Neynar

Celebrating the experimental launch of hubs.neynar.com

I have 4 uninterrupted hours of focus time today and I’m going to figure out how to give my @gmfc101 bot a memory so it can have a conversation with someone and not just reply to one-off casts. 

Could not be happier. Wish me luck.

Giving my bot a memory of its conversation was a little more complicated than I thought it would be.  

Is this the right approach:
- Use neynar conversation API to retrieve all casts from a thread (https://docs.neynar.com/reference/lookup-cast-conversation)
- Apply filtering logic to only get casts between the user & the bot
- Pass the conversation history back into LLM prompt to remind bot of its past conversation

I couldn't get through the first 2 steps today and I ran out of time.  Will pick back up later but if anyone has done something similar and has any guidance/advice I'm all ears.

Cohost /gmfarcaster livestream Farcaster news show 🎙️💬 Founder Tenger Ways, bringing joy into work with DevOps practices 🤗 mom x 3 💗

Depends on what you’re trying to do with the conversation memory. In about all cases it’s better to distill it down in some way

Building Higher apparel at High Line 🧢 
Host at /clickcast /cpg || Husband to @shiwen

I can see how that can be challenging. Even if you managed to pull all conversations tagging the bot, it’s no guarantee it’d pickup on nuance from older conversations. My half measure without complex RAG systems would be to pull the last thread someone interacted with the bot (if there is one) and potentially use the LLM to distill the contents of it

Right now my bot will answer you if you tag it, but it only has the context of whatever you wrote in the cast where you tag it.

I want to be able to have it continue a conversation with someone. Just like how you and I might do in this thread... 

Like if I ask the bot a clarifying question, it won't remember what we were just talking about.