o Jjg=R@sdZddlmZddlZddlZddlmZddlmZddl m Z m Z m Z m Z mZmZmZmZddlmZddlmZmZmZdd lmZdd lmZdd lmZdd lmZdd l m!Z!ddl"m#Z#ddl$m%Z%ddl&m'Z'm(Z(m)Z)m*Z*ddl+m,Z,ddl-m.Z.ddl/m0Z0ddl1m2Z2ddl3m4Z4ddl5m6Z6eee7e7fefZ8dddZ9d,ddZ:Gd d!d!e'Z;Gd"d#d#e,ZdS)-*Chain for chatting with a vector database.) annotationsN)abstractmethod)Path)AnyCallableDictListOptionalTupleTypeUnion) deprecated)AsyncCallbackManagerForChainRunCallbackManagerForChainRun Callbacks)Document)BaseLanguageModel) BaseMessage)BasePromptTemplate) BaseRetriever)RunnableConfig) VectorStore) BaseModel ConfigDictFieldmodel_validator)Chain)BaseCombineDocumentsChain)StuffDocumentsChain)CONDENSE_QUESTION_PROMPT)LLMChain) load_qa_chainHuman: Assistant: )humanai chat_historyList[CHAT_TURN_TYPE]returnstrcCsd}|D]M}t|tr(t|jdkr't|j|jd}|d||j7}qt|trEd|d}d|d}|dd||g7}qt dt|d |d |S) Nrz:  r#r$z!Unsupported chat history format: z. Full chat history:  ) isinstancerlencontent _ROLE_MAPgettypetuplejoin ValueError)r'buffer dialogue_turn role_prefixr%r&r;f/var/www/html/zoom/venv/lib/python3.10/site-packages/langchain/chains/conversational_retrieval/base.py_get_chat_history)s&     r=c@s.eZdZUdZded< eedZded<dS) InputTypez,Input type for ConversationalRetrievalChain.r*questiondefault_factoryr(r'N)__name__ __module__ __qualname____doc____annotations__rlistr'r;r;r;r<r>>s r>cseZdZUdZded< ded< dZded< d Zd ed < d Zd ed < d Zd ed< dZ ded< dZ ded< e d d ddZ e d7ddZ d8d9ddZe d7dd Zed:d'd(Z d8d;d*d+Zedfd5d6 ZZS)? BaseConversationalRetrievalChainz!Chain for chatting with an index.rcombine_docs_chainr!question_generatoranswerr* output_keyTboolrephrase_questionFreturn_source_documentsreturn_generated_questionNz/Optional[Callable[[List[CHAT_TURN_TYPE]], str]]get_chat_historyz Optional[str]response_if_no_docs_foundforbid)populate_by_namearbitrary_types_allowedextrar) List[str]cCsddgS)z Input keys.r?r'r;selfr;r;r< input_keysisz+BaseConversationalRetrievalChain.input_keysconfigOptional[RunnableConfig]Type[BaseModel]cCstSN)r>)rYr[r;r;r<get_input_schemansz1BaseConversationalRetrievalChain.get_input_schemacCs,|jg}|jr |dg}|jr|dg}|S)z8Return the output keys. :meta private: source_documentsgenerated_question)rLrOrP)rY _output_keysr;r;r< output_keysss   z,BaseConversationalRetrievalChain.output_keysr?inputsDict[str, Any] run_managerrList[Document]cCdS Get docs.Nr;rYr?rdrfr;r;r< _get_docssz*BaseConversationalRetrievalChain._get_docs$Optional[CallbackManagerForChainRun]cCs|pt}|d}|jpt}||d}|r%|}|jj|||d}n|}dt|j j v} | r;|j |||d} n| ||} i} |j durUt | dkrU|j | |j <n"|} |jr`|| d<|| d<|jjd | |d| } | | |j <|jr~| | d<|jr|| d <| S Nr?r')r?r' callbacksrf)rfr)input_documentsror`rar;)rget_noop_managerrQr= get_childrJruninspect signaturerl parametersrRr0rLcopyrNrIrOrPrYrdrf _run_managerr?rQchat_history_strro new_questionaccepts_run_managerdocsoutput new_inputsrKr;r;r<_callsB     z&BaseConversationalRetrievalChain._callrcsdSrir;rkr;r;r< _aget_docssz+BaseConversationalRetrievalChain._aget_docs)Optional[AsyncCallbackManagerForChainRun]cs(|pt}|d}|jpt}||d}|r)|}|jj|||dIdH}n|}dt|j j v} | rB|j |||dIdH} n | ||IdH} i} |j dur_t | dkr_|j | |j <n%|} |jrj|| d<|| d<|jjd | |d| IdH} | | |j <|jr| | d<|jr|| d <| Srn)rrqrQr=rrrJarunrtrurrvrRr0rLrwrNrIrOrPrxr;r;r<_acallsD    z'BaseConversationalRetrievalChain._acall file_pathUnion[Path, str]Nonecs|jrtdt|dS)Nz7Chain not saveable when `get_chat_history` is not None.)rQr7supersave)rYr __class__r;r<rsz%BaseConversationalRetrievalChain.save)r)rWr^)r[r\r)r]r?r*rdrerfrr)rg)rdrerfrmr)rer?r*rdrerfrr)rg)rdrerfrr)re)rrr)r)rBrCrDrErFrLrNrOrPrQrRr model_configpropertyrZr_rcrrlrrrr __classcell__r;r;rr<rHGsL            +  +rHz0.1.17z^create_history_aware_retriever together with create_retrieval_chain (see example in docstring)z1.0)since alternativeremovalc@sdeZdZUdZded< dZded< d+d d Zd,ddZd-ddZe e dddddfd.d)d*Z dS)/ConversationalRetrievalChainaChain for having a conversation based on retrieved documents. This class is deprecated. See below for an example implementation using `create_retrieval_chain`. Additional walkthroughs can be found at https://python.langchain.com/docs/use_cases/question_answering/chat_history .. code-block:: python from langchain.chains import ( create_history_aware_retriever, create_retrieval_chain, ) from langchain.chains.combine_documents import create_stuff_documents_chain from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder from langchain_openai import ChatOpenAI retriever = ... # Your retriever llm = ChatOpenAI() # Contextualize question 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}"), ] ) history_aware_retriever = create_history_aware_retriever( llm, retriever, contextualize_q_prompt ) # Answer question 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}"), ] ) # Below we use create_stuff_documents_chain to feed all retrieved context # into the LLM. Note that we can also use StuffDocumentsChain and other # instances of BaseCombineDocumentsChain. question_answer_chain = create_stuff_documents_chain(llm, qa_prompt) rag_chain = create_retrieval_chain( history_aware_retriever, question_answer_chain ) # Usage: chat_history = [] # Collect chat history here (a sequence of messages) rag_chain.invoke({"input": query, "chat_history": chat_history}) This chain takes in chat history (a list of messages) and new questions, and then returns an answer to that question. The algorithm for this chain consists of three parts: 1. Use the chat history and the new question to create a "standalone question". This is done so that this question can be passed into the retrieval step to fetch relevant documents. If only the new question was passed in, then relevant context may be lacking. If the whole conversation was passed into retrieval, there may be unnecessary information there that would distract from retrieval. 2. This new question is passed to the retriever and relevant documents are returned. 3. The retrieved documents are passed to an LLM along with either the new question (default behavior) or the original question and chat history to generate a final response. Example: .. code-block:: python from langchain.chains import ( StuffDocumentsChain, LLMChain, ConversationalRetrievalChain ) from langchain_core.prompts import PromptTemplate from langchain_community.llms import OpenAI combine_docs_chain = StuffDocumentsChain(...) vectorstore = ... retriever = vectorstore.as_retriever() # This controls how the standalone question is generated. # Should take `chat_history` and `question` as input variables. template = ( "Combine the chat history and follow up question into " "a standalone question. Chat History: {chat_history}" "Follow up question: {question}" ) prompt = PromptTemplate.from_template(template) llm = OpenAI() question_generator_chain = LLMChain(llm=llm, prompt=prompt) chain = ConversationalRetrievalChain( combine_docs_chain=combine_docs_chain, retriever=retriever, question_generator=question_generator_chain, ) r retrieverNz Optional[int]max_tokens_limitr}rgr)cspt|}jr2tjtr2fdd|D}t|d|}|jkr2|d8}|||8}|jks#|d|S)Ncsg|] }jj|jqSr;)rI llm_chain_get_num_tokens page_content).0docrXr;r< yszKConversationalRetrievalChain._reduce_tokens_below_limit..r-)r0rr/rIrsum)rYr}num_docstokens token_countr;rXr<_reduce_tokens_below_limitss    z7ConversationalRetrievalChain._reduce_tokens_below_limitr?r*rdrerfrcCs"|jj|d|id}||S)rjror[)rinvokerrrrYr?rdrfr}r;r;r<rls  z&ConversationalRetrievalChain._get_docsrcs*|jj|d|idIdH}||S)rjrorN)rainvokerrrrr;r;r<rs   z'ConversationalRetrievalChain._aget_docsstuffFllmrcondense_question_promptr chain_typeverboserMcondense_question_llmOptional[BaseLanguageModel]combine_docs_chain_kwargsOptional[Dict]rorkwargsrrHc KsP|pi}t|f|||d|} |p|} t| |||d} |d|| | |d| S)a Convenience method to load chain from LLM and retriever. This provides some logic to create the `question_generator` chain as well as the combine_docs_chain. Args: llm: The default language model to use at every part of this chain (eg in both the question generation and the answering) retriever: The retriever to use to fetch relevant documents from. condense_question_prompt: The prompt to use to condense the chat history and new question into a standalone question. chain_type: The chain type to use to create the combine_docs_chain, will be sent to `load_qa_chain`. verbose: Verbosity flag for logging to stdout. condense_question_llm: The language model to use for condensing the chat history and new question into a standalone question. If none is provided, will default to `llm`. combine_docs_chain_kwargs: Parameters to pass as kwargs to `load_qa_chain` when constructing the combine_docs_chain. callbacks: Callbacks to pass to all subchains. kwargs: Additional parameters to pass when initializing ConversationalRetrievalChain )rrro)rpromptrro)rrIrJroNr;r"r!) clsrrrrrrrror doc_chain_llmcondense_question_chainr;r;r<from_llms2$z%ConversationalRetrievalChain.from_llm)r}rgr)rgrr)rrrrrrrr*rrMrrrrrorrrr)rH) rBrCrDrErFrrrlr classmethodr rr;r;r;r<rs"  t     rc@seZdZUdZeddZded<dZded<eedZ d ed <e d0d dZ e dde d1ddZd2ddZd3dd Ze ed!d"d"fd4d.d/Zd"S)5ChatVectorDBChainr vectorstore)aliasrinttop_k_docs_for_contextr@dict search_kwargsr)r*cCrh)Nzchat-vector-dbr;rXr;r;r< _chain_typeszChatVectorDBChain._chain_typebefore)modevaluesrrcCstd|S)Nzj`ChatVectorDBChain` is deprecated - please use `from langchain.chains import ConversationalRetrievalChain`)warningswarn)rrr;r;r<raise_deprecationsz#ChatVectorDBChain.raise_deprecationr?rdrerfrrgcCs4|di}i|j|}|jj|fd|ji|S)rjvectordbkwargsk)r3rrsimilarity_searchr)rYr?rdrfr full_kwargsr;r;r<rls zChatVectorDBChain._get_docsrcs td)rjz(ChatVectorDBChain does not support async)NotImplementedErrorrkr;r;r<rszChatVectorDBChain._aget_docsrNrrrrrrrrorrrHc KsD|pi}t|f||d|}t|||d} |d||| |d|S)zLoad chain from LLM.)rro)rrro)rrIrJroNr;r) rrrrrrrorrrr;r;r<rs( zChatVectorDBChain.from_llm)r)r*)rrr)rrr)rrrrrrrr*rrrorrrr)rH)rBrCrDrErrrFrrrrrrrrrlrr rr;r;r;r<rs$      r)r'r(r)r*)?rE __future__rrtrabcrpathlibrtypingrrrr r r r r langchain_core._apirlangchain_core.callbacksrrrlangchain_core.documentsrlangchain_core.language_modelsrlangchain_core.messagesrlangchain_core.promptsrlangchain_core.retrieversrlangchain_core.runnablesrlangchain_core.vectorstoresrpydanticrrrrlangchain.chains.baser'langchain.chains.combine_documents.baser(langchain.chains.combine_documents.stuffr1langchain.chains.conversational_retrieval.promptsr langchain.chains.llmr!#langchain.chains.question_answeringr"r*CHAT_TURN_TYPEr2r=r>rHrrr;r;r;r<sJ   (                 *d