pub fn build_http_client() -> Result<Client>Expand description
Build a client suitable for multi-KB JSON prompts and slow inference APIs.
- Long connect (5 min) and overall (1 h) timeouts for slow inference.
- HTTP/1.1 only: avoids sporadic HTTP/2 stalls seen with some cloud gateways.
- TCP keepalive so idle long responses are less likely to be dropped.