Enterprise Search

From search portals to search engines, enterprise search has come on to the scene as a universal search tool for an organization’s internal information. Enterprise search has lessened the tribal knowledge and silos that come with an organization with loads of internal documentation. At the heart of enterprise search is:

• The ability to ingest and stay up-to-date on data

• The ability to make that data useful for users

Here is a lightweight demo video we put together for the Enterprise Search use case.

Data Ingestion and Data-Centric Actions

Integrations are paramount for data ingestion and actions as your customers’ data lives in dozens, if not hundreds of different third-party applications (i.e. CRMs like Salesforce or Hubspot, file storage apps like Google Drive or Sharepoint, ticketing platforms like Zendesk or ServiceNow, etc). Making that data useful often comes in two forms:

1. Optimally ranking document search results from a user's query

2. Giving users the ability to use AI and LLMs to extract important details from all of their data.

Looking forward, this could also mean performing write and update operations - not just reading data but taking actions in the 3rd-party apps with that context.

In this use case demo, we will be demoing an enterprise search application with integrations with Google Drive and Notion to show how data ingestion can be done at scale with Paragon as well as how you can perform agent triggered actions in those third-party applications, like transferring a Google Drive file to a Notion page via LLM prompting.

Data Ingestion

An enterprise search must be able to ingest and process data for retrieval. While it may seem simple to build out simple API commands that retrieve documents and data from third-party applications, ingestion jobs can be compute heavy, high-throughput jobs. Webhooks and polling mechanisms also must be considered to not just initially ingest data, but stay up-to-date with file changes in real time.

Paragon enables you to offer your users a fully white-labeled and embedded authentication experience for each 3rd party app, and provides the infrastructure for building ingestion jobs and webhook triggered workflows for your third-party integration use cases. This makes it easier for your engineers to add more and more third-party integrations and offload code maintenance of those integrations.

In this use case demo, we built a simple enterprise search application that can ingest and index data from uploaded PDF files, Google Drive files, and Notion pages.

With Paragon, the full array of file types from a file storage platform like Google Drive can be handled in the same workflow. Not only are these workflows developer friendly - allowing your engineering team to build out complex data ingestion pipelines quickly - Paragon’s workflow infrastructure scales and is able to offload large ingestion jobs away from your application’s infrastructure.

As mentioned, we wanted our application to not only handle initial file ingestion, but also capture real-time file changes. This is made simple with Paragon’s webhook triggers. Our Notion workflow is initiated when changes are made in our Notion page.

Data-Centric Actions

Your enterprise search likely has some level of third-party data ingestion. However, what will be a differentiator for your product is how you present the data to users and how you enable users to action off of that data. That’s where agentic actions and interacting with third-party APIs can come into play.

In our use case demo, we gave users the ability to interact with a chatbot agent for the selected document and gave our AI agent the ability to create a copy of the contents from the selected document in Notion.

This is enabled using another Paragon workflow that our agent triggers when a user is prompting for a Notion action. Again, we have developer friendly steps (in this example we needed to chunk our input text, as Notion’s API only allows blocks to be limited to 2000 characters), allowing your developers to interact with third-party APIs with a high level of customizability for your use case.

Wrapping Up

The value of an Enterprise Search platform comes down to its ability to index and make use of users' data: from ingesting data, to staying up-to-date with file changes, to actioning on that data. Integrations are a key piece in all three of those aspects, as data will flow from your users’ third-party applications to your search application and from your application back to your users’ third-parties. Paragon specializes in solving these integration challenges, helping customer like Frame and Pryon build out their enterprise search and AI capabilities.