Overview
I've been working on a Generative AI solution at Rightpoint for the past 5 months. We are calling it Spark OmniOne. Spark is our brand name for solution assets or accelerators. OmniOne playing on the concept of an all knowing one and inspiration from OpenAI o1 model naming.
I believe it is a composable way to build an Agentic AI solution. It builds on top of an Enterprise Search platform such as Coveo and takes a best in breed approach to assemble parts of the solution from leading platforms. Examples include:
- NextJS as the Front End
- Coveo for Enterprise Search and Passage / Chunk API using vector search
- Vercel AI SDK for dynamic tool AI orchestration
- Azure Logic Apps as middleware integration for 2 way actions in corporate systems (Salesforce, Jira, ServiceNow)
- LLM Models such as Azure OpenAI GPT-4o, Claude 3.5 sonnet
- Headless CMS Integration
By leveraging a composable architecture you can plug and play different parts of the stack as you need additional capability or vendors release new breakthroughs in AI. It doesn't keep you locked into paying one platform $30 a month per user for your entire corporation.
Design
In addition to a great technical platform, we also had our design team re-think the digital assistant workspace to build a fully personalized AI assistant UI that can bring you information from your network and also take action in your enterprise systems. Here is a peek at the design:
Coveo
We chose Coveo to power the platform due to the Passage Retrieval API and all of the benefits that it brings. Learn more on my past blog post. I believe that this solution can go head to head with many competitors in the industry such as Glean, Microsoft Copilot, Salesforce AgentForce just to name a few.
Demo
Check out our microsite and demo at: https://sparkomnione.ai