Pathway Retriever

Pathway is an open data processing framework. It allows you to easily develop data transformation pipelines and Machine Learning applications that work with live data sources and changing data.

This notebook demonstrates how to use a live data indexing pipeline with LlamaIndex. You can query the results of this pipeline from your LLM application using the provided PathwayRetriever. However, under the hood, Pathway updates the index on each data change giving you always up-to-date answers.

In this notebook, we will use a public demo document processing pipeline that:

Monitors several cloud data sources for data changes.
Builds a vector index for the data.

To have your own document processing pipeline check the hosted offering or build your own by following this notebook.

We will connect to the index using llama_index.retrievers.pathway.PathwayRetriever retriever, which implements the retrieve interface.

The basic pipeline described in this document allows to effortlessly build a simple index of files stored in a cloud location. However, Pathway provides everything needed to build realtime data pipelines and apps, including SQL-like able operations such as groupby-reductions and joins between disparate data sources, time-based grouping and windowing of data, and a wide array of connectors.

For more details about Pathway data ingestion pipeline and vector store, visit vector store pipeline.

Prerequisites

To use PathwayRetrievier you must install llama-index-retrievers-pathway package.

python

!pip install llama-index-retrievers-pathway

Create Retriever for llama-index

To instantiate and configure PathwayRetriever you need to provide either the url or the host and port of your document indexing pipeline. In the code below we use a publicly available demo pipeline, which REST API you can access at https://demo-document-indexing.pathway.stream. This demo ingests documents from Google Drive and Sharepoint and maintains an index for retrieving documents.

python

from llama_index.retrievers.pathway import PathwayRetriever

retriever = PathwayRetriever(
    url="https://demo-document-indexing.pathway.stream"
)
retriever.retrieve(str_or_query_bundle="what is pathway")

Your turn! Get your pipeline or upload new documents to the demo pipeline and retry the query!

Use in Query Engine

python

from llama_index.core.query_engine import RetrieverQueryEngine

query_engine = RetrieverQueryEngine.from_args(
    retriever,
)

python

response = query_engine.query("Tell me about Pathway")
print(str(response))

Building your own data processing pipeline

Prerequisites

Install pathway package. Then download sample data.

python

%pip install pathway
%pip install llama-index-embeddings-openai

python

!mkdir -p 'data/'
!wget 'https://gist.githubusercontent.com/janchorowski/dd22a293f3d99d1b726eedc7d46d2fc0/raw/pathway_readme.md' -O 'data/pathway_readme.md'

Define data sources tracked by Pathway

Pathway can listen to many sources simultaneously, such as local files, S3 folders, cloud storage and any data stream for data changes.

See pathway-io for more information.

python

import pathway as pw

data_sources = []
data_sources.append(
    pw.io.fs.read(
        "./data",
        format="binary",
        mode="streaming",
        with_metadata=True,
    )  # This creates a `pathway` connector that tracks
    # all the files in the ./data directory
)

# This creates a connector that tracks files in Google drive.
# please follow the instructions at https://pathway.com/developers/tutorials/connectors/gdrive-connector/ to get credentials
# data_sources.append(
#     pw.io.gdrive.read(object_id="17H4YpBOAKQzEJ93xmC2z170l0bP2npMy", service_user_credentials_file="credentials.json", with_metadata=True))

Create the document indexing pipeline

Let us create the document indexing pipeline. The transformations should be a list of TransformComponents ending with an Embedding transformation.

In this example, let's first split the text first using TokenTextSplitter, then embed with OpenAIEmbedding.

python

from pathway.xpacks.llm.vector_store import VectorStoreServer
from llama_index.embeddings.openai import OpenAIEmbedding
from llama_index.core.node_parser import TokenTextSplitter

embed_model = OpenAIEmbedding(embed_batch_size=10)

transformations_example = [
    TokenTextSplitter(
        chunk_size=150,
        chunk_overlap=10,
        separator=" ",
    ),
    embed_model,
]

processing_pipeline = VectorStoreServer.from_llamaindex_components(
    *data_sources,
    transformations=transformations_example,
)

# Define the Host and port that Pathway will be on
PATHWAY_HOST = "127.0.0.1"
PATHWAY_PORT = 8754

# `threaded` runs pathway in detached mode, we have to set it to False when running from terminal or container
# for more information on `with_cache` check out https://pathway.com/developers/api-docs/persistence-api
processing_pipeline.run_server(
    host=PATHWAY_HOST, port=PATHWAY_PORT, with_cache=False, threaded=True
)

Connect the retriever to the custom pipeline

python

from llama_index.retrievers.pathway import PathwayRetriever

retriever = PathwayRetriever(host=PATHWAY_HOST, port=PATHWAY_PORT)
retriever.retrieve(str_or_query_bundle="what is pathway")