ltcai 2.1.0 → 2.2.1

package/latticeai/server_app.py

@@ -1,6 +1,6 @@
 """
 Lattice AI MLX — Local LLM Bridge Server
-Apple Silicon (M1-M5) 전용 | mlx-lm 기반
+Apple Silicon (M1-M5) 전용 | MLX-VLM 기반
 """
 
 import asyncio

package/latticeai/services/model_catalog.py

@@ -17,8 +17,8 @@ from typing import Dict, List, Optional
 
 ENGINE_INSTALLERS = {
     "local_mlx": {
-        "command": [sys.executable, "-m", "pip", "install", "--upgrade", "mlx-lm", "mlx-vlm", "huggingface_hub[cli]"],
-        "label": "Install MLX runtime",
+        "command": [sys.executable, "-m", "pip", "install", "--upgrade", "mlx-vlm", "huggingface_hub[cli]"],
+        "label": "Install MLX-VLM runtime",
     },
     "openai": {
         "command": [sys.executable, "-m", "pip", "install", "openai"],
@@ -61,170 +61,109 @@ ENGINE_INSTALLERS = {
     },
 }
 
+def _model(
+    model_id: str,
+    name: str,
+    family: str,
+    tag: str,
+    size: str,
+    *,
+    source_country: str,
+    source_company: str,
+    execution_method: str,
+    internet_requirement: str = "모델을 다운로드할 때만 인터넷 필요; 실행 중에는 필요 없음",
+    pullable: bool = True,
+) -> Dict[str, object]:
+    clean_model_name = re.split(r"\s+via\s+", name, maxsplit=1)[0]
+    return {
+        "id": model_id,
+        "name": name,
+        "model_name": clean_model_name,
+        "family": family,
+        "tag": tag,
+        "size": size,
+        "pullable": pullable,
+        "modality": "multimodal",
+        "source_country": source_country,
+        "source_company": source_company,
+        "execution_method": execution_method,
+        "run_location": "내 컴퓨터에서만 실행",
+        "internet_requirement": internet_requirement,
+        "source_display_order": [
+            "source_country",
+            "source_company",
+            "execution_method",
+            "internet_requirement",
+            "model_name",
+        ],
+    }
+
+
+_RUNS_ON_THIS_COMPUTER = "내 컴퓨터에서만 실행"
+
+
 ENGINE_MODEL_CATALOG = {
     "local_mlx": [
-        {"id": "mlx-community/SmolLM-1.7B-Instruct-4bit", "name": "SmolLM 1.7B", "family": "SmolLM", "tag": "local-light", "size": "963MB", "pullable": True},
-        {"id": "mlx-community/gemma-3-1b-it-4bit", "name": "Gemma 3 1B", "family": "Gemma 3", "tag": "local-light", "size": "733MB", "pullable": True},
-        {"id": "mlx-community/Llama-3.2-1B-Instruct-4bit", "name": "Llama 3.2 1B", "family": "Llama 3.x", "tag": "local-light", "size": "1.3GB", "pullable": True},
-        {"id": "mlx-community/gemma-2-2b-it-4bit", "name": "Gemma 2 2B", "family": "Gemma 2", "tag": "local-light", "size": "1.6GB", "pullable": True},
-        {"id": "mlx-community/gemma-4-e2b-4bit", "name": "Gemma 4 E2B Base", "family": "Gemma 4", "tag": "local-vlm", "size": "3.6GB", "pullable": True},
-        {"id": "mlx-community/gemma-4-e2b-it-4bit", "name": "Gemma 4 E2B Instruct", "family": "Gemma 4", "tag": "local-vlm", "size": "3.6GB", "pullable": True},
-        {"id": "mlx-community/gemma-4-e4b-4bit", "name": "Gemma 4 E4B Base", "family": "Gemma 4", "tag": "local-vlm", "size": "5.2GB", "pullable": True},
-        {"id": "mlx-community/gemma-4-e4b-it-4bit", "name": "Gemma 4 E4B Instruct", "family": "Gemma 4", "tag": "local-vlm", "size": "5.2GB", "pullable": True},
-        {"id": "mlx-community/Qwen3-VL-4B-Instruct-4bit", "name": "Qwen3-VL 4B", "family": "Qwen3-VL", "tag": "local-vlm", "size": "2.7GB", "pullable": True},
-        {"id": "mlx-community/Qwen3-VL-8B-Instruct-4bit", "name": "Qwen3-VL 8B", "family": "Qwen3-VL", "tag": "local-vlm", "size": "4.8GB", "pullable": True},
-        {"id": "mlx-community/Qwen2.5-VL-7B-Instruct-4bit", "name": "Qwen2.5-VL 7B", "family": "Qwen2.5-VL", "tag": "local-vlm", "size": "4.4GB", "pullable": True},
-        {"id": "mlx-community/gemma-3-4b-it-4bit", "name": "Gemma 3 4B", "family": "Gemma 3", "tag": "local-vlm", "size": "3.3GB", "pullable": True},
-        {"id": "mlx-community/Llama-3.2-3B-Instruct-4bit", "name": "Llama 3.2 3B", "family": "Llama 3.x", "tag": "local-general", "size": "2.0GB", "pullable": True},
-        {"id": "mlx-community/Llama-3.1-8B-Instruct-4bit", "name": "Llama 3.1 8B", "family": "Llama 3.1", "tag": "local-general", "size": "4.7GB", "pullable": True},
-        {"id": "mlx-community/gemma-2-9b-it-4bit", "name": "Gemma 2 9B", "family": "Gemma 2", "tag": "local-general", "size": "5.4GB", "pullable": True},
-        {"id": "mlx-community/gemma-3-12b-it-4bit", "name": "Gemma 3 12B", "family": "Gemma 3", "tag": "local-vlm", "size": "8.0GB", "pullable": True},
-        {"id": "mlx-community/Phi-3.5-mini-instruct-4bit", "name": "Phi 3.5 Mini", "family": "Phi", "tag": "local-coding", "size": "2.2GB", "pullable": True},
-        {"id": "mlx-community/Phi-4-mini-instruct-4bit", "name": "Phi 4 Mini", "family": "Phi", "tag": "local-coding", "size": "2.2GB", "pullable": True},
-        {"id": "mlx-community/phi-4-4bit", "name": "Phi 4", "family": "Phi", "tag": "local-coding", "size": "8.3GB", "pullable": True},
-        {"id": "mlx-community/Mistral-7B-Instruct-v0.3-4bit", "name": "Mistral 7B Instruct v0.3", "family": "Mistral", "tag": "local-general", "size": "4.1GB", "pullable": True},
-        {"id": "mlx-community/Ministral-8B-Instruct-2410-4bit", "name": "Ministral 8B Instruct", "family": "Mistral", "tag": "local-general", "size": "4.5GB", "pullable": True},
-        {"id": "mlx-community/Mistral-Small-24B-Instruct-2501-4bit", "name": "Mistral Small 24B", "family": "Mistral", "tag": "local-large", "size": "13.3GB", "pullable": True},
-        {"id": "mlx-community/Qwen2.5-Coder-32B-Instruct-4bit", "name": "Qwen2.5 Coder 32B", "family": "Qwen2.5", "tag": "local-coding", "size": "18.5GB", "pullable": True},
-        {"id": "mlx-community/Qwen3-VL-30B-A3B-Instruct-4bit", "name": "Qwen3-VL 30B A3B", "family": "Qwen3-VL", "tag": "local-vlm", "size": "18GB", "pullable": True},
-        {"id": "mlx-community/gemma-3-27b-it-4bit", "name": "Gemma 3 27B", "family": "Gemma 3", "tag": "local-vlm", "size": "17GB", "pullable": True},
-        {"id": "mlx-community/gemma-4-26b-a4b-it-4bit", "name": "Gemma 4 26B A4B Instruct", "family": "Gemma 4", "tag": "local-vlm", "size": "15.6GB", "pullable": True},
-        {"id": "mlx-community/gemma-4-31b-it-4bit", "name": "Gemma 4 31B Instruct", "family": "Gemma 4", "tag": "local-vlm", "size": "18.4GB", "pullable": True},
-        {"id": "mlx-community/gpt-oss-20b-MXFP4-Q8", "name": "GPT-OSS 20B", "family": "GPT-OSS", "tag": "local-reasoning", "size": "12.1GB", "pullable": True},
-        {"id": "mlx-community/gpt-oss-120b-MXFP4-Q4", "name": "GPT-OSS 120B", "family": "GPT-OSS", "tag": "local-large", "size": "62.3GB", "pullable": True},
-        {"id": "mlx-community/Llama-3.3-70B-Instruct-4bit", "name": "Llama 3.3 70B", "family": "Llama 3.x", "tag": "local-general", "size": "40GB+", "pullable": True},
-        {"id": "mlx-community/Llama-3.1-70B-Instruct-4bit", "name": "Llama 3.1 70B", "family": "Llama 3.1", "tag": "local-general", "size": "40GB+", "pullable": True},
+        _model("mlx-community/gemma-4-e2b-4bit", "Gemma 4 E2B Base", "Gemma 4", "local-vlm", "3.6GB", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("mlx-community/gemma-4-e2b-it-4bit", "Gemma 4 E2B Instruct", "Gemma 4", "local-vlm", "3.6GB", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("mlx-community/gemma-4-e4b-4bit", "Gemma 4 E4B Base", "Gemma 4", "local-vlm", "5.2GB", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("mlx-community/gemma-4-e4b-it-4bit", "Gemma 4 E4B Instruct", "Gemma 4", "local-vlm", "5.2GB", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("mlx-community/gemma-4-12b-it-4bit", "Gemma 4 12B Instruct", "Gemma 4", "local-vlm", "7.6GB", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("mlx-community/gemma-4-26b-a4b-it-4bit", "Gemma 4 26B A4B Instruct", "Gemma 4", "local-vlm", "15.6GB", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("mlx-community/gemma-4-31b-it-4bit", "Gemma 4 31B Instruct", "Gemma 4", "local-vlm", "18.4GB", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("mlx-community/Qwen3-VL-4B-Instruct-4bit", "Qwen3-VL 4B", "Qwen3-VL", "local-vlm", "2.7GB", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("mlx-community/Qwen3-VL-8B-Instruct-4bit", "Qwen3-VL 8B", "Qwen3-VL", "local-vlm", "4.8GB", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("mlx-community/Qwen3-VL-30B-A3B-Instruct-4bit", "Qwen3-VL 30B A3B", "Qwen3-VL", "local-vlm", "18GB", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("mlx-community/Llama-4-Scout-17B-16E-Instruct-4bit", "Llama 4 Scout 17B 16E", "Llama 4", "local-vlm", "11.8GB", source_country="미국", source_company="Meta", execution_method=_RUNS_ON_THIS_COMPUTER),
     ],
     "ollama": [
-        {"id": "ollama:qwen3-vl:4b", "name": "Qwen3-VL 4B via Ollama", "family": "Qwen3-VL", "tag": "local-vlm", "size": "pull required", "pullable": True},
-        {"id": "ollama:qwen3-vl:8b", "name": "Qwen3-VL 8B via Ollama", "family": "Qwen3-VL", "tag": "local-vlm", "size": "pull required", "pullable": True},
-        {"id": "ollama:qwen3-vl:30b", "name": "Qwen3-VL 30B via Ollama", "family": "Qwen3-VL", "tag": "local-vlm", "size": "pull required", "pullable": True},
-        {"id": "ollama:gpt-oss:20b", "name": "GPT-OSS 20B via Ollama", "family": "GPT-OSS", "tag": "local-reasoning", "size": "pull required", "pullable": True},
-        {"id": "ollama:gpt-oss:120b", "name": "GPT-OSS 120B via Ollama", "family": "GPT-OSS", "tag": "local-large", "size": "pull required", "pullable": True},
-        {"id": "ollama:hf.co/ggml-org/gemma-4-31B-it-GGUF:Q4_K_M", "name": "Gemma 4 31B Q4 via Ollama", "family": "Gemma 4", "tag": "local-vlm", "size": "18.7GB", "pullable": True},
-        {"id": "ollama:qwen3:8b", "name": "Qwen3 8B via Ollama", "family": "Qwen", "tag": "local-server", "size": "pull required", "pullable": True},
-        {"id": "ollama:qwen2.5-coder:14b", "name": "Qwen2.5 Coder 14B via Ollama", "family": "Qwen", "tag": "local-coding", "size": "pull required", "pullable": True},
-        {"id": "ollama:gemma3:1b", "name": "Gemma 3 1B via Ollama", "family": "Gemma", "tag": "local-light", "size": "pull required", "pullable": True},
-        {"id": "ollama:gemma3:4b", "name": "Gemma 3 4B via Ollama", "family": "Gemma", "tag": "local-server", "size": "pull required", "pullable": True},
-        {"id": "ollama:gemma3:4b-it-q4_K_M", "name": "Gemma 3 4B q4_K_M via Ollama", "family": "Gemma", "tag": "quantized", "size": "pull required", "pullable": True},
-        {"id": "ollama:gemma3:12b", "name": "Gemma 3 12B via Ollama", "family": "Gemma", "tag": "local-server", "size": "pull required", "pullable": True},
-        {"id": "ollama:gemma3:12b-it-q4_K_M", "name": "Gemma 3 12B q4_K_M via Ollama", "family": "Gemma", "tag": "quantized", "size": "pull required", "pullable": True},
-        {"id": "ollama:gemma3:27b", "name": "Gemma 3 27B via Ollama", "family": "Gemma", "tag": "local-large", "size": "pull required", "pullable": True},
-        {"id": "ollama:llama3.2:1b", "name": "Llama 3.2 1B via Ollama", "family": "Llama 3.x", "tag": "local-light", "size": "pull required", "pullable": True},
-        {"id": "ollama:llama3.2:3b", "name": "Llama 3.2 3B via Ollama", "family": "Llama 3.x", "tag": "local-server", "size": "pull required", "pullable": True},
-        {"id": "ollama:llama3.1:8b", "name": "Llama 3.1 8B via Ollama", "family": "Llama 3.1", "tag": "local-server", "size": "pull required", "pullable": True},
-        {"id": "ollama:llama3.1:8b-instruct-q4_0", "name": "Llama 3.1 8B q4_0 via Ollama", "family": "Llama 3.1", "tag": "quantized", "size": "pull required", "pullable": True},
-        {"id": "ollama:llama3.1:8b-instruct-q8_0", "name": "Llama 3.1 8B q8_0 via Ollama", "family": "Llama 3.1", "tag": "quantized", "size": "pull required", "pullable": True},
-        {"id": "ollama:llama3.1:70b", "name": "Llama 3.1 70B via Ollama", "family": "Llama 3.1", "tag": "local-server", "size": "pull required", "pullable": True},
-        {"id": "ollama:llama3.3:70b", "name": "Llama 3.3 70B via Ollama", "family": "Llama 3.x", "tag": "local-large", "size": "pull required", "pullable": True},
-        {"id": "ollama:mistral:7b", "name": "Mistral 7B via Ollama", "family": "Mistral", "tag": "local-server", "size": "pull required", "pullable": True},
-        {"id": "ollama:mixtral:8x7b", "name": "Mixtral 8x7B via Ollama", "family": "Mistral", "tag": "local-large", "size": "pull required", "pullable": True},
-        {"id": "ollama:phi4-mini", "name": "Phi 4 Mini via Ollama", "family": "Phi", "tag": "local-coding", "size": "pull required", "pullable": True},
-        {"id": "ollama:phi4", "name": "Phi 4 via Ollama", "family": "Phi", "tag": "local-coding", "size": "pull required", "pullable": True},
-        {"id": "ollama:smollm2:1.7b", "name": "SmolLM2 1.7B via Ollama", "family": "SmolLM", "tag": "local-light", "size": "pull required", "pullable": True},
-        {"id": "ollama:deepseek-r1:1.5b", "name": "DeepSeek-R1 1.5B via Ollama", "family": "DeepSeek", "tag": "local-light", "size": "pull required", "pullable": True},
-        {"id": "ollama:deepseek-r1:7b", "name": "DeepSeek-R1 7B via Ollama", "family": "DeepSeek", "tag": "local-reasoning", "size": "pull required", "pullable": True},
-        {"id": "ollama:deepseek-r1:8b", "name": "DeepSeek-R1 8B via Ollama", "family": "DeepSeek", "tag": "local-reasoning", "size": "pull required", "pullable": True},
-        {"id": "ollama:deepseek-r1:14b", "name": "DeepSeek-R1 14B via Ollama", "family": "DeepSeek", "tag": "local-reasoning", "size": "pull required", "pullable": True},
-        {"id": "ollama:deepseek-r1:32b", "name": "DeepSeek-R1 32B via Ollama", "family": "DeepSeek", "tag": "local-large", "size": "pull required", "pullable": True},
-        {"id": "ollama:deepseek-coder-v2:16b", "name": "DeepSeek-Coder-V2 16B via Ollama", "family": "DeepSeek", "tag": "local-coding", "size": "pull required", "pullable": True},
+        _model("ollama:qwen3-vl:4b", "Qwen3-VL 4B via Ollama", "Qwen3-VL", "local-vlm", "pull required", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("ollama:qwen3-vl:8b", "Qwen3-VL 8B via Ollama", "Qwen3-VL", "local-vlm", "pull required", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("ollama:qwen3-vl:30b", "Qwen3-VL 30B via Ollama", "Qwen3-VL", "local-vlm", "pull required", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("ollama:hf.co/ggml-org/gemma-4-12B-it-GGUF:Q4_K_M", "Gemma 4 12B Q4 via Ollama", "Gemma 4", "local-vlm", "7.9GB", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("ollama:hf.co/ggml-org/gemma-4-31B-it-GGUF:Q4_K_M", "Gemma 4 31B Q4 via Ollama", "Gemma 4", "local-vlm", "18.7GB", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("ollama:hf.co/ggml-org/Llama-4-Scout-17B-16E-Instruct-GGUF:Q4_K_M", "Llama 4 Scout Q4 via Ollama", "Llama 4", "local-vlm", "12GB", source_country="미국", source_company="Meta", execution_method=_RUNS_ON_THIS_COMPUTER),
     ],
     "vllm": [
-        {"id": "vllm:openai/gpt-oss-20b", "name": "GPT-OSS 20B via vLLM", "family": "GPT-OSS", "tag": "local-reasoning", "size": "server model", "pullable": True},
-        {"id": "vllm:openai/gpt-oss-120b", "name": "GPT-OSS 120B via vLLM", "family": "GPT-OSS", "tag": "local-large", "size": "server model", "pullable": True},
-        {"id": "vllm:Qwen/Qwen3-VL-4B-Instruct", "name": "Qwen3-VL 4B via vLLM", "family": "Qwen3-VL", "tag": "local-vlm", "size": "server model", "pullable": True},
-        {"id": "vllm:Qwen/Qwen3-VL-8B-Instruct", "name": "Qwen3-VL 8B via vLLM", "family": "Qwen3-VL", "tag": "local-vlm", "size": "server model", "pullable": True},
-        {"id": "vllm:Qwen/Qwen3-VL-30B-A3B-Instruct", "name": "Qwen3-VL 30B A3B via vLLM", "family": "Qwen3-VL", "tag": "local-vlm", "size": "server model", "pullable": True},
-        {"id": "vllm:Qwen/Qwen2.5-VL-7B-Instruct", "name": "Qwen2.5-VL 7B via vLLM", "family": "Qwen2.5-VL", "tag": "local-vlm", "size": "server model", "pullable": True},
-        {"id": "vllm:google/gemma-2-2b", "name": "Gemma 2 2B Base via vLLM", "family": "Gemma", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "vllm:google/gemma-2-2b-it", "name": "Gemma 2 2B via vLLM", "family": "Gemma", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "vllm:google/gemma-2-9b", "name": "Gemma 2 9B Base via vLLM", "family": "Gemma", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "vllm:google/gemma-2-9b-it", "name": "Gemma 2 9B via vLLM", "family": "Gemma", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "vllm:google/gemma-3-4b-it", "name": "Gemma 3 4B via vLLM", "family": "Gemma", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "vllm:google/gemma-3-12b-it", "name": "Gemma 3 12B via vLLM", "family": "Gemma", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "vllm:microsoft/Phi-3.5-mini-instruct", "name": "Phi 3.5 Mini via vLLM", "family": "Phi", "tag": "local-coding", "size": "server model", "pullable": True},
-        {"id": "vllm:microsoft/Phi-4-mini-instruct", "name": "Phi 4 Mini via vLLM", "family": "Phi", "tag": "local-coding", "size": "server model", "pullable": True},
-        {"id": "vllm:microsoft/phi-4", "name": "Phi 4 via vLLM", "family": "Phi", "tag": "local-coding", "size": "server model", "pullable": True},
-        {"id": "vllm:mistralai/Mistral-7B-Instruct-v0.3", "name": "Mistral 7B via vLLM", "family": "Mistral", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "vllm:mistralai/Ministral-8B-Instruct-2410", "name": "Ministral 8B via vLLM", "family": "Mistral", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "vllm:mistralai/Mistral-Small-24B-Instruct-2501", "name": "Mistral Small 24B via vLLM", "family": "Mistral", "tag": "local-large", "size": "server model", "pullable": True},
-        {"id": "vllm:meta-llama/Llama-3.2-3B-Instruct", "name": "Llama 3.2 3B via vLLM", "family": "Llama 3.x", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "vllm:meta-llama/Llama-3.1-8B-Instruct", "name": "Llama 3.1 8B via vLLM", "family": "Llama 3.1", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "vllm:meta-llama/Llama-3.3-70B-Instruct", "name": "Llama 3.3 70B via vLLM", "family": "Llama 3.x", "tag": "local-large", "size": "server model", "pullable": True},
-        {"id": "vllm:meta-llama/Llama-3.1-70B-Instruct", "name": "Llama 3.1 70B via vLLM", "family": "Llama 3.1", "tag": "local-server", "size": "server model", "pullable": True},
+        _model("vllm:Qwen/Qwen3-VL-4B-Instruct", "Qwen3-VL 4B via vLLM", "Qwen3-VL", "local-vlm", "실행 도구에서 관리", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("vllm:Qwen/Qwen3-VL-8B-Instruct", "Qwen3-VL 8B via vLLM", "Qwen3-VL", "local-vlm", "실행 도구에서 관리", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("vllm:Qwen/Qwen3-VL-30B-A3B-Instruct", "Qwen3-VL 30B A3B via vLLM", "Qwen3-VL", "local-vlm", "실행 도구에서 관리", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("vllm:google/gemma-4-12b-it", "Gemma 4 12B via vLLM", "Gemma 4", "local-vlm", "실행 도구에서 관리", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("vllm:suitch/gemma-4-31B-it-4bit", "Gemma 4 31B via vLLM", "Gemma 4", "local-vlm", "실행 도구에서 관리", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("vllm:meta-llama/Llama-4-Scout-17B-16E-Instruct", "Llama 4 Scout via vLLM", "Llama 4", "local-vlm", "실행 도구에서 관리", source_country="미국", source_company="Meta", execution_method=_RUNS_ON_THIS_COMPUTER),
     ],
     "lmstudio": [
-        {"id": "lmstudio:openai/gpt-oss-20b", "name": "GPT-OSS 20B via LM Studio", "family": "GPT-OSS", "tag": "local-reasoning", "size": "server model", "pullable": True},
-        {"id": "lmstudio:openai/gpt-oss-120b", "name": "GPT-OSS 120B via LM Studio", "family": "GPT-OSS", "tag": "local-large", "size": "server model", "pullable": True},
-        {"id": "lmstudio:ggml-org/gemma-4-31B-it-GGUF", "name": "Gemma 4 31B 4-bit via LM Studio", "family": "Gemma 4", "tag": "local-vlm", "size": "server model", "pullable": True},
-        {"id": "lmstudio:Qwen/Qwen3-VL-4B-Instruct", "name": "Qwen3-VL 4B via LM Studio", "family": "Qwen3-VL", "tag": "local-vlm", "size": "server model", "pullable": True},
-        {"id": "lmstudio:Qwen/Qwen3-VL-8B-Instruct", "name": "Qwen3-VL 8B via LM Studio", "family": "Qwen3-VL", "tag": "local-vlm", "size": "server model", "pullable": True},
-        {"id": "lmstudio:Qwen/Qwen3-VL-30B-A3B-Instruct", "name": "Qwen3-VL 30B A3B via LM Studio", "family": "Qwen3-VL", "tag": "local-vlm", "size": "server model", "pullable": True},
-        {"id": "lmstudio:Qwen/Qwen2.5-VL-7B-Instruct", "name": "Qwen2.5-VL 7B via LM Studio", "family": "Qwen2.5-VL", "tag": "local-vlm", "size": "server model", "pullable": True},
-        {"id": "lmstudio:google/gemma-2-2b-it", "name": "Gemma 2 2B via LM Studio", "family": "Gemma", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "lmstudio:google/gemma-2-9b-it", "name": "Gemma 2 9B via LM Studio", "family": "Gemma", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "lmstudio:google/gemma-3-4b-it", "name": "Gemma 3 4B via LM Studio", "family": "Gemma", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "lmstudio:google/gemma-3-12b-it", "name": "Gemma 3 12B via LM Studio", "family": "Gemma", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "lmstudio:microsoft/Phi-3.5-mini-instruct", "name": "Phi 3.5 Mini via LM Studio", "family": "Phi", "tag": "local-coding", "size": "server model", "pullable": True},
-        {"id": "lmstudio:microsoft/Phi-4-mini-instruct", "name": "Phi 4 Mini via LM Studio", "family": "Phi", "tag": "local-coding", "size": "server model", "pullable": True},
-        {"id": "lmstudio:microsoft/phi-4", "name": "Phi 4 via LM Studio", "family": "Phi", "tag": "local-coding", "size": "server model", "pullable": True},
-        {"id": "lmstudio:mistralai/Mistral-7B-Instruct-v0.3", "name": "Mistral 7B via LM Studio", "family": "Mistral", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "lmstudio:mistralai/Ministral-8B-Instruct-2410", "name": "Ministral 8B via LM Studio", "family": "Mistral", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "lmstudio:mistralai/Mistral-Small-24B-Instruct-2501", "name": "Mistral Small 24B via LM Studio", "family": "Mistral", "tag": "local-large", "size": "server model", "pullable": True},
-        {"id": "lmstudio:meta-llama/Llama-3.2-3B-Instruct", "name": "Llama 3.2 3B via LM Studio", "family": "Llama 3.x", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "lmstudio:meta-llama/Llama-3.1-8B-Instruct", "name": "Llama 3.1 8B via LM Studio", "family": "Llama 3.1", "tag": "local-server", "size": "server model", "pullable": True},
-        {"id": "lmstudio:meta-llama/Llama-3.3-70B-Instruct", "name": "Llama 3.3 70B via LM Studio", "family": "Llama 3.x", "tag": "local-large", "size": "server model", "pullable": True},
-        {"id": "lmstudio:meta-llama/Llama-3.1-70B-Instruct", "name": "Llama 3.1 70B via LM Studio", "family": "Llama 3.1", "tag": "local-server", "size": "server model", "pullable": True},
+        _model("lmstudio:Qwen/Qwen3-VL-4B-Instruct", "Qwen3-VL 4B via LM Studio", "Qwen3-VL", "local-vlm", "실행 도구에서 관리", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("lmstudio:Qwen/Qwen3-VL-8B-Instruct", "Qwen3-VL 8B via LM Studio", "Qwen3-VL", "local-vlm", "실행 도구에서 관리", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("lmstudio:Qwen/Qwen3-VL-30B-A3B-Instruct", "Qwen3-VL 30B A3B via LM Studio", "Qwen3-VL", "local-vlm", "실행 도구에서 관리", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("lmstudio:ggml-org/gemma-4-12B-it-GGUF", "Gemma 4 12B 4-bit via LM Studio", "Gemma 4", "local-vlm", "실행 도구에서 관리", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("lmstudio:ggml-org/gemma-4-31B-it-GGUF", "Gemma 4 31B 4-bit via LM Studio", "Gemma 4", "local-vlm", "실행 도구에서 관리", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("lmstudio:meta-llama/Llama-4-Scout-17B-16E-Instruct", "Llama 4 Scout via LM Studio", "Llama 4", "local-vlm", "실행 도구에서 관리", source_country="미국", source_company="Meta", execution_method=_RUNS_ON_THIS_COMPUTER),
     ],
     "llamacpp": [
-        {"id": "llamacpp:ggml-org/gpt-oss-20b-GGUF", "name": "GPT-OSS 20B GGUF via llama.cpp", "family": "GPT-OSS", "tag": "gguf-q4", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:ggml-org/gpt-oss-120b-GGUF", "name": "GPT-OSS 120B GGUF via llama.cpp", "family": "GPT-OSS", "tag": "gguf-q4", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:ggml-org/gemma-4-31B-it-GGUF", "name": "Gemma 4 31B GGUF via llama.cpp", "family": "Gemma 4", "tag": "gguf-q4", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:Qwen/Qwen3-VL-4B-Instruct-GGUF", "name": "Qwen3-VL 4B GGUF via llama.cpp", "family": "Qwen3-VL", "tag": "gguf-vlm", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:Qwen/Qwen3-VL-8B-Instruct-GGUF", "name": "Qwen3-VL 8B GGUF via llama.cpp", "family": "Qwen3-VL", "tag": "gguf-vlm", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:unsloth/gemma-2-2b-it-GGUF", "name": "Gemma 2 2B GGUF via llama.cpp", "family": "Gemma", "tag": "gguf-q4", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:unsloth/gemma-2-9b-it-GGUF", "name": "Gemma 2 9B GGUF via llama.cpp", "family": "Gemma", "tag": "gguf-q4", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:unsloth/gemma-3-4b-it-GGUF", "name": "Gemma 3 4B GGUF via llama.cpp", "family": "Gemma", "tag": "gguf-q4", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:bartowski/Mistral-7B-Instruct-v0.3-GGUF", "name": "Mistral 7B GGUF via llama.cpp", "family": "Mistral", "tag": "gguf-q4", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:bartowski/Phi-3.5-mini-instruct-GGUF", "name": "Phi 3.5 Mini GGUF via llama.cpp", "family": "Phi", "tag": "gguf-q4", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:bartowski/phi-4-GGUF", "name": "Phi 4 GGUF via llama.cpp", "family": "Phi", "tag": "gguf-q4", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:bartowski/Llama-3.2-3B-Instruct-GGUF", "name": "Llama 3.2 3B GGUF via llama.cpp", "family": "Llama 3.x", "tag": "gguf-q4", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:bartowski/Llama-3.1-8B-Instruct-GGUF", "name": "Llama 3.1 8B GGUF via llama.cpp", "family": "Llama 3.1", "tag": "local-server", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:bartowski/Llama-3.3-70B-Instruct-GGUF", "name": "Llama 3.3 70B GGUF via llama.cpp", "family": "Llama 3.x", "tag": "local-large", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:bartowski/Llama-3.1-70B-Instruct-GGUF", "name": "Llama 3.1 70B GGUF via llama.cpp", "family": "Llama 3.1", "tag": "local-server", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:unsloth/DeepSeek-R1-GGUF", "name": "DeepSeek-R1 GGUF via llama.cpp", "family": "DeepSeek", "tag": "gguf-q4", "size": "gguf", "pullable": True},
-        {"id": "llamacpp:bartowski/DeepSeek-Coder-V2-Lite-Instruct-GGUF", "name": "DeepSeek-Coder-V2 Lite GGUF via llama.cpp", "family": "DeepSeek", "tag": "gguf-q4", "size": "gguf", "pullable": True},
+        _model("llamacpp:Qwen/Qwen3-VL-4B-Instruct-GGUF", "Qwen3-VL 4B GGUF via llama.cpp", "Qwen3-VL", "gguf-vlm", "gguf", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("llamacpp:Qwen/Qwen3-VL-8B-Instruct-GGUF", "Qwen3-VL 8B GGUF via llama.cpp", "Qwen3-VL", "gguf-vlm", "gguf", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("llamacpp:Qwen/Qwen3-VL-30B-A3B-Instruct-GGUF", "Qwen3-VL 30B GGUF via llama.cpp", "Qwen3-VL", "gguf-vlm", "gguf", source_country="중국", source_company="Alibaba", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("llamacpp:ggml-org/gemma-4-12B-it-GGUF", "Gemma 4 12B GGUF via llama.cpp", "Gemma 4", "gguf-vlm", "gguf", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("llamacpp:ggml-org/gemma-4-31B-it-GGUF", "Gemma 4 31B GGUF via llama.cpp", "Gemma 4", "gguf-vlm", "gguf", source_country="미국", source_company="Google", execution_method=_RUNS_ON_THIS_COMPUTER),
+        _model("llamacpp:ggml-org/Llama-4-Scout-17B-16E-Instruct-GGUF", "Llama 4 Scout GGUF via llama.cpp", "Llama 4", "gguf-vlm", "gguf", source_country="미국", source_company="Meta", execution_method=_RUNS_ON_THIS_COMPUTER),
     ],
 }
 
 MODEL_ENGINE_ALIASES = {
-    "gpt-oss-20b": {
-        "local_mlx": "mlx-community/gpt-oss-20b-MXFP4-Q8",
-        "ollama": "gpt-oss:20b",
-        "vllm": "openai/gpt-oss-20b",
-        "lmstudio": "openai/gpt-oss-20b",
-        "llamacpp": "ggml-org/gpt-oss-20b-GGUF",
-    },
-    "openai/gpt-oss-20b": {
-        "local_mlx": "mlx-community/gpt-oss-20b-MXFP4-Q8",
-        "ollama": "gpt-oss:20b",
-        "vllm": "openai/gpt-oss-20b",
-        "lmstudio": "openai/gpt-oss-20b",
-        "llamacpp": "ggml-org/gpt-oss-20b-GGUF",
-    },
-    "gpt-oss-120b": {
-        "local_mlx": "mlx-community/gpt-oss-120b-MXFP4-Q4",
-        "ollama": "gpt-oss:120b",
-        "vllm": "openai/gpt-oss-120b",
-        "lmstudio": "openai/gpt-oss-120b",
-        "llamacpp": "ggml-org/gpt-oss-120b-GGUF",
-    },
-    "openai/gpt-oss-120b": {
-        "local_mlx": "mlx-community/gpt-oss-120b-MXFP4-Q4",
-        "ollama": "gpt-oss:120b",
-        "vllm": "openai/gpt-oss-120b",
-        "lmstudio": "openai/gpt-oss-120b",
-        "llamacpp": "ggml-org/gpt-oss-120b-GGUF",
+    "gemma-4-12b-it-4bit": {
+        "local_mlx": "mlx-community/gemma-4-12b-it-4bit",
+        "ollama": "hf.co/ggml-org/gemma-4-12B-it-GGUF:Q4_K_M",
+        "vllm": "google/gemma-4-12b-it",
+        "lmstudio": "ggml-org/gemma-4-12B-it-GGUF",
+        "llamacpp": "ggml-org/gemma-4-12B-it-GGUF",
+    },
+    "mlx-community/gemma-4-12b-it-4bit": {
+        "local_mlx": "mlx-community/gemma-4-12b-it-4bit",
+        "ollama": "hf.co/ggml-org/gemma-4-12B-it-GGUF:Q4_K_M",
+        "vllm": "google/gemma-4-12b-it",
+        "lmstudio": "ggml-org/gemma-4-12B-it-GGUF",
+        "llamacpp": "ggml-org/gemma-4-12B-it-GGUF",
     },
     "gemma-4-31b-it-4bit": {
         "local_mlx": "mlx-community/gemma-4-31b-it-4bit",
@@ -247,13 +186,26 @@ MODEL_ENGINE_ALIASES = {
         "lmstudio": "ggml-org/gemma-4-31B-it-GGUF",
         "llamacpp": "ggml-org/gemma-4-31B-it-GGUF",
     },
+    "qwen3-vl-8b": {
+        "local_mlx": "mlx-community/Qwen3-VL-8B-Instruct-4bit",
+        "ollama": "qwen3-vl:8b",
+        "vllm": "Qwen/Qwen3-VL-8B-Instruct",
+        "lmstudio": "Qwen/Qwen3-VL-8B-Instruct",
+        "llamacpp": "Qwen/Qwen3-VL-8B-Instruct-GGUF",
+    },
+    "llama-4-scout": {
+        "local_mlx": "mlx-community/Llama-4-Scout-17B-16E-Instruct-4bit",
+        "ollama": "hf.co/ggml-org/Llama-4-Scout-17B-16E-Instruct-GGUF:Q4_K_M",
+        "vllm": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
+        "lmstudio": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
+        "llamacpp": "ggml-org/Llama-4-Scout-17B-16E-Instruct-GGUF",
+    },
 }
 
 _VERSIONED_MODEL_PATTERNS = (
     ("gemma", re.compile(r"\bgemma[-\s]?(\d+(?:\.\d+)?)", re.IGNORECASE)),
     ("qwen", re.compile(r"\bqwen[-\s]?(\d+(?:\.\d+)?)", re.IGNORECASE)),
     ("llama", re.compile(r"\bllama[-\s]?(\d+(?:\.\d+)?)", re.IGNORECASE)),
-    ("phi", re.compile(r"\bphi[-\s]?(\d+(?:\.\d+)?)", re.IGNORECASE)),
 )
 
 

package/latticeai/services/model_recommendation.py

@@ -3,7 +3,8 @@
 Given a detected system profile (from :func:`auto_setup.probe`) this module
 classifies every model in :data:`model_catalog.ENGINE_MODEL_CATALOG` into one of
 three states — **recommended**, **compatible**, or **not_recommended** — and
-groups the result by model family (Gemma, Qwen, Llama, Phi, DeepSeek, …).
+groups the result by current multimodal model family (Gemma 4, Qwen3-VL,
+Llama 4).
 
 It is intentionally pure and dependency-light: the only input is a plain dict
 describing the machine, so it is fully unit-testable without touching real
@@ -28,12 +29,11 @@ NOT_RECOMMENDED = "not_recommended"
 # Apple-Silicon only.  Used to decide platform availability before sizing.
 _APPLE_ONLY_ENGINES = {"local_mlx"}
 
-# Family display order for the grouped view (best/newest first within a brand).
+# Family display order for the grouped view (newest multimodal generations first).
 _FAMILY_ORDER = [
-    "Gemma 4", "Gemma 3", "Gemma 2", "Gemma",
-    "Qwen3-VL", "Qwen2.5-VL", "Qwen2.5", "Qwen",
-    "Llama 3.x", "Llama 3.1", "Llama",
-    "Mistral", "Phi", "GPT-OSS", "DeepSeek", "SmolLM",
+    "Gemma 4",
+    "Qwen3-VL",
+    "Llama 4",
 ]
 
 _SIZE_RE = re.compile(r"([\d.]+)\s*(TB|GB|MB)", re.IGNORECASE)
@@ -44,7 +44,7 @@ def parse_size_gb(size: Any) -> Optional[float]:
     """Parse a catalog ``size`` string (``"4.7GB"``, ``"963MB"``, ``"40GB+"``).
 
     Returns ``None`` when the size is non-numeric (e.g. ``"pull required"`` or
-    ``"server model"``) so callers can treat it as "size unknown".
+    ``"실행 도구에서 관리"``) so callers can treat it as "size unknown".
     """
     if not isinstance(size, str):
         return None
@@ -92,30 +92,38 @@ def _classify_one(
     need_gb = estimated_ram_gb(size_gb) if size_gb is not None else None
 
     if not engine_available:
-        status, reason = NOT_RECOMMENDED, "Requires Apple Silicon (MLX runtime)"
+        status, reason = NOT_RECOMMENDED, "Apple Silicon과 MLX-VLM이 필요합니다"
     elif need_gb is None:
-        # Server/pull models have no fixed on-disk size — treat as compatible
-        # (the engine streams/pulls weights on demand).
-        status, reason = COMPATIBLE, "Served/pulled on demand by the engine"
+        # Tool-managed/pull models have no fixed on-disk size, so treat them as
+        # compatible and let the execution tool validate the exact model.
+        status, reason = COMPATIBLE, "선택한 실행 방식에서 필요할 때 모델을 받습니다"
     elif ram_gb <= 0:
-        status, reason = COMPATIBLE, "Memory unknown — verify before loading"
-    elif need_gb <= ram_gb * 0.6:
-        status, reason = RECOMMENDED, f"Fits comfortably (~{need_gb:.0f} GB of {ram_gb:.0f} GB RAM)"
+        status, reason = COMPATIBLE, "메모리 정보를 확인하지 못했습니다. 불러오기 전에 검증합니다"
+    elif need_gb <= ram_gb * 0.75:
+        status, reason = RECOMMENDED, f"현재 메모리에서 안정적으로 사용할 가능성이 높습니다 (~{need_gb:.0f} GB / {ram_gb:.0f} GB)"
     elif need_gb <= ram_gb * 0.9:
-        status, reason = COMPATIBLE, f"Runs but tight (~{need_gb:.0f} GB of {ram_gb:.0f} GB RAM)"
+        status, reason = COMPATIBLE, f"사용 가능하지만 여유가 적습니다 (~{need_gb:.0f} GB / {ram_gb:.0f} GB)"
     else:
-        status, reason = NOT_RECOMMENDED, f"Needs ~{need_gb:.0f} GB RAM (have {ram_gb:.0f} GB)"
+        status, reason = NOT_RECOMMENDED, f"권장 메모리가 부족합니다 (~{need_gb:.0f} GB 필요, 현재 {ram_gb:.0f} GB)"
 
     return {
         "id": model.get("id"),
         "name": model.get("name"),
+        "model_name": model.get("model_name") or model.get("name"),
         "family": model.get("family"),
         "tag": model.get("tag"),
+        "modality": model.get("modality") or "multimodal",
         "size": model.get("size"),
         "size_gb": size_gb,
         "required_ram_gb": need_gb,
         "status": status,
         "reason": reason,
+        "source_country": model.get("source_country"),
+        "source_company": model.get("source_company"),
+        "execution_method": model.get("execution_method"),
+        "run_location": model.get("run_location"),
+        "internet_requirement": model.get("internet_requirement"),
+        "source_display_order": model.get("source_display_order"),
     }
 
 
@@ -132,7 +140,10 @@ def recommend_catalog(profile: Dict[str, Any], *, engine: str = "local_mlx") ->
     ``profile`` is a dict shaped like ``auto_setup.SystemProfile.to_json()``
     (``os``, ``arch``, ``ram_mb``, ``gpu={vendor,vram_mb}`` …).
     """
-    models = ENGINE_MODEL_CATALOG.get(engine, [])
+    models = [
+        model for model in ENGINE_MODEL_CATALOG.get(engine, [])
+        if str(model.get("modality") or "").lower() == "multimodal"
+    ]
     engine_available = _engine_available(engine, profile)
     ram_gb = _ram_gb(profile)
 

package/latticeai/services/model_runtime.py

@@ -64,7 +64,7 @@ INVITE_GATE_ENABLED = False
 ALLOW_PLAINTEXT_API_KEYS = False
 CORS_ALLOW_NETWORK = False
 PUBLIC_MODEL = "openai:gpt-4o-mini"
-LOCAL_MODEL = "mlx-community/SmolLM-1.7B-Instruct-4bit"
+LOCAL_MODEL = "mlx-community/gemma-4-12b-it-4bit"
 IS_PUBLIC_MODE = False
 keyring = None
 
@@ -889,7 +889,7 @@ def ensure_llamacpp_server(model_name: str) -> None:
 
 def engine_installed(engine: str) -> bool:
     if engine == "local_mlx":
-        return bool(importlib.util.find_spec("mlx") and importlib.util.find_spec("mlx_lm"))
+        return bool(importlib.util.find_spec("mlx") and importlib.util.find_spec("mlx_vlm"))
     if engine == "ollama":
         return local_binary("ollama") is not None
     if engine == "vllm":