Building a Farcaster bot in public (cont.)

Someone asked me why I put a limit on the length of my bot's conversation (4 replies max). The main reason is to control cost, while still in development. If someone wants to get in an infinite conversation with a bot, there are plenty of other options out there.  Doesn't need to be mine.

And that got me thinking about what differentiates my bot: its access to /gmfarcaster101 videos. 

The transcripts I fed it weren't perfect.  It thought @nounishprof was Nannish Prav, and it thought degen was dGen.  

I decided to focus on improving its responses, to correctly mention Nounish Prof and refer to degen correctly. (Garbage in, garbage out)

Solution: Search & replace the transcript embeddings and refresh my vector DB  

It was a good learning experience.  Instead of focusing on additional bot capabilities I'll keep focused on context for a while - improving the transcript quality AND maybe start loading regular /gmfarcaster episodes in addition to Farcaster 101.

It's kind of fun to do the impossible."— Walt Disney @gyoza