Increase LLM timeouts to 3m per turn and 10m per request (#8789)

Author: begelundmuller

runtime/ai/ai.go

@@ -1078,7 +1078,7 @@ func (s *Session) CallTool(ctx context.Context, role Role, toolName string, out,
 	})
 }
 
-const llmRequestTimeout = 60 * time.Second
+const llmRequestTimeout = 3 * time.Minute
 
 // CompleteOptions provides options for Session.Complete.
 type CompleteOptions struct {
@@ -1249,6 +1249,9 @@ func (s *Session) Complete(ctx context.Context, name string, out any, opts *Comp
 
 			// Handle LLM completion error
 			if err != nil {
+				if errors.Is(err, llmCtx.Err()) && errors.Is(err, context.DeadlineExceeded) {
+					return nil, fmt.Errorf("LLM request timed out after %s: %w", llmRequestTimeout, err)
+				}
 				return nil, fmt.Errorf("completion failed: %w", err)
 			}
 

runtime/server/server.go

@@ -300,7 +300,7 @@ func timeoutSelector(fullMethodName string) time.Duration {
 	}
 
 	if fullMethodName == runtimev1.RuntimeService_Complete_FullMethodName || fullMethodName == runtimev1.RuntimeService_CompleteStreaming_FullMethodName {
-		return time.Minute * 5
+		return time.Minute * 10
 	}
 
 	if fullMethodName == runtimev1.RuntimeService_Health_FullMethodName || fullMethodName == runtimev1.RuntimeService_InstanceHealth_FullMethodName {