When rollama is chewing through a longer queue of requests, they time out at some point:
I tried httr2::req_timeout(), which sets the libcurl option timeout_ms, at first, but the 300,000 milliseconds are actually set by connecttimeout_ms. As far as I can see, the timer for this starts ticking as soon as the first request is sent, so that's why it starts to fail with a couple 100 prompts (just a theory I tested).
To fix it, I'm increasing both timeout_ms and connecttimeout_ms to 24 hours. That fixes it, but wanted to keep a record of what's going on.
When
rollamais chewing through a longer queue of requests, they time out at some point:I tried
httr2::req_timeout(), which sets the libcurl optiontimeout_ms, at first, but the 300,000 milliseconds are actually set byconnecttimeout_ms. As far as I can see, the timer for this starts ticking as soon as the first request is sent, so that's why it starts to fail with a couple 100 prompts (just a theory I tested).To fix it, I'm increasing both
timeout_msandconnecttimeout_msto 24 hours. That fixes it, but wanted to keep a record of what's going on.