Templates (#12294)

Co-authored-by: Harrison Chase <hw.chase.17@gmail.com>
Co-authored-by: Lance Martin <lance@langchain.dev>
Co-authored-by: Jacob Lee <jacoblee93@gmail.com>

templates/rag-elasticsearch/rag_elasticsearch/__init__.py

@@ -0,0 +1,3 @@
+from rag_elasticsearch.chain import chain
+
+__all__ = ["chain"]

templates/rag-elasticsearch/rag_elasticsearch/chain.py

@@ -0,0 +1,56 @@
+from langchain.chat_models import ChatOpenAI
+from langchain.schema.output_parser import StrOutputParser
+from langchain.schema.runnable import RunnablePassthrough, RunnableMap
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain.vectorstores.elasticsearch import ElasticsearchStore
+from langchain.schema import format_document
+from typing import Tuple, List
+from operator import itemgetter
+from .prompts import CONDENSE_QUESTION_PROMPT, LLM_CONTEXT_PROMPT, DOCUMENT_PROMPT
+from .connection import es_connection_details
+
+# Setup connecting to Elasticsearch
+vectorstore = ElasticsearchStore(
+    **es_connection_details,
+    embedding=HuggingFaceEmbeddings(
+        model_name="all-MiniLM-L6-v2", model_kwargs={"device": "cpu"}
+    ),
+    index_name="workplace-search-example",
+)
+retriever = vectorstore.as_retriever()
+
+# Set up LLM to user
+llm = ChatOpenAI(temperature=0)
+
+
+def _combine_documents(
+    docs, document_prompt=DOCUMENT_PROMPT, document_separator="\n\n"
+):
+    doc_strings = [format_document(doc, document_prompt) for doc in docs]
+    return document_separator.join(doc_strings)
+
+
+def _format_chat_history(chat_history: List[Tuple]) -> str:
+    buffer = ""
+    for dialogue_turn in chat_history:
+        human = "Human: " + dialogue_turn[0]
+        ai = "Assistant: " + dialogue_turn[1]
+        buffer += "\n" + "\n".join([human, ai])
+    return buffer
+
+
+_inputs = RunnableMap(
+    standalone_question=RunnablePassthrough.assign(
+        chat_history=lambda x: _format_chat_history(x["chat_history"])
+    )
+    | CONDENSE_QUESTION_PROMPT
+    | llm
+    | StrOutputParser(),
+)
+
+_context = {
+    "context": itemgetter("standalone_question") | retriever | _combine_documents,
+    "question": lambda x: x["standalone_question"],
+}
+
+chain = _inputs | _context | LLM_CONTEXT_PROMPT | llm | StrOutputParser()

templates/rag-elasticsearch/rag_elasticsearch/connection.py

@@ -0,0 +1,15 @@
+import os
+
+ELASTIC_CLOUD_ID = os.getenv("ELASTIC_CLOUD_ID")
+ELASTIC_USERNAME = os.getenv("ELASTIC_USERNAME", "elastic")
+ELASTIC_PASSWORD = os.getenv("ELASTIC_PASSWORD")
+ES_URL = os.getenv("ES_URL", "http://localhost:9200")
+
+if ELASTIC_CLOUD_ID and ELASTIC_USERNAME and ELASTIC_PASSWORD:
+    es_connection_details = {
+        "es_cloud_id": ELASTIC_CLOUD_ID,
+        "es_user": ELASTIC_USERNAME,
+        "es_password": ELASTIC_PASSWORD,
+    }
+else:
+    es_connection_details = {"es_url": ES_URL}

templates/rag-elasticsearch/rag_elasticsearch/prompts.py

@@ -0,0 +1,39 @@
+from langchain.prompts import ChatPromptTemplate, PromptTemplate
+
+# Used to condense a question and chat history into a single question
+condense_question_prompt_template = """Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question, in its original language. If there is no chat history, just rephrase the question to be a standalone question.
+
+Chat History:
+{chat_history}
+Follow Up Input: {question}
+"""
+CONDENSE_QUESTION_PROMPT = PromptTemplate.from_template(
+    condense_question_prompt_template
+)
+
+# RAG Prompt to provide the context and question for LLM to answer
+# We also ask the LLM to cite the source of the passage it is answering from
+llm_context_prompt_template = """
+Use the following passages to answer the user's question.
+Each passage has a SOURCE which is the title of the document. When answering, cite source name of the passages you are answering from below the answer in a unique bullet point list.
+
+If you don't know the answer, just say that you don't know, don't try to make up an answer.
+
+----
+{context}
+----
+Question: {question}
+"""
+
+LLM_CONTEXT_PROMPT = ChatPromptTemplate.from_template(llm_context_prompt_template)
+
+# Used to build a context window from passages retrieved
+document_prompt_template = """
+---
+NAME: {name}
+PASSAGE:
+{page_content}
+---
+"""
+
+DOCUMENT_PROMPT = PromptTemplate.from_template(document_prompt_template)