{
  "_note": "离线模型档策略 (2026-09-07). 判断在代码, 模型只问答/转述/初筛; 换模型改这里 + ollama pull, 不改结论.",
  "runtime": "ollama", "endpoint": "http://127.0.0.1:11434",
  "tiers": {
    "default_qa":  {"model": "qwen3:8b",        "vram_gb": 6,  "note": "默认问答, think=false; 校闸不过自动升档一次"},
    "escalate_qa": {"model": "qwen3:32b",       "vram_gb": 22, "note": "复杂题升档（★2026-09-22 用户人工测试 7 实逮: 原写 qwen3.8:27b, 本机没这个 tag ⇒ 升档必失败、ask 最终 error; 已改为本机实装的 qwen3:32b）"},
    "review":      {"model": "deepseek-r1:14b", "vram_gb": 10, "note": "交叉审核初筛票 (非独立审级; 确诊/¥/上云仍需云端跨厂商双审)"},
    "embed":       {"model": "bge-m3",          "vram_gb": 2,  "note": "检索向量"}
  },
  "profiles": {
    "gpu_24gb_plus": ["default_qa", "escalate_qa", "review", "embed"],
    "gpu_12_16gb":   ["default_qa", "review", "embed"],
    "cpu_only":      ["default_qa", "embed"]
  },
  "pull_commands": ["ollama pull qwen3:8b", "ollama pull qwen3:32b", "ollama pull deepseek-r1:14b", "ollama pull bge-m3"]
}