Tripling your Data, Keeping the Price

Q&A AI Chatbots are built from 'facts'. One fact can be a web page, a page in a PDF file, or some other piece of fact that contributes to the AI chatbot's abilities. Today we are happy to inform you that all of your cloudlets have been upgraded, and they can now hold up to 3 times more information than before. And, we're doing this without changing our prices in an ways.
This implies that we can deliver AI chatbots and agents with 3x as much information as before, and you get to leverage this from today, without any extra costs.
Basically, 3x more data, same inexpensive price!
How?
An AI chatbot built upon RAG needs a VSS library. A VSS library again is a tool that allows us to match questions from our users to "training snippets" or "facts" from your database. Such a database can be built by uploading PDF files, scraping and crawling websites, or manually creating individual facts for that matter. This implies that the larger the database, the "smarter" the AI chatbot/agent.
We've just completely replaced our legacy VSS library with a brand new library, that instead of storing the whole index in memory, can parse directly from disc. This implies that previously a database of 30,000 snippets required several gigabytes of memory to just keep this "index" in memory. By parsing directly from disc instead, we can scan the same database without consuming more than some 50 to 100 megabytes of data.
The way we pulled this through, was by collaborating with Marco Bambini on his SQLite vector library. Marco was extremely helpful, and even allowed us to embed his work into our platform, in addition to spending several hours with us to make sure his library was performing according to our standards. And we are now proud to announce ...
We basically have the best VSS library that exists out there
This again results in no Pinecone requirements, just doing the VSS filtering straight into our SQLite database. Which again results in zero latency, much less complex code and architecture, and basically an AI chatbot or agent that simply performs superior, and can do things no other chatbots/agent platform can do. The latter is true because we can join within SQLite, allowing us to do things others simply can't.
Why?
There are a bajillion AI chatbot companies out there, yet still a lot of companies choose to even implement their own - Something I can understand considering the quality of most of "our competitors". In an evolutionary perspective, this makes perfect sense, given the point in time we're at now. 30 years ago, "everybody" was still building their websites manually. Today WordPress owns 80% of the World Wide Web or something.
However, contrary to most others in the segment, we're her to build quality products that are unique and can deliver what others cannot. This sometimes makes us slower in the short run, since others could build their tech stack on top of Pinecone, and other "soon to be bankrupt ideas". To understand why, is to understand that 2 years ago Pinecone was the only real alternative, if you wanted to come out with a product without sacrifising 500,000 on C developers. So all our competitors built on Pinecone and similar "hosted VSS database experiences".
However, when you separate your primary database need, which is to run queries, containing joins, aggregates, and pivot tables (group by SQL), and your primary VSS and vector database - Then executing even ridiculously simple stuff such as "where date_created > now() - '20 days'" (pseudo code) becomes ipso facto impossible. I always knew this of course, so instead of taking short cuts, we literally built our own. It wasn't as "scalable" as the others, but I'd rather get my architectural foundation correctly applied first, than to accept "a bajillion" customers onto an inferior platform. So we instead chose to build our own, which allowed us to perform much more complex querying into our database, which fused together the VSS parts with the traditional RDBMS parts.
This gave is a "long term unique salespoint", due to the fact that we can simply answer yes to all the stuff your existing AI chatbot provider has to answer no to - So hence, I would deeply advice you to examine what our platform can do for you, that your existing one cannot - Because I'm fairly confident in that the answer to that question will blow you out of the water ...
And to further the amount of disruption, we've even made the whole platform available as open source. Why you may ask? Well, here's my answer ...
There are simply so many garbage AI chatbot providers out there, that somebody had to do something about it. And if you can't afford to actually pay for one, and that's the reason you're using the junk you're currently using? Well, feel free to use ours for free in a DIY manner. Because NOBODY should have to use the junk you're currently using! 😉
Now today of course, the above is beginning to become a commodity, and most of these chatbot providers are desperately trying to swap out their existing "SaaS" VSS database provider with some database plugin, which exists for both PostgreSQL, MySQL, Microsoft SQL Studio, and (as of now, thx to us and Marco) also SQLite. Most of them will fail though, because of their need to avoid breaking backwards compatibility for their existing clients, so their only option for survival is to hope their customers doesn't "discover" the alternatives out there, and pray no customers ever need a simple join
SQL ...
This implies that every single time you ask your existing AI chatbot provider or AI agent provider asks the following question; "Can you guys do this?" - And the answer they give you is ...
I am sorry Sir, but we cannot do that!
... then our answer will highly likely be; "Of course mate, do you want it before lunch?"
Sorry, it's nothing personal, just biznizz 😎
And yes, if you're now wondering "how much is really 100,000 training snippets"? Well, I guarantee you that you'd probably pay more for that at Pinecone, doing nothing but your existing AI chatbot provider's VSS database ... 😂
And yes, I am sorry to tell my competition that 43 years of software development experience made me realise that VSS is just a "column type". I tried to tell you, but you wouldn't listen to me. I guess you were too busy becoming "bajillionairs" to have time for me ... 😉