Vectara Portal helps non-developers build AI apps that chat with data. Here’s how it works:

Vectara just made it a lot easier to develop generative AI. The Palo Alto, California-based company, a pioneer in the field of search augmentation generative (RAG), has announced the Vectara Portal, an open-source environment that enables anyone to build AI applications that interact with data.

While there are many commercial services that allow users to get instant answers from documentation, what sets Vectara Portal apart is its accessibility and ease of use: by following a few basic steps, anyone, regardless of technical skill or knowledge, can freely use search, summarization, and chat apps based on their datasets – without writing a single line of code.

The service has the potential to enable non-developers to power a variety of use cases within organizations, from policies to invoice search. However, it’s important to note that the tool is still very new, with only a handful of customers testing it in beta, so the jury is still out on performance.

Ofer Mendelevitch, head of developer relations at Vectara, told VentureBeat that because Portal runs on Vectara’s proprietary RAG-as-a-service platform, the company expects to see mass adoption from non-developers, which will boost support for the company’s full-fledged, enterprise-grade products.

“We’re eager to see what users build with the Vectara Portal, and we hope that the level of accuracy and relevance enhanced by users’ documents will demonstrate the full power of (Vectara’s) enterprise RAG system,” he said.

How does Vectara Portal work?

The Portal is available as a Vectara hosted app and as an open source product under the Apache 2.0 license. Vectara Portal is based on the idea that users can create portals (custom applications) and then share them with their intended audience for use.

First, users need to create a portal account using their main Vectara account credentials and configure their profile with their Vectara ID, API key, and OAuth client ID. Once the profile is ready, users just need to go to the “Create Portal” button and fill in some basic details like the name of their planned app, description, whether it will function as a semantic search tool, summarization app, or conversational chat assistant. After that, they can press the create button and the app will be added to the portal management page of the tool.

Once created, users can open the portal through the portal admin screen, go to settings, and add any number of documents to customize the app to their data. Once these files are uploaded, they are indexed by Vecatara’s RAG services platform, which powers the portal’s backend, to provide accurate and unambiguous answers.

“This (platform) means a powerful search engine, state-of-the-art Boomerang embedding models, multilingual re-ranking, reduced artifacts, and improved overall quality of responses to user questions on the Portal. It’s a no-code product, so developers can quickly create AI products with just a few clicks,” Mendelevich said.

The Developer Relations Officer pointed out that when a user creates a portal and adds documents, the tool’s backend builds a “corpus” specific to that data in the user’s main Vectara account. This corpus acts as a place to hold all documents related to the portal. So when a user asks a question on the portal, Vectara’s RAG API runs that query against the relevant corpus to derive the most relevant answer.

The platform first picks the most relevant parts of the document needed to answer the user’s question (in the search step) and feeds them into a large-scale language model (LLM). Vectara gives users the option to choose from a variety of LLMs, including its own Mockingbird LLM and OpenAI’s LLM.

“For Vectara Scale (the company’s larger plan) customers, Portal will use the best of Vectara features, including the most powerful LLM,” Mendelevitch added. The app is public by default and can be shared via link, but users can also limit it to specific groups of users.

Aiming to increase corporate customers

By offering both hosted and open-source no-code products, Vectara hopes to empower more enterprise users with the ability to build powerful generative AI apps targeting a variety of use cases. The company hopes that increased sign-ups and buzz around its flagship RAG service will ultimately lead to higher conversion rates.

“RAG is a very powerful use case for many enterprise developers and we wanted to expose it to no-code builders so they can understand the power of Vectara’s end-to-end platform. The Portal does just that and we think it will be a valuable tool for product managers, general managers and other C-level executives to understand how Vectara can help their gen AI use cases,” said Mendelevitch.

The company has raised over $50 million in funding to date and has approximately 50 production customers, including Obeikan Group, Juniper Networks, Sonosim and Qumulo.