fix(p2p): adapt to backend changes, general improvements (#5889)

The binary is now named "llama-cpp-rpc-server" for p2p workers.

We also decrease the default token rotation interval, in this way
peer discovery is much more responsive.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

core/cli/worker/worker_llamacpp.go

@@ -1,6 +1,7 @@
 package worker
 
 import (
+	"errors"
 	"fmt"
 	"os"
 	"path/filepath"
@@ -16,6 +17,10 @@ type LLamaCPP struct {
 	WorkerFlags `embed:""`
 }
 
+const (
+	llamaCPPRPCBinaryName = "llama-cpp-rpc-server"
+)
+
 func findLLamaCPPBackend(backendSystemPath string) (string, error) {
 	backends, err := gallery.ListSystemBackends(backendSystemPath)
 	if err != nil {
@@ -33,12 +38,12 @@ func findLLamaCPPBackend(backendSystemPath string) (string, error) {
 	}
 
 	if backendPath == "" {
-		return "", fmt.Errorf("llama-cpp backend not found")
+		return "", errors.New("llama-cpp backend not found, install it first")
 	}
 
 	grpcProcess := filepath.Join(
 		backendPath,
-		"grpc-server",
+		llamaCPPRPCBinaryName,
 	)
 
 	return grpcProcess, nil