This is true. I don’t feel that my current level of development competence with LLMs has matured enough to share anything yet. While I’ve used Github’s Copilot for more than a year, all my work with it is for Stream It while developing analytics systems for DOTs, which is hardly relevant to this community. Frankly, it’s boring stuff.
Correct - this blog post swayed me to believe the best approach is to use something like Pinecone. Technically, as you pointed out, it’s not required, but it sure is easier unless you are as deeply skilled as you and @curt.kennedy are in building database systems and have a deeper understanding of math, cosine, etc. Now that I’ve seen both worlds, I would still favor outsourcing for many projects, but that’s just my historical experience kicking in with a decided bias toward scalability.
Available time has not been favourable to me lately. My wife is very ill, and I’m juggling two jobs - one building the software for this, and another as a consultant trying to pay medical bills.
Yep - can do - I’ll work on that this week. It’s been my desire to add some meat to the general concept of copilots, but these things take time, especially for 70+ year old dudes with modest technology successes.