Application Development

August 21 , 2024 by Deryk DeGuzman & Adeel & Ofer Mendelevitch

Introduction

Vectara is the trusted RAG-as-a-service platform and as such it is primarily used by developers, who utilize the API to ingest data, execute queries, or manage chat sessions, as part of their generative AI applications.

Today we are happy to announce Vectara Portal (beta), a service aimed at non-developers who want to build a great generative AI application but don’t have the budget to hire developers or the time to learn the required programming skills.

If that’s you, please keep on reading and we’ll show you how easy this can be.

What is Vectara Portal?

It all began as the topic of a lunch conversation: “What about non-developers?”

More specifically: how do we let users chat with, get answers from, or search through their data without having to type a single line of code?

We’ve seen services like ChatBase, ChatPDF, PDF.AI, ChatFuel and about a thousand others provide an easy-to-use interface to build simple AI assistants based on uploaded PDF documents or website crawls. Although the quality of their RAG is not always that great, it did serve as a quick and easy way to generate LLM chatbots or question-answering applications.

So with that in mind, we developed Vectara Portal to enable exactly this use case – enabling any Vectara user to generate an AI assistant without having to code. Vectara Portal is powered by our trusted and secure end-to-end RAG platform, and provides higher accuracy of responses with reduced hallucinations. We wanted to provide a full-fledged application to users who don’t want to create one themselves. AI, after all, is here to stay and should be accessible to anyone.

And with that, the idea of Vectara Portal was born.

Setting up Vectara Portal

Before you can start using Vectara Portal, you need to sign up for a Vectara account (if you don’t have one already) and then for a Vectara Portal account.

Image 1: Vectara Portal login screen

Once you login to your account, you need to set up your profile by providing your Vectara credentials that will allow it to generate all of the apps for you automatically, as shown below:

Image 2: Vectara Portal profile settings

• Your email will be pre-filled based on the email address you used to sign up for Vectara Portal
• Copy your Vectara account ID, and your Personal API key into the appropriate boxes in your profile
• Then go to the Authentication page in your Vectara console and copy your OAuth 2.0 client ID and client Secret. If you do not yet have an app client, you’ll need to create one
• Click “Save” to save all of your settings and you’re ready to go

If you have a Vectara Scale account, click on the button marked “are you a scale customer”, to ensure your portals use our most powerful LLMs and our state-of-the-art reranker, making the responses from portal even better.

Great, now you are ready to use Vectara Portal.

Creating Portals

With your Vectara Portal account setup, let’s see how you can create a portal. A portal is a single application, connected to Vectara’s RAG-as-a-service, that can be customized by you with only a single configuration parameter.

Start by clicking on “create a portal”, and you will see the portal creation screen:

Image 3: creating a portal

Simply enter your portal name, a short description and whether you want the user interface to be “search” (semantic search), “summary” (RAG), or “chat” (AI assistant).

In this example, we’ll create a portal based on a few PDF documents from the CFPB website, covering various topics in financial regulation. We choose the “Chat” portal type for a chat interface, but that can of course be changed later.

Click “Create It”, and your portal is ready to go.

You will now be transferred to your “portals” view and notice the “cfpb” portal that has been added. 

Once you click on it (opening in a new window), click on the “settings” icon on the top right, and then near “documents,” click on the plus sign (+) to add documents.

In this case, I’m going to add 7 CFPB reports (PDF files):

Image 4: adding documents to your portal

Once the files are finished uploading, click “Save”, and you can start using the Portal and chat with the data:

Image 5: example chat in your portal

We’ve made this specific portal with CFPB data public, so you can just try it yourself.

As you chat with your data, clicking reference links will open the references drawer, giving full visibility into the source of the provided response.

Sharing Portals

The portal we just created is shared. In fact you can easily share any portal you create with others in your organization: simply click on the “Share” button next to the settings and the portal URL is copied to your clipboard.

By default, portals are public so a link to the URL is all you need to share them with your friends or colleagues. If you want to keep a portal “private,” just turn on the “restricted” button in the portal settings. In private mode, the portal can only be used by you, or other users you’ve added to “Your users” in your profile.

Vectara Portal as an Open Source Repository

The Portal app is hosted by Vectara so you can use it immediately at https://portal.demo.vectara.com.

At the same time, we made the source code available under the Apache 2.0 license.

If you want to install Vectara Portal within your organization, please check the detailed instructions to set it up.

We welcome your feedback, comments, and contributions to the project that can make it better.

Conclusion

Vectara Portal is a service that provides a no-code path for building a GenAI application using Vectara: create a portal, upload a document, and you have a working search, question-answering, or an AI assistant in minutes.

Now everyone can chat with one or more uploaded documents, using the latest RAG technology from Vectara, and showcase the results to their friends and colleagues. We have made the Vectara Portal source code available as an open-source repository, under the Apache 2.0 license, and we would love to hear your feedback and feature suggestions on our discussion forums or Discord server.