complete RAG, History RAG

.config/config.properties

@@ -1 +1 @@
-API_KEY=9d2d36b2c8b20bb329fd44fe058b7ac2.UF14LzIdcAH2Ob21
+API_KEY=86df1c93a84174062eaf6a38e331efc1.AJV5y2jkKASbZ3YD

.idea/workspace.xml

@@ -4,7 +4,11 @@
     <option name="autoReloadType" value="SELECTIVE" />
   </component>
   <component name="ChangeListManager">
-    <list default="true" id="db7f4434-7cc2-4790-b42e-92159334a8b4" name="Changes" comment="second commit" />
+    <list default="true" id="db7f4434-7cc2-4790-b42e-92159334a8b4" name="Changes" comment="complete RAG, History RAG">
+      <change beforePath="$PROJECT_DIR$/.idea/workspace.xml" beforeDir="false" afterPath="$PROJECT_DIR$/.idea/workspace.xml" afterDir="false" />
+      <change beforePath="$PROJECT_DIR$/requirements.txt" beforeDir="false" afterPath="$PROJECT_DIR$/requirements.txt" afterDir="false" />
+      <change beforePath="$PROJECT_DIR$/src/serve/RAG.py" beforeDir="false" afterPath="$PROJECT_DIR$/src/serve/RAG.py" afterDir="false" />
+    </list>
     <option name="SHOW_DIALOG" value="false" />
     <option name="HIGHLIGHT_CONFLICTS" value="true" />
     <option name="HIGHLIGHT_NON_ACTIVE_CHANGELIST" value="false" />
@@ -26,6 +30,20 @@
     <option name="RECENT_GIT_ROOT_PATH" value="$PROJECT_DIR$" />
     <option name="RESET_MODE" value="HARD" />
   </component>
+  <component name="GitHubPullRequestSearchHistory">{
+  &quot;lastFilter&quot;: {
+    &quot;state&quot;: &quot;OPEN&quot;,
+    &quot;assignee&quot;: &quot;YDzzz&quot;
+  }
+}</component>
+  <component name="GithubPullRequestsUISettings">
+    <option name="selectedUrlAndAccountId">
+      <UrlAndAccount>
+        <option name="accountId" value="c877182e-149b-4367-a384-4c094fd3c76e" />
+        <option name="url" value="https://github.com/YDzzz/LLM.git" />
+      </UrlAndAccount>
+    </option>
+  </component>
   <component name="MarkdownSettingsMigration">
     <option name="stateVersion" value="1" />
   </component>
@@ -42,11 +60,13 @@
   </component>
   <component name="PropertiesComponent"><![CDATA[{
   "keyToString": {
+    "Python.HistoryRAG.executor": "Run",
     "Python.RAG.executor": "Run",
     "Python.embedding.executor": "Run",
     "RunOnceActivity.OpenProjectViewOnStart": "true",
     "RunOnceActivity.ShowReadmeOnStart": "true",
-    "git-widget-placeholder": "91444e75",
+    "com.google.cloudcode.ide_session_index": "20240518_0001",
+    "git-widget-placeholder": "detached",
     "last_opened_file_path": "G:/code/py/LLM",
     "node.js.detected.package.eslint": "true",
     "node.js.detected.package.tslint": "true",
@@ -84,6 +104,12 @@
       <workItem from="1715345875476" duration="11056000" />
       <workItem from="1715441734127" duration="1201000" />
       <workItem from="1715516642242" duration="6557000" />
+      <workItem from="1715878258483" duration="36000" />
+      <workItem from="1715878326308" duration="244000" />
+      <workItem from="1715878583964" duration="209000" />
+      <workItem from="1715935000078" duration="18704000" />
+      <workItem from="1715964027274" duration="201000" />
+      <workItem from="1715964238142" duration="2613000" />
     </task>
     <task id="LOCAL-00001" summary="init">
       <option name="closed" value="true" />
@@ -109,7 +135,15 @@
       <option name="project" value="LOCAL" />
       <updated>1715530309702</updated>
     </task>
-    <option name="localTasksCounter" value="4" />
+    <task id="LOCAL-00004" summary="complete RAG, History RAG">
+      <option name="closed" value="true" />
+      <created>1715963552332</created>
+      <option name="number" value="00004" />
+      <option name="presentableId" value="LOCAL-00004" />
+      <option name="project" value="LOCAL" />
+      <updated>1715963552332</updated>
+    </task>
+    <option name="localTasksCounter" value="5" />
     <servers />
   </component>
   <component name="TypeScriptGeneratedFilesManager">
@@ -126,7 +160,7 @@
                   <entry key="branch">
                     <value>
                       <list>
-                        <option value="master" />
+                        <option value="detached" />
                       </list>
                     </value>
                   </entry>
@@ -141,10 +175,12 @@
   <component name="VcsManagerConfiguration">
     <MESSAGE value="init" />
     <MESSAGE value="second commit" />
-    <option name="LAST_COMMIT_MESSAGE" value="second commit" />
+    <MESSAGE value="complete RAG, History RAG" />
+    <option name="LAST_COMMIT_MESSAGE" value="complete RAG, History RAG" />
   </component>
   <component name="com.intellij.coverage.CoverageDataManagerImpl">
-    <SUITE FILE_PATH="coverage/LLM$RAG.coverage" NAME="RAG Coverage Results" MODIFIED="1715346172432" SOURCE_PROVIDER="com.intellij.coverage.DefaultCoverageFileProvider" RUNNER="coverage.py" COVERAGE_BY_TEST_ENABLED="true" COVERAGE_TRACING_ENABLED="false" WORKING_DIRECTORY="$PROJECT_DIR$/src/serve" />
-    <SUITE FILE_PATH="coverage/LLM$embedding.coverage" NAME="embedding Coverage Results" MODIFIED="1715354020292" SOURCE_PROVIDER="com.intellij.coverage.DefaultCoverageFileProvider" RUNNER="coverage.py" COVERAGE_BY_TEST_ENABLED="true" COVERAGE_TRACING_ENABLED="false" WORKING_DIRECTORY="$PROJECT_DIR$/src/serve" />
+    <SUITE FILE_PATH="coverage/LLM$RAG.coverage" NAME="RAG Coverage Results" MODIFIED="1715966784292" SOURCE_PROVIDER="com.intellij.coverage.DefaultCoverageFileProvider" RUNNER="coverage.py" COVERAGE_BY_TEST_ENABLED="true" COVERAGE_TRACING_ENABLED="false" WORKING_DIRECTORY="$PROJECT_DIR$/src/serve" />
+    <SUITE FILE_PATH="coverage/LLM$embedding.coverage" NAME="embedding Coverage Results" MODIFIED="1715952325378" SOURCE_PROVIDER="com.intellij.coverage.DefaultCoverageFileProvider" RUNNER="coverage.py" COVERAGE_BY_TEST_ENABLED="true" COVERAGE_TRACING_ENABLED="false" WORKING_DIRECTORY="$PROJECT_DIR$/src/serve" />
+    <SUITE FILE_PATH="coverage/LLM$HistoryRAG.coverage" NAME="HistoryRAG Coverage Results" MODIFIED="1715963407636" SOURCE_PROVIDER="com.intellij.coverage.DefaultCoverageFileProvider" RUNNER="coverage.py" COVERAGE_BY_TEST_ENABLED="true" COVERAGE_TRACING_ENABLED="false" WORKING_DIRECTORY="$PROJECT_DIR$/src/serve" />
   </component>
 </project>

requirements.txt

@@ -1,7 +1,7 @@
-langchain~=0.1.19
-httpx
-pyjwt
-transformers
-text2vec~=1.2.9
-chardet
-pypdf~=4.2.0
+langchain>=0.1.20
+pypdf>=4.2.0
+zhipuai~=2.0.1.20240423.1
+langchainhub~=0.1.15
+chromadb
+httpx-sse~=0.4.0
+chardet~=5.2.0

src/base/PdfLoader.py

@@ -1,15 +0,0 @@
-from langchain_community.document_loaders import PyPDFLoader
-
-
-class PDFLoader:
-    @staticmethod
-    def loader(file_path: str):
-        content = ""
-        loader = PyPDFLoader(file_path)
-        for page in loader.load():
-            content += page.page_content
-        return content
-
-
-if __name__ == '__main__':
-    print(PDFLoader.loader("C:\\Users\\16922\\Desktop\\文档1.pdf"))

src/base/README.md

@@ -1,10 +0,0 @@
-## Document Loader Interface
-| Method Name | Explanation                                                                                                                 |
-|-------------|-----------------------------------------------------------------------------------------------------------------------------|
-| lazy_load	  | Used to load documents one by one lazily. Use for production code.                                                          |
-| alazy_load  | Async variant of lazy_load                                                                                                  |
-| load        | Used to load all the documents into memory eagerly. Use for prototyping or interactive work.                                |
-| aload	      | Used to load all the documents into memory eagerly. Use for prototyping or interactive work. Added in 2024-04 to LangChain. |
-- The load methods is a convenience method meant solely for prototyping work – it just invokes list(self.lazy_load()).
-- The alazy_load has a default implementation that will delegate to lazy_load. If you’re using async, we recommend overriding the default implementation and providing a native async implementation.
-

src/base/ZhuPuAiEmbeddings.py

@@ -0,0 +1,47 @@
+import os
+from abc import ABC
+from typing import List
+
+from zhipuai import ZhipuAI
+
+from src.init.property import Property
+
+from langchain_core.embeddings import Embeddings
+
+os.environ["ZHIPUAI_API_KEY"] = Property.get_property("API_KEY")
+client = ZhipuAI()
+
+
+def _text_qualify(embedding_text):
+    """
+    using ZhipuAI Embedding API to get embedding 1024 dimension support
+    :param embedding_text:
+    :return:
+    """
+    if type(embedding_text) == str:
+        e_t = embedding_text
+    else:
+        e_t = embedding_text.page_content
+    # print usage of token number:
+    #   response.usage.total_tokens
+    # embedding support:
+    #   response.data[0].embedding
+    response = client.embeddings.create(
+        model="embedding-2",
+        input=e_t,
+    )
+    return response.data[0].embedding
+
+
+class ZhuPuAiEmbedding(Embeddings, ABC):
+    def embed_documents(self, texts: List[str]) -> List[List[float]]:
+        embeddings = []
+        for i in texts:
+            embeddings.append(_text_qualify(i))
+        return embeddings
+
+    def embed_query(self, text: str) -> List[float]:
+        return _text_qualify(text)
+
+    def __init__(self):
+        super().__init__()

src/base/chinese_text_splitter.py

@@ -1,59 +0,0 @@
-from langchain.text_splitter import CharacterTextSplitter
-import re
-from typing import List
-
-
-class ChineseTextSplitter(CharacterTextSplitter):
-    def __init__(self, pdf: bool = False, sentence_size: int = 250, **kwargs):
-        super().__init__(**kwargs)
-        self.pdf = pdf
-        self.sentence_size = sentence_size
-
-    def split_text1(self, text: str) -> List[str]:
-        if self.pdf:
-            text = re.sub(r"\n{3,}", "\n", text)
-            text = re.sub('\s', ' ', text)
-            text = text.replace("\n\n", "")
-        sent_sep_pattern = re.compile('([﹒﹔﹖﹗．。！？]["’”」』]{0,2}|(?=["‘“「『]{1,2}|$))')  # del ：；
-        sent_list = []
-        for ele in sent_sep_pattern.split(text):
-            if sent_sep_pattern.match(ele) and sent_list:
-                sent_list[-1] += ele
-            elif ele:
-                sent_list.append(ele)
-        return sent_list
-
-    def split_text(self, text: str) -> List[str]:   ##此处需要进一步优化逻辑
-        if self.pdf:
-            text = re.sub(r"\n{3,}", r"\n", text)
-            text = re.sub('\s', " ", text)
-            text = re.sub("\n\n", "", text)
-
-        text = re.sub(r'([;；.!?。！？\?])([^”’])', r"\1\n\2", text)  # 单字符断句符
-        text = re.sub(r'(\.{6})([^"’”」』])', r"\1\n\2", text)  # 英文省略号
-        text = re.sub(r'(\…{2})([^"’”」』])', r"\1\n\2", text)  # 中文省略号
-        text = re.sub(r'([;；!?。！？\?]["’”」』]{0,2})([^;；!?，。！？\?])', r'\1\n\2', text)
-        # 如果双引号前有终止符，那么双引号才是句子的终点，把分句符\n放到双引号后，注意前面的几句都小心保留了双引号
-        text = text.rstrip()  # 段尾如果有多余的\n就去掉它
-        # 很多规则中会考虑分号;，但是这里我把它忽略不计，破折号、英文双引号等同样忽略，需要的再做些简单调整即可。
-        ls = [i for i in text.split("\n") if i]
-        for ele in ls:
-            if len(ele) > self.sentence_size:
-                ele1 = re.sub(r'([,，.]["’”」』]{0,2})([^,，.])', r'\1\n\2', ele)
-                ele1_ls = ele1.split("\n")
-                for ele_ele1 in ele1_ls:
-                    if len(ele_ele1) > self.sentence_size:
-                        ele_ele2 = re.sub(r'([\n]{1,}| {2,}["’”」』]{0,2})([^\s])', r'\1\n\2', ele_ele1)
-                        ele2_ls = ele_ele2.split("\n")
-                        for ele_ele2 in ele2_ls:
-                            if len(ele_ele2) > self.sentence_size:
-                                ele_ele3 = re.sub('( ["’”」』]{0,2})([^ ])', r'\1\n\2', ele_ele2)
-                                ele2_id = ele2_ls.index(ele_ele2)
-                                ele2_ls = ele2_ls[:ele2_id] + [i for i in ele_ele3.split("\n") if i] + ele2_ls[
-                                                                                                       ele2_id + 1:]
-                        ele_id = ele1_ls.index(ele_ele1)
-                        ele1_ls = ele1_ls[:ele_id] + [i for i in ele2_ls if i] + ele1_ls[ele_id + 1:]
-
-                id = ls.index(ele)
-                ls = ls[:id] + [i for i in ele1_ls if i] + ls[id + 1:]
-        return ls

src/base/embedding.py

@@ -0,0 +1,26 @@
+from langchain_community.document_loaders import PyPDFLoader
+from langchain_community.vectorstores import Chroma
+from langchain_text_splitters import RecursiveCharacterTextSplitter
+
+from src.base.ZhuPuAiEmbeddings import ZhuPuAiEmbedding
+
+
+class SentenceEmbedding:
+    def __init__(self, file_path: str):
+        self.file_path = file_path
+        docs = PyPDFLoader(file_path).load()
+        text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200, add_start_index=True)
+        sentences = text_splitter.split_documents(docs)
+        self.vectorstore = Chroma.from_documents(sentences, ZhuPuAiEmbedding())
+
+    def get_vectorstore(self):
+        return self.vectorstore
+
+    def search(self, query: str) -> str:
+        docs = self.vectorstore.similarity_search(query)
+        return docs[0].page_content
+
+
+if __name__ == '__main__':
+    a = SentenceEmbedding("C:\\Users\\16922\\Desktop\\文档1.pdf")
+    print(a.search("We will dedicate a segment of our project to discussing these ethical issues"))

src/serve/HistoryRAG.py

@@ -0,0 +1,66 @@
+from langchain.chains.combine_documents import create_stuff_documents_chain
+from langchain.chains.history_aware_retriever import create_history_aware_retriever
+from langchain.chains.retrieval import create_retrieval_chain
+from langchain_core.messages import HumanMessage
+from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
+
+from src.serve.RAG import RAG
+
+__contextualize_q_system_prompt = """Given a chat history and the latest user question \
+which might reference context in the chat history, formulate a standalone question \
+which can be understood without the chat history. Do NOT answer the question, \
+just reformulate it if needed and otherwise return it as is."""
+_contextualize_q_prompt = ChatPromptTemplate.from_messages(
+    [
+        ("system", __contextualize_q_system_prompt),
+        MessagesPlaceholder("chat_history"),
+        ("human", "{input}"),
+    ]
+)
+
+__qa_system_prompt = """You are an assistant for question-answering tasks. \
+Use the following pieces of retrieved context to answer the question. \
+If you don't know the answer, just say that you don't know. \
+Use three sentences maximum and keep the answer concise.\
+
+{context}"""
+_qa_prompt = ChatPromptTemplate.from_messages(
+    [
+        ("system", __qa_system_prompt),
+        MessagesPlaceholder("chat_history"),
+        ("human", "{input}"),
+    ]
+)
+
+
+class HistoryRAG(RAG):
+
+    def __init__(self, file_path):
+        global _contextualize_q_prompt
+        global _qa_prompt
+        super().__init__(file_path)
+        self.__chat_history = []
+
+        history_aware_retriever = create_history_aware_retriever(
+            HistoryRAG._llm, self._retriever, _contextualize_q_prompt
+        )
+        question_answer_chain = create_stuff_documents_chain(HistoryRAG._llm, _qa_prompt)
+        self.__rag_chain = create_retrieval_chain(history_aware_retriever, question_answer_chain)
+
+    def get_chat(self, question: str):
+        ai_msg = self.__rag_chain.invoke({"input": question, "chat_history": self.__chat_history})
+        self.__chat_history.extend([HumanMessage(content=question), ai_msg["answer"]])
+        return ai_msg["answer"]
+
+    def clear_history(self):
+        self.__chat_history.clear()
+
+    def select_prompt(self, prompt_index: int = 1):
+        pass
+
+
+if __name__ == '__main__':
+    hr = HistoryRAG("C:\\Users\\16922\\Desktop\\文档1.pdf")
+    print(hr.get_chat("what can Multimodal Agent AI systems do?"))
+    print(hr.get_chat(input()))
+

src/serve/RAG.py

@@ -1,11 +1,28 @@
 import os
 
-from langchain_community.document_loaders import PyPDFLoader
 from langchain_core.callbacks import StreamingStdOutCallbackHandler, CallbackManager
-
-from src.init.property import Property
 from langchain_community.chat_models.zhipuai import ChatZhipuAI
+from langchain_core.output_parsers import StrOutputParser
 from langchain_core.prompts import ChatPromptTemplate
+from langchain import hub
+from langchain_core.runnables import RunnablePassthrough
+
+from src.init.property import Property
+from src.base.embedding import SentenceEmbedding
+
+# 初始化prompt工程
+_prompt = {
+    'prompt_1': ChatPromptTemplate.from_template(
+        """你是问答任务的助手。使用以下检索到的上下文来回答问题。如果你不知道答案，就说你不知道。最多使用三个句子并保持答案简洁。
+
+        {context}
+
+        Question: {question}
+
+        Helpful Answer:"""
+    ),
+    'prompt_2': hub.pull("rlm/rag-prompt")
+}
 
 
 def format_docs(docs):
@@ -16,14 +33,8 @@ class RAG:
     # 初始化ZHIPU API KEY
     os.environ["ZHIPUAI_API_KEY"] = Property.get_property("API_KEY")
 
-    # 初始化prompt工程
-    __prompt = ChatPromptTemplate.from_messages([
-        ("system", "You are a world class technical documentation writer."),
-        ("user", "{input}")
-    ])
-
     # 初始化模型
-    __llm = ChatZhipuAI(
+    _llm = ChatZhipuAI(
         temperature=0.95,
         model="glm-4"
     )
@@ -35,22 +46,38 @@ class RAG:
         callback_manager=CallbackManager([StreamingStdOutCallbackHandler()]),
     )
 
-    # 构建langchain
-    chain = __prompt | __llm
-
     def __init__(self, file_path: str):
+        try:
+            global _prompt
+            self._example_prompt = _prompt["prompt_1"]
+        except NameError:
+            pass
         self.__file_path = file_path
-        loader = PyPDFLoader(self.__file_path)
-        file = loader.load()
+        self.__sentenceEmbedding = SentenceEmbedding(file_path)
+        self._retriever = self.__sentenceEmbedding.get_vectorstore().as_retriever(search_type="similarity",
+                                                                                  search_kwargs={"k": 5})
+
+        try:
+            self.__rag_chain = (
+                    {"context": self._retriever | format_docs, "question": RunnablePassthrough()}
+                    | self._example_prompt
+                    | RAG._llm
+                    | StrOutputParser()
+            )
+        except AttributeError:
+            pass
 
-    def get_answer(self, message) -> str:
-        return RAG.__llm.invoke(message).content
+    def get_chat(self, message: str):
+        for chunk in self.__rag_chain.stream(message):
+            print(chunk, end="", flush=True)
 
-    def get_streaming_chat(self, message) -> str:
-        return RAG.__streaming_chat.invoke(message).content
+    def select_prompt(self, prompt_index: int = 1):
+        global _prompt
+        prompt_name = "prompt_" + str(prompt_index)
+        self._example_prompt = _prompt[prompt_name]
 
 
 if __name__ == '__main__':
     r = RAG("C:\\Users\\16922\\Desktop\\文档1.pdf")
-    print(r.get_streaming_chat("hello"))
-    print(r.get_streaming_chat("what can you do"))
+    r.select_prompt(2)
+    r.get_chat("what can Multimodal Agent AI systems do?")

src/serve/embedding.py

@@ -1,28 +0,0 @@
-from langchain_community.document_loaders import PyPDFLoader
-from langchain_community.vectorstores import Chroma
-from text2vec import SentenceModel
-from src.base.chinese_text_splitter import ChineseTextSplitter
-
-
-class SentenceEmbedding:
-    __model = SentenceModel('shibing624/text2vec-base-chinese')
-    def __init__(self, file_path: str):
-        self.file_path = file_path
-        content = ""
-        loader = PyPDFLoader(file_path)
-        for page in loader.load():
-            content += page.page_content
-        sentences = ChineseTextSplitter(True).split_text(content)
-        embeddings = SentenceEmbedding.__model.encode(sentences)
-        self.vectorstore = Chroma.add_texts(iter(sentences), embeddings)
-
-    def get_vectorstore(self):
-        return self.vectorstore
-
-    def search(self, query:str):
-        embeddings = SentenceEmbedding.__model.encode(query)
-        self.vectorstore
-
-
-if __name__ == '__main__':
-    SentenceEmbedding("C:\\Users\\16922\\Desktop\\文档1.pdf").