Why it matters
This is one of the few events that treats inference reliability as a first-class engineering topic: retries, circuit breakers, rate limiting, and caching are exactly the controls that keep agent spend predictable in production.
The tokenmaxxing angle
Caching and circuit breakers are cost controls as much as reliability ones, they stop retried or runaway calls from multiplying token spend, and the workshop builds them alongside the RAG pipeline instead of as an afterthought.
From the organizers
Led by host Miya L. at 399 Boylston St, Boston, Tue Aug 18, 6-9pm; attendees code along on laptops and clone a repo covering ChatClient, streaming, pgvector RAG, retries, and observability.