Closed ErezSha closed 4 months ago
@asaf I closed the previous PR (#144 ) because I needed to fix the commits. I'll copy the comments here:
@ErezSha Is this another LLM call and going to increase latency? we'r already at > 20s
That's the only way to do it reliably. It calls gpt-4o, so it's faster. Also, it's happening asynchronously, in parallel, with the other call entry point makes, so it's doesn't increase latency.
implements an on_topic guardrail for the graph
closes #138