Post

How do y'all avoid having seven+ terminal windows open with agent conversations that are probably over? but maybe you might want to pick back up a few days from now? seems a shame to throw away all that exploration & context, right?

we chunk and summarize everyone’s agent transcripts, compute text embeddings, and do RAG over the result. it works insanely well

This, but just use agentsview.io to do it for you instead of building it yourself. Shout out to @wesmckinney.com

we deliberately wanted something that worked without needing to scrape the local session logs but this is also cool 👍

How do you do it without?

we forked agentgateway.dev to robustly persist all inference request/response bodies to s3 as gzipped jsonl batches

session reconstruction from the raw inference call logs happens out of band

session reconstruction must be hairy right? i use a hook that emits turn-by-turn tracing, but the drawback there is I need to install it on the client

most sessions are driven by a single writer and most agents pass a persistent ID around (even for sub agent sessions) so it’s actually pretty easy. esp since you’re always sending the full context window so it’s not like you’re stitching stuff together