langchain/templates/rag-momento-vector-index/rag_momento_vector_index/chain.py
2024-01-03 13:28:05 -08:00

63 lines
1.8 KiB
Python

import os
from langchain_community.chat_models import ChatOpenAI
from langchain_community.embeddings import OpenAIEmbeddings
from langchain_community.vectorstores import MomentoVectorIndex
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.pydantic_v1 import BaseModel
from langchain_core.runnables import RunnablePassthrough
from momento import (
CredentialProvider,
PreviewVectorIndexClient,
VectorIndexConfigurations,
)
API_KEY_ENV_VAR_NAME = "MOMENTO_API_KEY"
if os.environ.get(API_KEY_ENV_VAR_NAME, None) is None:
raise Exception(f"Missing `{API_KEY_ENV_VAR_NAME}` environment variable.")
MOMENTO_INDEX_NAME = os.environ.get("MOMENTO_INDEX_NAME", "langchain-test")
### Sample Ingest Code - this populates the vector index with data
### Run this on the first time to seed with data
# from rag_momento_vector_index import ingest
# ingest.load(API_KEY_ENV_VAR_NAME, MOMENTO_INDEX_NAME)
vectorstore = MomentoVectorIndex(
embedding=OpenAIEmbeddings(),
client=PreviewVectorIndexClient(
configuration=VectorIndexConfigurations.Default.latest(),
credential_provider=CredentialProvider.from_environment_variable(
API_KEY_ENV_VAR_NAME
),
),
index_name=MOMENTO_INDEX_NAME,
)
retriever = vectorstore.as_retriever()
# RAG prompt
template = """Answer the question based only on the following context:
{context}
Question: {question}
"""
prompt = ChatPromptTemplate.from_template(template)
# RAG
model = ChatOpenAI()
chain = (
{"context": retriever, "question": RunnablePassthrough()}
| prompt
| model
| StrOutputParser()
)
# Add typing for input
class Question(BaseModel):
__root__: str
chain = chain.with_types(input_type=Question)