feat(llama.cpp): add reranking

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

core/backend/options.go

@@ -94,6 +94,11 @@ func grpcModelOpts(c config.BackendConfig) *pb.ModelOptions {
 		lowVRAM = *c.LowVRAM
 	}
 
+	reranking := false
+	if c.Reranking != nil {
+		reranking = *c.Reranking
+	}
+
 	mmap := false
 	if c.MMap != nil {
 		mmap = *c.MMap
@@ -178,6 +183,7 @@ func grpcModelOpts(c config.BackendConfig) *pb.ModelOptions {
 		RopeFreqScale:       c.RopeFreqScale,
 		NUMA:                c.NUMA,
 		Embeddings:          embeddings,
+		Reranking:           reranking,
 		LowVRAM:             lowVRAM,
 		NGPULayers:          int32(nGPULayers),
 		MMap:                mmap,

core/config/backend_config.go

@@ -120,6 +120,7 @@ type LLMConfig struct {
 	MMap            *bool    `yaml:"mmap"`
 	MMlock          *bool    `yaml:"mmlock"`
 	LowVRAM         *bool    `yaml:"low_vram"`
+	Reranking       *bool    `yaml:"reranking"`
 	Grammar         string   `yaml:"grammar"`
 	StopWords       []string `yaml:"stopwords"`
 	Cutstrings      []string `yaml:"cutstrings"`
@@ -372,6 +373,10 @@ func (cfg *BackendConfig) SetDefaults(opts ...ConfigLoaderOption) {
 		cfg.Embeddings = &falseV
 	}
 
+	if cfg.Reranking == nil {
+		cfg.Reranking = &falseV
+	}
+
 	if threads == 0 {
 		// Threads can't be 0
 		threads = 4