mirror of https://github.com/hwchase17/langchain
zep/rag conversation zep template (#12762)
LangServe template for a RAG Conversation App using Zep. @baskaryan, @eyurtsev --------- Co-authored-by: Erick Friis <erick@langchain.dev>pull/12865/head
parent
ea1ab391d4
commit
f41f4c5e37
@ -0,0 +1,21 @@
|
||||
MIT License
|
||||
|
||||
Copyright (c) 2023 LangChain, Inc.
|
||||
|
||||
Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
of this software and associated documentation files (the "Software"), to deal
|
||||
in the Software without restriction, including without limitation the rights
|
||||
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
copies of the Software, and to permit persons to whom the Software is
|
||||
furnished to do so, subject to the following conditions:
|
||||
|
||||
The above copyright notice and this permission notice shall be included in all
|
||||
copies or substantial portions of the Software.
|
||||
|
||||
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
SOFTWARE.
|
@ -0,0 +1,37 @@
|
||||
# Ingest Documents into a Zep Collection
|
||||
import os
|
||||
|
||||
from langchain.document_loaders import WebBaseLoader
|
||||
from langchain.embeddings import FakeEmbeddings
|
||||
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
||||
from langchain.vectorstores.zep import CollectionConfig, ZepVectorStore
|
||||
|
||||
ZEP_API_URL = os.environ.get("ZEP_API_URL", "http://localhost:8000")
|
||||
ZEP_API_KEY = os.environ.get("ZEP_API_KEY", None)
|
||||
ZEP_COLLECTION_NAME = os.environ.get("ZEP_COLLECTION", "langchaintest")
|
||||
|
||||
collection_config = CollectionConfig(
|
||||
name=ZEP_COLLECTION_NAME,
|
||||
description="Zep collection for LangChain",
|
||||
metadata={},
|
||||
embedding_dimensions=1536,
|
||||
is_auto_embedded=True,
|
||||
)
|
||||
|
||||
# Load
|
||||
loader = WebBaseLoader("https://lilianweng.github.io/posts/2023-06-23-agent/")
|
||||
data = loader.load()
|
||||
|
||||
# Split
|
||||
text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)
|
||||
all_splits = text_splitter.split_documents(data)
|
||||
|
||||
# Add to vectorDB
|
||||
vectorstore = ZepVectorStore.from_documents(
|
||||
documents=all_splits,
|
||||
collection_name=ZEP_COLLECTION_NAME,
|
||||
config=collection_config,
|
||||
api_url=ZEP_API_URL,
|
||||
api_key=ZEP_API_KEY,
|
||||
embedding=FakeEmbeddings(size=1),
|
||||
)
|
File diff suppressed because it is too large
Load Diff
@ -0,0 +1,28 @@
|
||||
[tool.poetry]
|
||||
name = "rag_conversation_zep"
|
||||
version = "0.0.1"
|
||||
description = "A RAG application built with Zep. Zep provides a VectorStore implementation to the chain."
|
||||
authors = ["Daniel Chalef <daniel@getzep.com>"]
|
||||
readme = "README.md"
|
||||
|
||||
[tool.poetry.dependencies]
|
||||
python = ">=3.8.1,<4.0"
|
||||
langchain = ">=0.0.313, <0.1"
|
||||
openai = "^0.28.1"
|
||||
zep-python = "^1.4.0"
|
||||
tiktoken = "^0.5.1"
|
||||
beautifulsoup4 = "^4.12.2"
|
||||
bs4 = "^0.0.1"
|
||||
|
||||
[tool.poetry.group.dev.dependencies]
|
||||
langchain-cli = ">=0.0.15"
|
||||
fastapi = "^0.104.0"
|
||||
sse-starlette = "^1.6.5"
|
||||
|
||||
[tool.langserve]
|
||||
export_module = "rag_conversation_zep"
|
||||
export_attr = "chain"
|
||||
|
||||
[build-system]
|
||||
requires = ["poetry-core"]
|
||||
build-backend = "poetry.core.masonry.api"
|
@ -0,0 +1,3 @@
|
||||
from rag_conversation_zep.chain import chain
|
||||
|
||||
__all__ = ["chain"]
|
@ -0,0 +1,144 @@
|
||||
import os
|
||||
from operator import itemgetter
|
||||
from typing import List, Tuple
|
||||
|
||||
from langchain.chat_models import ChatOpenAI
|
||||
from langchain.prompts import ChatPromptTemplate, MessagesPlaceholder
|
||||
from langchain.prompts.prompt import PromptTemplate
|
||||
from langchain.pydantic_v1 import BaseModel, Field
|
||||
from langchain.schema import AIMessage, HumanMessage, format_document
|
||||
from langchain.schema.document import Document
|
||||
from langchain.schema.messages import BaseMessage
|
||||
from langchain.schema.output_parser import StrOutputParser
|
||||
from langchain.schema.runnable import (
|
||||
ConfigurableField,
|
||||
RunnableBranch,
|
||||
RunnableLambda,
|
||||
RunnableMap,
|
||||
RunnablePassthrough,
|
||||
)
|
||||
from langchain.schema.runnable.utils import ConfigurableFieldSingleOption
|
||||
from langchain.vectorstores.zep import CollectionConfig, ZepVectorStore
|
||||
|
||||
ZEP_API_URL = os.environ.get("ZEP_API_URL", "http://localhost:8000")
|
||||
ZEP_API_KEY = os.environ.get("ZEP_API_KEY", None)
|
||||
ZEP_COLLECTION_NAME = os.environ.get("ZEP_COLLECTION", "langchaintest")
|
||||
|
||||
collection_config = CollectionConfig(
|
||||
name=ZEP_COLLECTION_NAME,
|
||||
description="Zep collection for LangChain",
|
||||
metadata={},
|
||||
embedding_dimensions=1536,
|
||||
is_auto_embedded=True,
|
||||
)
|
||||
|
||||
vectorstore = ZepVectorStore(
|
||||
collection_name=ZEP_COLLECTION_NAME,
|
||||
config=collection_config,
|
||||
api_url=ZEP_API_URL,
|
||||
api_key=ZEP_API_KEY,
|
||||
embedding=None,
|
||||
)
|
||||
|
||||
# Zep offers native, hardware-accelerated MMR. Enabling this will improve
|
||||
# the diversity of results, but may also reduce relevance. You can tune
|
||||
# the lambda parameter to control the tradeoff between relevance and diversity.
|
||||
# Enabling is a good default.
|
||||
retriever = vectorstore.as_retriever().configurable_fields(
|
||||
search_type=ConfigurableFieldSingleOption(
|
||||
id="search_type",
|
||||
options={"Similarity": "similarity", "Similarity with MMR Reranking": "mmr"},
|
||||
default="mmr",
|
||||
name="Search Type",
|
||||
description="Type of search to perform: 'similarity' or 'mmr'",
|
||||
),
|
||||
search_kwargs=ConfigurableField(
|
||||
id="search_kwargs",
|
||||
name="Search kwargs",
|
||||
description=(
|
||||
"Specify 'k' for number of results to return and 'lambda_mult' for tuning"
|
||||
" MMR relevance vs diversity."
|
||||
),
|
||||
),
|
||||
)
|
||||
|
||||
# Condense a chat history and follow-up question into a standalone question
|
||||
_template = """Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question, in its original language.
|
||||
Chat History:
|
||||
{chat_history}
|
||||
Follow Up Input: {question}
|
||||
Standalone question:""" # noqa: E501
|
||||
CONDENSE_QUESTION_PROMPT = PromptTemplate.from_template(_template)
|
||||
|
||||
# RAG answer synthesis prompt
|
||||
template = """Answer the question based only on the following context:
|
||||
<context>
|
||||
{context}
|
||||
</context>"""
|
||||
ANSWER_PROMPT = ChatPromptTemplate.from_messages(
|
||||
[
|
||||
("system", template),
|
||||
MessagesPlaceholder(variable_name="chat_history"),
|
||||
("user", "{question}"),
|
||||
]
|
||||
)
|
||||
|
||||
# Conversational Retrieval Chain
|
||||
DEFAULT_DOCUMENT_PROMPT = PromptTemplate.from_template(template="{page_content}")
|
||||
|
||||
|
||||
def _combine_documents(
|
||||
docs: List[Document],
|
||||
document_prompt: PromptTemplate = DEFAULT_DOCUMENT_PROMPT,
|
||||
document_separator: str = "\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, str]]) -> List[BaseMessage]:
|
||||
buffer: List[BaseMessage] = []
|
||||
for human, ai in chat_history:
|
||||
buffer.append(HumanMessage(content=human))
|
||||
buffer.append(AIMessage(content=ai))
|
||||
return buffer
|
||||
|
||||
|
||||
_condense_chain = (
|
||||
RunnablePassthrough.assign(
|
||||
chat_history=lambda x: _format_chat_history(x["chat_history"])
|
||||
)
|
||||
| CONDENSE_QUESTION_PROMPT
|
||||
| ChatOpenAI(temperature=0)
|
||||
| StrOutputParser()
|
||||
)
|
||||
|
||||
_search_query = RunnableBranch(
|
||||
# If input includes chat_history, we condense it with the follow-up question
|
||||
(
|
||||
RunnableLambda(lambda x: bool(x.get("chat_history"))).with_config(
|
||||
run_name="HasChatHistoryCheck"
|
||||
),
|
||||
# Condense follow-up question and chat into a standalone_question
|
||||
_condense_chain,
|
||||
),
|
||||
# Else, we have no chat history, so just pass through the question
|
||||
RunnableLambda(itemgetter("question")),
|
||||
)
|
||||
|
||||
|
||||
# User input
|
||||
class ChatHistory(BaseModel):
|
||||
chat_history: List[Tuple[str, str]] = Field(..., extra={"widget": {"type": "chat"}})
|
||||
question: str
|
||||
|
||||
|
||||
_inputs = RunnableMap(
|
||||
{
|
||||
"question": lambda x: x["question"],
|
||||
"chat_history": lambda x: _format_chat_history(x["chat_history"]),
|
||||
"context": _search_query | retriever | _combine_documents,
|
||||
}
|
||||
).with_types(input_type=ChatHistory)
|
||||
|
||||
chain = _inputs | ANSWER_PROMPT | ChatOpenAI() | StrOutputParser()
|
Loading…
Reference in New Issue