Merge branch 'master' into feat-request-middleware

Makefile

@@ -8,7 +8,7 @@ DETECT_LIBS?=true
 # llama.cpp versions
 GOLLAMA_REPO?=https://github.com/go-skynet/go-llama.cpp
 GOLLAMA_VERSION?=2b57a8ae43e4699d3dc5d1496a1ccd42922993be
-CPPLLAMA_VERSION?=3ec9fd4b77b6aca03a3c2bf678eae3f9517d6904
+CPPLLAMA_VERSION?=d774ab3acc4fee41fbed6dbfc192b57d5f79f34b
 
 # whisper.cpp version
 WHISPER_REPO?=https://github.com/ggerganov/whisper.cpp
@@ -24,7 +24,7 @@ BARKCPP_VERSION?=v1.0.0
 
 # stablediffusion.cpp (ggml)
 STABLEDIFFUSION_GGML_REPO?=https://github.com/leejet/stable-diffusion.cpp
-STABLEDIFFUSION_GGML_VERSION?=5eb15ef4d022bef4a391de4f5f6556e81fbb5024
+STABLEDIFFUSION_GGML_VERSION?=d46ed5e184b97c2018dc2e8105925bdb8775e02c
 
 ONNX_VERSION?=1.20.0
 ONNX_ARCH?=x64

core/http/endpoints/openai/chat.go

@@ -396,6 +396,11 @@ func ChatEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, evaluat
 				log.Debug().Msgf("Text content to return: %s", textContentToReturn)
 				noActionsToRun := len(results) > 0 && results[0].Name == noActionName || len(results) == 0
 
+				finishReason := "stop"
+				if len(input.Tools) > 0 {
+					finishReason = "tool_calls"
+				}
+
 				switch {
 				case noActionsToRun:
 					result, err := handleQuestion(config, input, ml, startupOptions, results, s, predInput)
@@ -403,19 +408,18 @@ func ChatEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, evaluat
 						log.Error().Err(err).Msg("error handling question")
 						return
 					}
+
 					*c = append(*c, schema.Choice{
-						Message: &schema.Message{Role: "assistant", Content: &result}})
+						FinishReason: finishReason,
+						Message:      &schema.Message{Role: "assistant", Content: &result}})
 				default:
 					toolChoice := schema.Choice{
+						FinishReason: finishReason,
 						Message: &schema.Message{
 							Role: "assistant",
 						},
 					}
 
-					if len(input.Tools) > 0 {
-						toolChoice.FinishReason = "tool_calls"
-					}
-
 					for _, ss := range results {
 						name, args := ss.Name, ss.Arguments
 						if len(input.Tools) > 0 {
@@ -433,7 +437,7 @@ func ChatEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, evaluat
 								},
 							)
 						} else {
-							// otherwise we return more choices directly
+							// otherwise we return more choices directly (deprecated)
 							*c = append(*c, schema.Choice{
 								FinishReason: "function_call",
 								Message: &schema.Message{

gallery/index.yaml

@@ -523,6 +523,36 @@
     - filename: Nohobby_L3.3-Prikol-70B-v0.4-Q4_K_M.gguf
       sha256: e1d67a40bdf0526bdfcaa16c6e4dfeecad41651e201b4009b65f4f444b773604
       uri: huggingface://bartowski/Nohobby_L3.3-Prikol-70B-v0.4-GGUF/Nohobby_L3.3-Prikol-70B-v0.4-Q4_K_M.gguf
+- !!merge <<: *llama33
+  name: "arliai_llama-3.3-70b-arliai-rpmax-v1.4"
+  urls:
+    - https://huggingface.co/ArliAI/Llama-3.3-70B-ArliAI-RPMax-v1.4
+    - https://huggingface.co/bartowski/ArliAI_Llama-3.3-70B-ArliAI-RPMax-v1.4-GGUF
+  description: |
+    RPMax is a series of models that are trained on a diverse set of curated creative writing and RP datasets with a focus on variety and deduplication. This model is designed to be highly creative and non-repetitive by making sure no two entries in the dataset have repeated characters or situations, which makes sure the model does not latch on to a certain personality and be capable of understanding and acting appropriately to any characters or situations.
+  overrides:
+    parameters:
+      model: ArliAI_Llama-3.3-70B-ArliAI-RPMax-v1.4-Q4_K_M.gguf
+  files:
+    - filename: ArliAI_Llama-3.3-70B-ArliAI-RPMax-v1.4-Q4_K_M.gguf
+      sha256: 7c79e76e5c057cfe32529d930360fbebd29697948e5bac4e4b2eb6d2ee596e31
+      uri: huggingface://bartowski/ArliAI_Llama-3.3-70B-ArliAI-RPMax-v1.4-GGUF/ArliAI_Llama-3.3-70B-ArliAI-RPMax-v1.4-Q4_K_M.gguf
+- !!merge <<: *llama33
+  name: "black-ink-guild_pernicious_prophecy_70b"
+  icon: https://huggingface.co/Black-Ink-Guild/Pernicious_Prophecy_70B/resolve/main/header.gif
+  urls:
+    - https://huggingface.co/Black-Ink-Guild/Pernicious_Prophecy_70B
+    - https://huggingface.co/bartowski/Black-Ink-Guild_Pernicious_Prophecy_70B-GGUF
+  description: |
+    Pernicious Prophecy 70B is a Llama-3.3 70B-based, two-step model designed by Black Ink Guild (SicariusSicariiStuff and invisietch) for uncensored roleplay, assistant tasks, and general usage.
+    NOTE: Pernicious Prophecy 70B is an uncensored model and can produce deranged, offensive, and dangerous outputs. You are solely responsible for anything that you choose to do with this model.
+  overrides:
+    parameters:
+      model: Black-Ink-Guild_Pernicious_Prophecy_70B-Q4_K_M.gguf
+  files:
+    - filename: Black-Ink-Guild_Pernicious_Prophecy_70B-Q4_K_M.gguf
+      sha256: d8d4874b837993546b750db3faf1c6e5d867883a6750f04f1f4986973d7c107b
+      uri: huggingface://bartowski/Black-Ink-Guild_Pernicious_Prophecy_70B-GGUF/Black-Ink-Guild_Pernicious_Prophecy_70B-Q4_K_M.gguf
 - &rwkv
   url: "github:mudler/LocalAI/gallery/rwkv.yaml@master"
   name: "rwkv-6-world-7b"
@@ -1448,7 +1478,7 @@
       sha256: 6063cf3cf90f72cfb6ad7564bca8229806cb9823a055adcbce3dc539c2a75765
       uri: huggingface://bartowski/AGI-0_Art-Skynet-3B-GGUF/AGI-0_Art-Skynet-3B-Q4_K_M.gguf
 - !!merge <<: *llama32
-  name: "localai-functioncall-llama3.2-3b-v0.5"
+  name: "LocalAI-functioncall-llama3.2-3b-v0.5"
   icon: https://cdn-uploads.huggingface.co/production/uploads/647374aa7ff32a81ac6d35d4/Dzbdzn27KEc3K6zNNi070.png
   urls:
     - https://huggingface.co/mudler/LocalAI-functioncall-llama3.2-3b-v0.5
@@ -3591,6 +3621,20 @@
     - filename: rubenroy_Gilgamesh-72B-Q4_K_M.gguf
       sha256: c6842b3bc882082c63243e762234ae697c1727bebed18b5241eb97e019f0cf68
       uri: huggingface://bartowski/rubenroy_Gilgamesh-72B-GGUF/rubenroy_Gilgamesh-72B-Q4_K_M.gguf
+- !!merge <<: *qwen25
+  name: "tiger-lab_qwen2.5-32b-instruct-cft"
+  urls:
+    - https://huggingface.co/TIGER-Lab/Qwen2.5-32B-Instruct-CFT
+    - https://huggingface.co/bartowski/TIGER-Lab_Qwen2.5-32B-Instruct-CFT-GGUF
+  description: |
+    Qwen2.5-32B-Instruct-CFT is a 32B parameter model fine-tuned using our novel Critique Fine-Tuning (CFT) approach. Built upon the Qwen2.5-32B-Instruct base model, this variant is trained to critique and analyze responses rather than simply imitate them, leading to enhanced reasoning capabilities.
+  overrides:
+    parameters:
+      model: TIGER-Lab_Qwen2.5-32B-Instruct-CFT-Q4_K_M.gguf
+  files:
+    - filename: TIGER-Lab_Qwen2.5-32B-Instruct-CFT-Q4_K_M.gguf
+      sha256: 57e87e246db368f39f31f38e44ba8e9dc838a026f729f5a123aacc2aeb5a9402
+      uri: huggingface://bartowski/TIGER-Lab_Qwen2.5-32B-Instruct-CFT-GGUF/TIGER-Lab_Qwen2.5-32B-Instruct-CFT-Q4_K_M.gguf
 - &llama31
   url: "github:mudler/LocalAI/gallery/llama3.1-instruct.yaml@master" ## LLama3.1
   icon: https://avatars.githubusercontent.com/u/153379578