Langchain Python

Indexify complements LangChain by providing a robust platform for indexing large volume of multi-modal content such as PDFs, raw text, audio and video. It provides a retriever API to retrieve context for LLMs.

Install the Indexify Langchain retriever package -

pip install indexify-langchain

Add some raw text or documents in Indexify following the Getting Started Guide

Instantiate the Retriever

# Initialize retriever
params = {"name": "minilml6.embedding", "top_k": 9}
retriever = IndexifyRetriever(client=client, params=params)

Here we are initializing the retriever to retrieve from the minilml6.embedding index and asking for top 9 results to be returned.

Setup Chat Prompt Template

from langchain.prompts import ChatPromptTemplate

template = """Answer the question based only on the following context:
  {context}

  Question: {question}"""
prompt = ChatPromptTemplate.from_template(template)

Create some prompt templates.

Ask llm question with retriever context

from langchain_openai import ChatOpenAI
from langchain.schema.runnable import RunnablePassthrough
from langchain.schema.output_parser import StrOutputParser

llm = ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0)

rag_chain = (
    {"context": retriever, "question": RunnablePassthrough()}
    | prompt
    | llm
    | StrOutputParser()
)

Pass in the retriever created above into the chain so that langchain uses that for retreival.

Ask LLM Question

query = "Where is Lucas?"
print(rag_chain.invoke(query))

After that you can just query the chain