154, 0.015654892, -0.01019365, -0.005135016, -0.005788498, 0.005652381, 0.018197127, -0.048258357, 0.018697621, 0.009251598, 0.013933649,-0.05764918, 0.017524255, -0.055997822, 0.00016845911, -0.02751447, 0.032793835, 0.0019559327, -0.0075224875, -0.01907267, -0.008904949, -0.030548248, 0.013870509, 0.0069685555, -0.028905436, -0.02268364, -0.020575576, -0.0030170542, 0.015130529, -0.00084239285, -0.029587327, 0.037500374, -0.027868606, -0.006660683, 0.031205097, 0.012189993, -0.000101473976, 0.012176301, -0.012304269, 0.016419522, -0.031622086, -0.031915437, -0.07447143, 0.044259857, 0.034456633, -0.030144012, 0.006155522,-0.047044933, -0.053129446, -0.003883295, -0.02238673, -0.036794078, 0.018963076, 0.038487703, 0.043547265, 0.020530019, 0.013525153, 0.03025931, -0.012935889, 0.0006987865, 0.001170972, 0.049213756, -0.021247994, -0.010136263, -0.008933942, 0.040884793, -0.014472714, 0.04323616, -0.024579003, 0.044722065, -0.00044178023, 0.022410542, -0.043617256, 0.022241985, 0.04859234, -0.034833476, 0.013860875, 0.023002412, 0.00644085, -0.01903299, -0.030982777, -0.02251995, 0.025486361, 0.013646234,-0.025401177, 0.028737532, -0.0021397755, 0.005839515, 0.031907808, -0.017933525, 0.0054767667, 0.011915933, 0.04648308, -0.0016207868, -0.02563286, -0.044512413, 0.03895756, 0.026000049, 0.028605616, 0.0051885676, 0.0020558056, 0.0013344906, 0.018722871, 0.01694392, -0.048954893, -0.0022045183, -0.04172797, -0.021244593, 0.013639434, 0.0023175476, -0.057096064, -0.017740823, -0.041350298, -0.0043968866, -0.03799386, -0.0042910026, -0.09530829, -0.038248345, -0.030624647, -0.0110857375,-0.010244377, 0.046464287, 0.11650956, 0.015119723, -0.069281384, 0.01305812, 0.012854464, -0.0055132634, 0.018360458, -0.0165773, -0.0032716265, 0.031324998, -0.047736414, 0.014519401, 0.048135165, -0.035682708, -0.0365996, 0.03389063, -0.0043802285, -0.01877874, 0.04836236, 0.020456195, -0.056787007, 0.010132782, 0.0004068694, -0.047474485, 0.03445135, -0.02724392, -0.03328944, -0.0029203526, -0.026272407, -0.00029890105, 0.013430369, 0.013834825, 0.060670216, 0.012882629, 0.06432347,0.015768427, -0.010093849, 0.024928765, 0.012602784, -0.064383924, -0.0050611207, -0.044316787, -0.041941397, -0.04322314, 0.006486478, 0.0071116807, -0.012460606, -0.0074164756, 0.029274957, -0.024530469, 0.018598046, 0.03159384, -0.001951544, 0.00523427, 0.010126147, -0.017179206, -0.018108787, -0.007169589, 0.009502527, 0.008932048, -0.069397494, 0.044141974, -0.005750729, -0.045750074, 0.023508383, 0.03368248, 0.027255893, -0.015447886, 0.056365486, 0.008277988, 0.008696923, -0.059466016,-0.0071224286, 0.020547476, 0.028926386, -0.031088999, 0.0034343835, -0.009711087, 0.0011468836, -0.013077483, -0.025331382, -0.016229326, 0.02545839, 0.011994319, 0.008591205, -0.0015080342, 0.02231805, -0.04337707, 0.040335864, 0.026538009, -0.0062174527, 0.050592676, 0.029670224, -0.013418461, 0.0727519, -0.037769295, 0.04368287, 0.03542585, -0.020641856, -0.05900921, -0.027732523, 0.0072757383, -0.019184208, -0.013169488, 0.022744527, -0.010796951, 0.013560294, -0.035595182, 0.023965288,0.010551866, 0.024181299, -0.03136934, -0.044932388, -0.020812577, -0.054782774, 0.04152298, -0.011101999, 0.024159305, 0.019186165, -0.062283333, 0.042109832, -0.0031904012, -0.056608956, 0.030812602, 0.020276703, 0.03021552, -0.039611798, 0.03263928, -0.008093682, -0.029934445, -0.003328774, 0.04017525, 0.05742037, 0.065777995, 0.011823178, -0.022649916, -0.030095868, -0.029602733, 3.6471364e-09, 0.019888746, 0.042886823, -0.013049726, 0.037731685, -0.036016267, -0.046768334, 0.060499705,0.00012572506, 0.016042653, 0.024610577, 0.033402827, 0.004141988, -0.0006018001, -0.055084515, -0.039775707, 0.008617055, -0.01653752, -0.013155907, 0.016346993, -0.023589248, 0.024627982, -0.012017998, -0.053784538, 0.037820525, 0.03518804, 0.046451963, 0.016566351, 0.019617809, 0.03032722, 0.051768657, 0.003973617, 0.011890001, 0.030245746, -0.009497901, -0.001179198, -0.026856216, 0.015133692, 0.04234234, 0.007812807, -0.02503269, 0.017157003, -0.061224688, 0.04974642, 0.033100616,-0.0151869515, 0.024570305, 0.018643305, 0.06040034, 0.038901594, 0.04535261, -0.04548045, 0.0020181928, -0.034227967, -0.0013509532, -0.030452356, -0.018487962, -0.023855759, 0.05625316, -0.00090770074, -0.00640004, -0.014777776, 0.027941104, 0.016304482, -0.038740773, 0.011337658, 0.07173977, -0.03550307, 0.023911772, -0.0016779553, -0.033899304, -0.022336006, -0.015419301, 0.001863339, 0.0008760349, 0.009762301, -0.028453615, -0.011362094, 0.04275605, -0.043521807, 0.029829519, -0.02416117,0.0335993, -0.006421268, 0.04060249, -0.032474253, -0.07600941, 0.004471598, -0.04126713, -0.02453735, 0.011764305, 0.015372352, 0.0028955622, -0.0291478, 0.05877516, -0.009823992, -0.054267995, 0.0067844344, 0.03217804, -0.011592191, -0.00035714993, 0.028778687, 0.0087011475, -1.7158678e-27, 0.0017932786, 0.03569676, 0.054138917, -0.008628211, 0.043432187, 0.023903232, -0.0067582014, -0.03050816, -0.028248005, -0.02057966, -0.057666212, 0.031576764, -0.0151130445, -0.027907273, 0.019604214,0.019528171, -0.030740337, 0.013142259, 0.04407093, -0.0029533103, 0.0063910317, 0.017176246, 0.032132074, 0.022815445, -0.050552312, -0.020800956, 0.018182818, -0.037349436, 0.014202859, -0.00087027677, 0.058879256, 0.0056527536, -0.005539146, 0.032238558, 0.022045877, 0.03943093, 0.0077616903, 0.02564012, 0.06469182, 0.020434259, -0.028960383, -0.01953917, 0.016642196, -0.04865382, -0.009785435, 0.00070976716, 0.005543073, -0.00053107564, 0.019826816, -0.046817478, 0.02328427, 0.0038257786,0.041383106, -0.049317617, -0.012191691, 0.040143386, -0.020945452, -0.015941637, 0.014165448, -0.014123277, 0.0048813364, 0.009677354, -0.001581121, 0.03251474, 0.042997185, -0.015337093, -0.021302974, -0.03482498, -0.02814613, 0.014423916, 0.030641912, -0.0033474728, 0.019615268, -0.046341933, -0.007550764, 0.014528103, -0.024804426, -0.04388465, -0.00038302864, 0.018112512, 0.016945953, 0.022862429, -0.06407195, 0.02999205, -0.029377965, 0.038923685, -0.003280456, -0.0140948305,-0.016372854, -0.0024543288, -0.019321902, 0.029282054, -0.0075866296, -0.018350568, 0.014755779, -0.050756894, -0.038549382, -0.026098477, 0.0016459071, 0.023259882, -6.1448135e-24, -0.022952693, 0.022025546, 0.014744688, 0.009926959, 0.044778008, 0.018253125, 0.04506746, -0.019753184, 0.046412323, 0.00015234029, 0.060078166, -0.060511682, 0.055275027, 0.02107507, 0.0074826907, -0.026740791, -0.012698739, 0.017056648, 0.050418142]}, 'metadata': {}}, {'content':'Destiny is called nature, nature is called Tao, and cultivating the Tao is called teaching. \nThe Tao cannot be separated in a moment (yú); it can be separated, but it is not the Tao. Therefore, a gentleman should be cautious about what he does not see and be afraid of what he does not hear. Don't see (xian) what is hidden, don't show what is subtle, so a gentleman should be cautious about being alone. \nWhen joy, anger, sorrow, and joy have not arisen, it is called the middle; when they arise, they are all in the middle, it is called the harmony. The one who is in the middle is the foundation of the world; the one who is in harmony is the way of the world. To achieve neutrality, the heaven will have its place and all things will be nurtured. \n\nZhongni said: "A gentleman is the mean, and a villain is against the mean. A gentleman is in the mean, and a gentleman is in the middle at all times; a villain is against the mean, and a villain has no scruples."\n\nConfucius said: "The mean is the end! People rarely last long. ! "\n\nConfucius said: "I know that the way is not good: those who know it can't do it, and the fool can't do it. I know that the way can't be done; the wise person can't do it, and the unworthy person can't do it."', 'embedding': {'__default__': [0.025426213, 0.034025285, -0.031320818, 0.016555943, 0.06301127, 0.0053874906, -0.03851282, 0.05617112, -0.026647208, -0.04409193, 0.07942905, -0.016034704, 0.0020301498,0.009465688, 0.022152705, -0.02026594, -0.05420044, 0.048457634, 0.029878676, -0.023549352, 0.016439304, 0.034842428, 0.020412622, -0.0025755076, 0.0027205113, -0.022589676, 0.020291883, 0.03922448, 0.009154288, -0.007037486, 0.018945666, -0.038242333, -0.0059177317, 0.035270065, 0.023194453, -0.024663404, 0.045518205, 0.078890085, -0.0049626785, -0.0013358546, -0.020847179, 0.008833306, -0.057070132, -0.046780605, 0.0070816474, 0.009134139, -0.019501934, -0.034442246, -0.03516072,0.006713291, -0.006998874, 0.032105204, 0.052466325, 0.02138813, -0.009691537, -0.009268791, -0.04773306, -0.01108876, -0.03728646, 0.06281455, -0.015532677, 0.012268486, 0.005829601, -0.0013577868, 0.006930408, -0.05296581, -0.023193905, -0.046643063, 0.011809876, 0.00023975057, 0.050946303, -0.03136553, -0.06390407, 0.03504418, -0.01805315, -0.04344443, 0.00677673, 0.013735469, 0.045064267, 0.029002164, 0.03938319, 0.034683738, 0.006368947, -0.00906264, 0.035292383, 0.058860265,-0.028741615, -0.0036960396, 0.012889149, -0.008738161, -0.012072681, 0.044848487, 0.047913834, 0.0025441165, 0.033541866, 0.0077599497, 0.025643757, -0.016167128, 0.02875357, -0.012898535, -0.03224147, -0.021762878, -0.015384127, -0.022707786, -0.0061772997, -0.020073174, 0.02529043, 0.014178167, -0.018244589, -0.008239636, 0.02126484, 0.0034560761, 0.051880326, 0.06363023, 0.01969195, -0.011307083, 0.005372883, -0.019629491, 0.0017945681, 0.011739278, -0.023668867, 0.009739918, 0.059829958,0.034342274, -0.0019169106, 0.03975126, 0.0023158423, -0.044209726, -0.0027910431, 0.00031879256, -0.040849816, 0.04776571, 0.0020381147, 0.01910589, 0.0027710563, 0.011146467, 0.020007843, 0.0042484123, 0.005176938, 0.021434778, 0.0777557, -0.0038533013, 0.047306042, 0.0034837353, 0.042881086, -0.06315707, -0.052171838, -0.04829033, 0.019041482, 0.031125337, 0.045959365, 0.04621173, -1.540408e-13, -0.006997954, -0.020921359, 0.037427846, -0.07051564, 0.00041891588, 0.026310083, 0.02499563,0.0017974637, 0.002034413, -0.023606593, 9.611427e-18, -0.009685626, -0.019306758, 0.033706345, 0.0356712, 0.039736345, -0.012066647, -0.0059178593, 0.031503696, 0.023679575, -0.011107136, -0.048489127, 0.0053288713, -0.0123187, -0.049243163, 0.00041808805, 0.0003394062, 0.015456655, -0.0028465304, 0.014629242, -0.011859338, -0.029523129, -0.026531836, -0.004875002, -0.009635127, 0.007637854, 0.016192164, -0.027906612, 0.0043651653, -0.011299037, 0.009702859, 0.020838758, 0.0027266808,0.013553497, -0.020599183, -0.052582372, 0.047934312, -0.0211804, -0.014517992, -0.06661431, 0.057032656, -0.048424397, 0.011499752, 0.0056307977, -0.016409324, 0.044111975, -0.04115124, -0.026549118, 0.01797299, 0.022250125, -0.020612499, 0.060236566, -0.006571854, -0.00393553, -0.031184353, 0.064215384, 0.046786495, -0.00046514318, -0.00701734, -0.02017527, 0.008758273, 0.0016055728, -0.02660426, 0.009220425, -0.014724887, 0.010843108, -0.044277404, 0.010815744, 0.034662224, -0.026707007,-0.005213057, -0.011781672, -0.019246181, -0.019320516, -0.0016598016, 0.005209462, 0.028500574, 0.013622352, -0.026014416, -0.03194754, -0.010626347, 0.070779, -0.02930651, -0.049735054, 0.019265305, -0.017812809, -0.03864494, -0.057057705, 0.047812644, -0.021030407, -0.06017914, -0.008436314, -0.00565766, 0.040196173, -0.0015437013, 0.03359702, 0.07130042, 0.019466452, 0.030842118, 0.030507375, -0.007607868, -0.014626105, -0.011076136, -2.4094108e-05, -0.0024054602, 0.007530281, 0.029631741,0.031458937, -0.006896458, 0.014952638, -0.0036152329, 0.027877389, 0.016777638, -0.021922763, -0.03586302, -0.0010009537, -0.016428495, 0.008924784, 0.023603037, -0.020115128, 0.00865001, -0.026730474, -0.033528563, 0.013149746, 0.030888822, -0.005040427, -0.0014151632, 0.020947043, 0.028981369, 0.00765713, 0.032147344, -0.023203162, -0.0044152536, -0.045422826, 0.016876237, 0.036611527, 0.013162975, 0.026166268, -0.027787978, 0.034219008, -0.013293831, -0.019005, 0.055061992, 0.00763016,0.032072656, 0.014239542, 0.032104347, -0.037449952, -0.02217896, -0.032320846, 0.017871961, -0.03970623, 0.026835883, -0.054085482, -0.0027009265, 0.07634273, -0.024354596, -0.0029545398, 0.059289258, -0.027541274, 0.019158483, -0.02926486, -0.059198033, -0.00045453524, 0.03309172, -0.058500547, 0.03847318, 0.06686244, 0.0359078, 0.037715927, 0.026004314, -0.010441379, -0.044468235, -0.04735834, -0.054638885, -0.003099882, -0.016880812, 0.00814928, -0.010210958, -0.022301093, 0.0029216122,-0.012461757, 0.030258488, 0.00953299, -0.03189832, 0.015131103, 0.0012498442, 0.014835826, 0.05411683, 0.020760238, 0.017031388, 0.020908138, 0.059172362, 0.033446126, -0.0024910853, -0.0605189, -0.0029194264, -0.001088384, -0.042831995, 0.008137215, 0.020813528, 0.037613075, -0.005095976, -0.01884459, -0.005923714, 0.023792582, -0.028474923, 0.03757395, 0.024344636, -0.035215013, -0.023457121, 0.052875917, 0.026036683, 0.10764514, -0.051340614, -0.018821908, 0.016108429, 0.024424875,-0.053446505, -0.0054197996, -0.0060087196, -0.0041229264, -0.006034636, 0.007538079, -0.018776668, -0.013380194, 0.0028277775, -0.0048381737, 0.004624616, -0.030349467, -0.04140086, 0.009394284, -0.008468681, -0.036008462, -0.011696614, -0.020296633, -0.030620858, -0.063259475, 0.038780067, -0.010312165, -0.053780414, -0.011514602, -0.04333725, -0.010768698, 0.0016200787, -0.029752402, 0.007386246, 0.018586209, 0.009053951, 0.00036854058, 0.022906242, 0.028303033, 0.019804642, 0.036445,0.005007112, 0.014926994, 0.029018547, -0.015011582, 0.0065884907, 0.029475305, -0.016382167, 0.007041045, -0.024008943, 0.046547353, 0.027779775, -0.0071733906, 0.012406345, 0.011313683, 0.03918696, -0.056599338, -0.034151986, 0.047533363, -0.01459687, -0.0679629, -0.011175669, 0.030152632, 0.0382269, 0.0041881935, -0.007858538, -0.018000403, 0.022335991, -0.003267302, -0.048622385, 0.023882171, -0.034625404, -0.035592694, -0.042968493, -0.004153031, 0.019842803, -0.0016533697, -0.004881279,-0.045086186, -0.0011175375, 0.0021369208, -0.027217858, 0.042248532, -0.030293925, 0.064989135, -0.029612714, 0.024552464, -0.033408698, -0.0101805655, -0.0029412394, -0.0915347, 0.04138679, 0.013030276, 0.06705817, -0.004188915, -0.0062264856, -0.051195238, -0.02774683, 0.00731135, -0.013503835, -0.052448288, 0.021496458, 0.019841682, -0.023670819, 0.009778803, -0.0014320388, 0.014062996, 0.07120368, -0.04988873, -0.00012608718, -0.04031885, 0.018206028, 0.0640813, 0.070187114,-0.0015529695, 0.020645538, 0.0052610333, -0.043595728, 0.061265957, -0.042493634, -0.04516448, -0.005089449, 0.010595472, -0.015604816, 0.03951871, 0.03183545, -0.016843012, 0.0235672, -0.040382978, -0.03078044, 0.014075101, -0.0014767948, 0.042445626, 0.026724413, 0.017799338, -0.08596669, -0.018500224, 0.051851522, -0.009828736, -0.029204156, 0.011474476, 0.0055204323, -0.004064649, -0.033348784, 0.00523749, 0.022435933, -0.046652254, 0.024559079, -0.016480578, 0.040265355, -0.03305009,-0.031315356, -0.010046829, -0.01731933, -0.020532912, -0.0029909958, 0.011842756, -0.048824314, -0.012646131, -0.055367026, 0.0031566594, 0.014152518, -0.01904883, -0.024933103, 0.04189374, 0.028184457, 0.025812266, 0.02407379, 0.022633042, -0.007829883, -0.014562866, -0.032446302, 0.008651778, 0.026188472, 0.04299915, -0.02544324, 0.021127068, -0.018766753, -0.0063632224, 0.025388861, 0.0063838237, -0.02366371, -0.017125644, 0.0014911906, -0.022412458, 0.009111594, -0.008172784,-0.008471129, -0.023930583, -0.05943837, 0.00966703, -0.006407053, 0.012494917, -0.012754937, -0.026357513, 0.0407052, -0.024343831, -0.0037548298, 0.029267209, -0.008125634, -0.010630878, 0.02786125, -0.007368123, 0.0062804264, -0.032700405, -0.020568395, -0.043276288, 0.015865225, 0.0311623, -0.043471303, 0.055800557, -0.011932295, -0.045096148, -0.0050109215, -0.011723784, -0.005000813, 0.003097428, 0.0811478, 0.040164344, 0.0036458236, 0.012483547, 0.010060236, 0.0023018436, 0.006584019,0.024982, 0.03525057, 0.0041822596, -0.015556074, 0.0034614399, 0.038330406, -0.00889707, 0.03221703, 0.03959702, 6.4116706e-05, 0.0036850327, 0.038901724, -0.028086498, 0.010405911, 0.021698542, -0.015949449, 0.005926045, 0.033439852, 0.0038825811, -0.025638554, -0.023800563, 0.0133479, 0.0346594, -0.00953549, -0.047149554, 0.024543112, 0.017636258, -0.013193999, 0.017017592, 0.028411236, -0.0025103707, 0.018269507, 0.051845845, 0.062203504, -0.025815979, -0.051945943, 0.0033721116,0.07998744, 0.025422625, 0.009253838, -0.03604736, -0.042337134, 0.014561826, 0.04507746, -0.020416573, -0.015352677, -0.036153104, -0.027990274, 0.01924466, -0.045471456, -0.008539538, -0.040453684, -0.05723972, -0.01498072, -0.030530391, -0.007754218, -0.039093077, -0.03979653, -0.02228624, 0.0008455001, 0.0071356776, 0.025943242, 0.11981686, 0.014268126, -0.05869411, 0.008343074, -0.040158387, -0.016552536, 0.018982463, -0.03898772, -0.007980556, 0.022687193, -0.051658224, 0.025367606,0.01998329, 0.021392373, -0.0434203, 0.02121578, 0.018556163, 0.003245211, 0.018774172, 0.0005500793, -0.03479759, 0.035857406, 0.00040770217, -0.035999756, 0.028827438, -0.029466469, -0.047930814, 0.0265652, -0.05008721, -0.0010059318, -0.01251698, 0.003447119, 0.05898002, 0.05744354, 0.06840657, 0.007024658, 0.006566852, 0.030507237, -0.041550875, -0.04153647, 0.0010598813, -0.0794358, -0.041427277, -0.06114106, -0.019902522, -0.012520727, -0.01163048, 0.0007943832, 0.028543016,-0.009956491, 0.025655229, 0.043709233, 0.0011159213, -0.030077763, 0.039286032, -0.03297622, 0.0095170215, -0.039622717, 0.002401525, 0.021634275, -0.057686023, 0.04041061, -0.03395288, -0.04493925, 0.028587993, 0.046589497, 0.03245823, -0.02444273, 0.036353294, 0.008618879, 0.0035412086, -0.046254568, 0.0148467375, 0.016634878, 0.0306935, 0.011313652, -0.031087654, 0.009768224, 0.007366226, -0.024521042, 0.013069051, -0.036803946, 0.045954864, 0.00825386, 0.013556772, -0.028380545,0.01906237, -0.017373545, 0.02040689, -0.004991811, -0.011323219, 0.038479466, -0.010787252, -0.019826626, 0.06550503, -0.021512486, 0.04311003, 0.06399184, 0.01899978, -0.06393204, -0.052857753, 0.0026758595, -0.023297222, 0.0063398667, 0.029147884, 0.0076334844, 0.029746521, -0.016748957, -0.0021633625, 0.021256661, 0.022264387, -0.022884088, -0.016388448, -0.033683803, -0.02238929, 0.055876125, -0.0052925227, 0.013481841, -0.0209584, -0.03506728, 0.016223963, 0.0012435996, -0.046723302,-0.0005387966, 0.012555997, 0.04083436, -0.012232332, -0.016081795, 0.0062284134, -0.027968781, -0.022376344, 0.02119133, 0.034631502, 0.02318447, 0.044668313, 0.0023679107, -0.04467385, -0.01789442, 2.5628348e-09, -0.010842855, 0.026404202, 0.018929183, -0.008009062, -0.03902315, -0.03539446, 0.045142844, 0.0016768258, -0.0059571, 0.048917703, 0.05116004, 0.0005371061, 0.004794815, -0.016625853, -0.06221956, 0.0022883292, -0.037917495, -0.026748385, -0.01792442, -0.0045090592, 0.034487702,0.029278198, -0.035341892, 0.019185508, 0.023990307, 0.040781856, 0.033953443, -0.012868148, 0.052692164, 0.005785138, -0.005816314, -0.017770205, 0.052668773, -0.011497479, -0.0042478456, -0.041163344, 0.017727418, 0.019524269, 0.0028672623, 0.027206969, 0.0098760445, -0.01839258, 0.064081036, -0.018610574, -0.03167949, 0.028112203, 0.038009446, 0.027018046, 0.034964353, 0.06139204, -0.000934609, -0.013590096, -0.0112873, -0.016028268, -0.06521579, 0.009233209, -0.010975224, 0.014953974,-0.013712477, -0.035961337, -0.01200976, 0.016888669, -0.0254347, -0.03294505, -0.011085823, 0.074079536, -0.041230906, 0.008066781, 0.016872587, -0.009588438, 0.015129394, 0.004597649, -0.012082546, -0.0205048, -0.011996115, -0.04258334, 0.01090112, 0.025189904, -0.026545173, 0.011398977, -0.035342906, 0.011842086, -0.046653755, 0.0023188903, -0.027082337, -0.053023692, 0.009047406, -0.04184604, -0.0027126037, 0.022172572, -0.0019665314, -0.020220164, -0.0055029565, 0.059257552, -0.038126208,-0.03557876, 0.01125294, 0.024858668, -0.0056422777, 0.0049045635, 0.011553989, 0.026200697, -1.432319e-27, -0.0010860946, 0.09268514, 0.06090386, -0.005796361, 0.025179828, 0.06804877, -0.050707188, 0.02777679, -0.031054588, -0.019260377, -0.04869623, 0.04939678, -0.024753766, -0.0072507886, 0.06201284, 0.038341094, -0.023940781, -0.0044804853, 0.0602865, 0.012955638, 0.041328467, 0.012215266, 0.041857783, 0.05578706, -0.04423121, -0.030089064, 0.039828192, -0.025488378, 0.044566587,0.01636707, 0.05779351, -0.029481513, 0.0053500645, 0.051323455, 0.03458597, 0.013696377, -0.035192043, 0.04330355, 0.058219656, 0.049582005, -0.02016202, -0.036638554, 0.020224864, -0.039496087, -0.015062504, 0.0016461068, 0.015035624, -0.03048053, -0.027442321, -0.019795708, 0.061282612, -0.0035766829, 0.049931448, -0.044324502, -0.034937955, -0.055052124, -0.002334567, 0.0011809984, -0.0005358032, 0.014579384, -0.006727375, 0.01961919, -0.024230124, 0.022260979, 0.041638635, 0.015936956,-0.0129846865, -0.028698431, -0.023311313, 0.020681119, -0.022004813, 0.0045342017, 0.10454673, -0.00907232, 0.018019272, 0.027693322, -0.045678418, 0.008820441, -0.018247915, 0.0022897322, -0.0045168954, 0.0330645, -0.046881303, -0.02397729, -0.03653085, 0.03674319, -0.027837344, -0.02493074, -0.025111457, -0.02385346, -0.026238898, 0.019991593, -0.00067811715, -0.014781632, 0.024634153, -0.016004443, -0.0052194847, 0.024627017, 0.0032142564, 0.013624659, -5.867911e-24, -9.8801356e-05,0.07649344, 0.02002069, 0.029181154, 0.03094678, 0.021331448, 0.029769164, -0.01106459, 0.035867818, 0.0059047374, 0.044848107, -0.034437712, 0.056731243, 0.0032158946, 0.018998314, -0.070217706, -0.029742291, 0.034644052, 0.024905458]}, 'metadata': {}}]

You can see that the results are output in the form of dict. Similarly, at this time, join is either not set, or set to False, which has no effect on the result. It cannot be set to True or a string, otherwise an error will be reported.

ppl.reranker = Reranker(name="ModuleReranker", model=lazyllm.OnlineEmbeddingModule(type="rerank", source="glm", embed_model_name="rerank"), topk=2, output_format="dict", join=True) | bind(query=ppl.input)

The error result is as follows:

AssertionError: Only content output can be joined

Next we set output_format to content:

ppl.reranker = Reranker(name="ModuleReranker", model=lazyllm.OnlineEmbeddingModule(type="rerank", source="glm", embed_model_name="rerank"), topk=2, output_format="content" ) | bind(query=ppl.input)

The result is as follows:

nodes: ['The way of observing the sky and the journey of adhering to the sky have been completed. \nTherefore, there are five thieves in heaven. Those who see them will prosper. \nThe five thieves are in the heart and execute them in heaven. \nThe universe depends on the hands, and everything depends on the body. \nIt is human nature and scheming. The way of establishing heaven is used to determine people. \nThe sky sends murderous intentions, and the stars move to their places; the earth sends murderous intentions, and dragons and snakes rise from the ground; humans send murderous intentions, and heaven and earth turn over and over again; heaven and man unite, and everything is established. \nSex is cunning and can be hidden. The evil of the nine orifices depends on the three essentials, which can cause movement and stillness. \nFire is born in wood, and disaster will be overcome; treachery is born in a country, and it will collapse in time. Those who know how to practice are called saints. \nHe is born to be killed by nature, which is the principle of Tao. The thief of all things in heaven and earth, the thief of all things by man, the thief of man and all things. The three thieves are both suitable and the three talents are safe. \nTherefore, it is said that if you eat at the right time, everything will be in order; if you use its mechanism, everything will be at peace. People know the gods of their gods, but they don’t know the reason why they are gods. \nThe sun and the moon are numbered, the size is determined, the divine power is born, and the gods are born. \nNo one in the world can see or know about his stolen machine. A gentleman will have a firm bow, while a villain will have a light life. \nThe blind person is good at hearing, and the deaf person is good at seeing. One source of absolute profit, use the teacher ten times. Return to day and night three times, and use your troops ten thousand times. \nThe heart is born from things and dies from things. The machine is in the eyes. \nHeaven is unkind but has great kindness. Thunder and strong wind are all stupid. \nThe most joyful nature, the most tranquil nature and incorruptibility. Heaven is supremely private, and its use is supremely public. \nThe control of birds lies in Qi. The living is the root of death, and the dead is the root of life. Grace is born from harm, harm is born from grace. \nFools regard heaven and earth, culture and science as sages, I regard time, objects, culture and philosophy as sages. ', 'Destiny is called nature, nature is called Tao, and cultivating Tao is called teaching. \nThe Tao cannot be separated in a moment (yú); it can be separated, but it is not the Tao. Therefore, a gentleman should be cautious about what he does not see and be afraid of what he does not hear. Don't see (xian) what is hidden, don't show what is subtle, so a gentleman should be cautious about being alone. \nWhen joy, anger, sorrow, and joy have not arisen, it is called the middle; when they arise, they are all in the middle, it is called the harmony. The one who is in the middle is the foundation of the world; the one who is in harmony is the way of the world. To achieve neutrality, the heaven will have its place and all things will be nurtured. \n\nZhongni said: "A gentleman is the mean, and a villain is against the mean. A gentleman is in the mean, and a gentleman is in the middle at all times; a villain is against the mean, and a villain has no scruples."\n\nConfucius said: "The mean is the end! People rarely last long. ! "\n\nConfucius said: "I know that the way is not good: those who know it can't do it, and the fool can't do it. I know that the way can't be done; the wise person can't do it, and the unworthy person can't do it."']

The result is output in the form of List[str].

What if join is set to True?

ppl.reranker = Reranker(name="ModuleReranker", model=lazyllm.OnlineEmbeddingModule(type="rerank", source="glm", embed_model_name="rerank"), topk=2, output_format="content", join=True) | bind(query=ppl.input)

The result is as follows:

nodes: The way of observing the sky and the journey of adhering to the sky are complete.
Therefore, there are five thieves in heaven, and those who see them will prosper.
The five thieves are in the heart and execute them in heaven.
The universe depends on the hands, and all transformations occur on the body.
Human beings are born with human nature and are scheming. The way of establishing heaven is used to determine people.
The sky sends murderous intentions, and the stars move to their places; the earth sends murderous intentions, and dragons and snakes rise from the ground; humans send murderous intentions, and heaven and earth turn over and over again; the combination of heaven and man creates the basis for all transformations.
Nature is clever and can be hidden. The evil of the nine orifices depends on the three essentials, which can cause movement and stillness.
If fire is born in wood, disaster will be overcome; if treachery is born in the country, it will collapse in time. Those who know how to practice are called saints.
God is born to kill, this is the principle of Tao. The thief of all things in heaven and earth, the thief of all things by man, the thief of man and all things. The three thieves are both suitable and the three talents are safe.
Therefore, it is said that if you eat at the right time, everything will be in order; if you move its mechanism, everything will be in peace. People know the gods of their gods, but they don’t know the reason why they are gods.
The sun and the moon are numbered, the size is determined, the divine power is generated, and the gods are revealed.
No one in the world can see or know how to steal a machine. A gentleman will have a firm bow, while a villain will have a light life.
The blind person is good at hearing, and the deaf person is good at seeing. One source of absolute profit, use the teacher ten times. Return to day and night three times, and use your troops ten thousand times.
The heart is born from things and dies from things. The mechanism lies in the eyes.
God is unfaithful but has great kindness. Thunder and strong wind are all stupid.
The most joyful nature, the most tranquil nature and incorruptibility. Heaven is supremely private, and its use is supremely public.
The control of birds lies in Qi. The living is the root of death, and the dead is the root of life. Grace is born from harm, harm is born from grace.
Fools regard heaven and earth, culture and science as sages, I regard time, objects, culture and philosophy as sages. Destiny is called nature, nature is called Tao, and cultivating Tao is called teaching.
The Tao cannot be separated for a moment; the Tao can be separated, but it is not the Tao. Therefore, a gentleman should be cautious about what he does not see and be afraid of what he does not hear. Don't see (xian) what is hidden, don't show what is subtle, so a gentleman should be cautious about being alone.
When joy, anger, sorrow, and joy are not expressed, it is called the middle; when they are expressed, they are all in the middle, it is called the harmony. The one who is in the middle is the foundation of the world; the one who is in harmony is the way of the world. To achieve neutrality, the heaven will have its place and all things will be nurtured.

Zhongni said: "A gentleman is moderate, and a villain is against the mean. A gentleman is moderate, and a gentleman is in the right time; a villain is against the mean, and a villain has no scruples."

Confucius said: "The golden mean is the best! The people can survive for a long time!"

Confucius said: "I know that the way is not good: those who know it are better than it, and those who are fools are not as good as it. I know that the way is unclear: the wise people are beyond it, and the unworthy people are not as good as it. People all eat and drink, and they rarely know the taste."

You can see that the result is output in the form of a string. What if join is set to '11111111111111111111111111111'?

ppl.reranker = Reranker(name="ModuleReranker", model=lazyllm.OnlineEmbeddingModule(type="rerank", source="glm", embed_model_name="rerank"), topk=2, output_format="content", join='11111111111111111111111111111') | bind(query=ppl.input)

The result looks like this:

nodes: The way of observing the sky and the journey of adhering to the sky are complete.
Therefore, there are five thieves in heaven, and those who see them will prosper.
The five thieves are in the heart and execute them in heaven.
The universe depends on the hands, and all transformations occur on the body.
Human beings are born with human nature and are scheming. The way of establishing heaven is used to determine people.
The sky sends murderous intentions, and the stars move to their places; the earth sends murderous intentions, and dragons and snakes rise from the ground; humans send murderous intentions, and heaven and earth turn over and over again; the combination of heaven and man creates the basis for all transformations.
Nature is clever and can be hidden. The evil of the nine orifices depends on the three essentials, which can cause movement and stillness.
If fire is born in wood, disaster will be overcome; if treachery is born in the country, it will collapse in time. Those who know how to practice are called saints.
God is born to kill, this is the principle of Tao. The thief of all things in heaven and earth, the thief of all things by man, the thief of man and all things. The three thieves are both suitable and the three talents are safe.
Therefore, it is said that if you eat at the right time, everything will be in order; if you move its mechanism, everything will be in peace. People know the gods of their gods, but they don’t know the reason why they are gods.
The sun and the moon are numbered, the size is determined, the divine power is generated, and the gods are revealed.
No one in the world can see or know how to steal a machine. A gentleman will have a firm bow, while a villain will have a light life.
The blind person is good at hearing, and the deaf person is good at seeing. One source of absolute profit, use the teacher ten times. Return to day and night three times, and use your troops ten thousand times.
The heart is born from things and dies from things. The mechanism lies in the eyes.
God is unfaithful but has great kindness. Thunder and strong wind are all stupid.
The most joyful nature, the most tranquil nature and incorruptibility. Heaven is supremely private, and its use is supremely public.
The control of birds lies in Qi. The living is the root of death, and the dead is the root of life. Grace is born from harm, harm is born from grace.
Fools regard heaven and earth, culture and science as sages, I regard time, objects, culture and philosophy as sages. 111111111111111111111111111111 Destiny is called nature, nature is called Tao, and cultivating Tao is called teaching.
The Tao cannot be separated for a moment; the Tao can be separated, but it is not the Tao. Therefore, a gentleman should be cautious about what he does not see and be afraid of what he does not hear. Don't see (xian) what is hidden, don't show what is subtle, so a gentleman should be cautious about being alone.
When joy, anger, sorrow, and joy are not expressed, it is called the middle; when they are expressed, they are all in the middle, it is called the harmony. The one who is in the middle is the foundation of the world; the one who is in harmony is the way of the world. To achieve neutrality, the heaven will have its place and all things will be nurtured.

Zhongni said: "A gentleman is moderate, and a villain is against the mean. A gentleman is moderate, and a gentleman is in the right time; a villain is against the mean, and a villain has no scruples."

Confucius said: "The golden mean is the best! The people can survive for a long time!"

Confucius said: "I know that the way is not good: those who know it are better than it, and those who are fools are not as good as it. I know that the way is unclear: the wise people are beyond it, and the unworthy people are not as good as it. People all eat and drink, and they rarely know the taste."

It can be seen that the result is composed of two text segments. It can be seen that the post-processing of Retriever and Reranker is the same.

[How to download the data set]