README.md

Milvus Documentation Search

This repo includes en example application that makes use of a Retrieval Augmented Generative architecture built with Haystack to do search on the Milvus documentation.

Screen.Recording.2023-08-28.at.13.51.59.mov

Install dependencies

pip install -r requirements.txt

The Indexing Pipeline

An indexing pipeline is used to write documents to a database. In this example, we use the MilvusDocumentStore as our database for the RAG pipeline. So, we need to write the Milvus documentation into our Milvus database. For demonstration purposes, we use the Crawler component to crawl everything under https://milvus.io/docs/

Once you have Milvus running locally on localhost:19530, you can use the indexing pipeline as follows:

python scripts/index_files.py

The RAG Pipeline

The RAG pipeline that we use is the following:

from haystack import Pipeline
from haystack.nodes import EmbeddingRetriever, PromptNode, PromptTemplate, AnswerParser
from milvus_haystack import MilvusDocumentStore

document_store = MilvusDocumentStore()

retriever = EmbeddingRetriever(document_store=document_store, embedding_model="sentence-transformers/multi-qa-mpnet-base-dot-v1")
template = PromptTemplate(prompt="deepset/question-answering", output_parser=AnswerParser())
prompt_node = PromptNode(model_name_or_path="gpt-4", default_prompt_template=template, api_key=OPENAI_API_KEY, max_length=500)

query_pipeline = Pipeline()
query_pipeline.add_node(component=retriever, name="Retriever", inputs=["Query"])
query_pipeline.add_node(component=prompt_node, name="PromptNode", inputs=["Retriever"])

To run it as a Streamlit App

streamlit run app.py

To run it as a standalone script

python scripts/documentation_qa.py