Not Another Chatbot: How Squid AI's Universal RAG Goes Beyond ChatGPT

Jen Person|Sep 03, 2024

Given that AI is at the heart of Squid’s technical offerings, it’s natural to wonder, “How is Squid AI different from ChatGPT?” As we attend events to connect with AI innovators and developers, we have been asked this question many times. The short answer is that Squid connects to your data sources and provides AI agent building blocks, allowing you to create custom AI agents and workflows that are experts on your resources. Squid’s AI agent building blocks work on your data no matter where it’s stored and no matter what scale, handling tables with millions of rows or multiple PDFs totaling thousands of pages. The long answer gets into why this distinction is so important, and since you’re not stopping by our booth between talks, we have time to get into the weeds.

Flexible and Unopinionated

Squid AI’s platform allows you to connect and secure any database or API and integrate your sources into your AI solution in only a few lines of code. Whether your database is SQL or NoSQL, whether your API is HTTP or GraphQL, you can connect your source in the Squid Console and get insights from your data right away. This feature is essential for companies looking to incorporate AI agents and workflows into their tech stack, as it makes it possible to add AI without the need for expensive and time-consuming migration and modernization. Squid meets companies where they are, providing a way to add GenAI solutions to enterprise at a speed that keeps up with the ever-changing landscape.

In addition to integrating with any data source, Squid AI integrates with any LLM, giving you the freedom to drop in the model that works best for your use case. Choose your model and be free to change your mind as new innovative models are released. Squid AI works with the following models:

  • GPT-4o
  • GPT-4o Mini
  • GPT-4o Turbo
  • GPT-4
  • GPT 3.5
  • Gemini Pro
  • Claude 3 Opus
  • Claude 3 Sonnet
  • Claude 3.5 Sonnet
  • Claude 3 Haiku

All models are served from an enterprise account, meaning your data is never used to train LLMs. In addition, if your company prefers to provide a custom LLM or to self-host a model like Llama, Squid will work with you to make it happen.

Squid AI’s goal is to be completely flexible and un-opinionated so that it works with your tech stack no matter what you use.

Expert AI Pipeline Implementation

Squid AI performs universal Retrieval-Augmented Generation (RAG) AI on your data to provide insights and answers that go beyond what was trained within a given model. The reason we call it universal is that it works for any data source. To find out more about our universal RAG implementation, check out our blog post on the topic. At a high level, all RAG consists of chunking a data source, turning those chunks into vectors using a very large series of mathematical equations known as an embedding model, and determining which vector chunks to provide a large language model so that it can respond based on how it compares this new vector information to its own vectorized knowledge. There are several tools and methods available for managing the steps of universal RAG. The experts at Squid opted to build our own universal RAG pipeline that yielded high quality results with little of the hallucination that LLMs are notorious for.

Natural Language Querying on your Data

With Squid’s Query with AI feature built into the console, you can ask questions about your data and Squid AI will perform the correct SQL or NoSQL query on your behalf, allowing you to get answers on the fly without needing to write code. You can even generate graphs and charts of your data right in the console. When you’re ready to integrate Query with AI into your application, the process is very straightforward, making it easy to build solutions your team can use without needing to be a technical expert. You can also automate your code so Squid can automatically create charts or run queries at your desired interval.

Since ChatGPT alone doesn’t know your schema, it cannot reliably generate database queries, nor can it provide results from your database. Therefore, it cannot create charts or graphs from data stored in your database without you first doing the work of getting the data and providing it to ChatGPT. While many individual database solutions are working on GenAI querying capabilities, they’re each limited to their own data offering and therefore cannot provide a complete solution to meet the needs of enterprise. Companies have multiple disparate data sources that they need to analyze, and Squid provides the connectors to run Query with AI on any of these sources, giving you a centralized and complete solution.

Incorporate Unstructured Data

In addition to performing universal RAG on your structured data sources, Squid AI can also use unstructured sources like documents as context. For any given prompt, you may need your AI agent to parse through thousands of pages of PDFs. You might want to extract images and get information from them. And you’ll likely want to connect this information with your structured data sources from databases and APIs. With Squid’s connectors and straightforward APIs, it’s easy to incorporate unstructured data into your universal RAG solution, be it on the fly or automated in data processing pipeline.

Your Complete GenAI Solution

While both ChatGPT and Squid AI offer RAG-based conversations, Squid transcends far beyond those capabilities, giving enterprise access to the myriad benefits of GenAI in less time and using less resources than ever possible. To summarize, Squid’s platform sets itself apart through the following key benefits:

  • Flexible and un-opinionated: Connect all of your data sources in minutes, allowing you to add AI agents and workflows without the need for tech stack migration.
  • Expert AI pipeline implementation: A fully custom AI pipeline built by the engineering experts at Squid.
  • Native solutions for structured data: Connect a data source and use Query with AI right away to get insights and generate charts.
  • Incorporate unstructured data: Go beyond attaching files. Connect unstructured and structured data to build solutions that move your company forward.

If you’re interested in seeing how Squid AI can bring your company into the modern AI age, then reach out to us to schedule a conversation.