{
  "schema_version": 1,
  "family": "sherpa_kws",
  "display_name": "sherpa-onnx Zipformer KWS",
  "description": "Streaming Zipformer transducer keyword spotting from sherpa-onnx model exports.",
  "category": "audio_tools",
  "status": "experimental",
  "tasks": [
    "wake_word"
  ],
  "modes": [
    "offline",
    "streaming"
  ],
  "languages": [
    "model_dependent"
  ],
  "capabilities": {
    "wake_word": [
      "speech_segments",
      "stream_events"
    ]
  },
  "runtime": {
    "tags": [
      "gguf",
      "stream"
    ]
  },
  "options": {
    "request": [
      {
        "name": "keywords",
        "type": "string",
        "description": "Newline- or slash-separated tokenized keyword rows; each row may include :score, #threshold, and @label fields.",
        "required": false
      },
      {
        "name": "max_active_paths",
        "type": "int",
        "description": "Maximum keyword decoder hypotheses.",
        "required": false,
        "default": 4,
        "min": 1,
        "max": 64
      },
      {
        "name": "keywords_score",
        "type": "float",
        "description": "Default context score added for each matched keyword token. Negative values are accepted but token scores are clamped to at least zero.",
        "required": false,
        "default": 1
      },
      {
        "name": "keywords_threshold",
        "type": "float",
        "description": "Default mean acoustic probability required to trigger a keyword.",
        "required": false,
        "default": 0.25,
        "min": 0,
        "max": 1
      },
      {
        "name": "num_trailing_blanks",
        "type": "int",
        "description": "Emit after strictly more than this many trailing blank frames; 1 requires 2 blanks, 0 requires 1.",
        "required": false,
        "default": 1,
        "min": 0
      }
    ],
    "session": [],
    "load": []
  },
  "package_defaults": {
    "download": {
      "kind": "huggingface_snapshot",
      "repo": "audio-cpp/audio.cpp-gguf",
      "revision": "main",
      "gated": false
    }
  },
  "packages": [
    {
      "id": "sherpa_kws_zipformer_zh_en_3m_f32",
      "display_name": "Sherpa KWS Zipformer Chinese-English 3M GGUF F32",
      "default": true,
      "format": "gguf",
      "precision": "f32",
      "target_directory": "Sherpa-KWS-GGUF",
      "files": [
        "Sherpa-KWS-GGUF/sherpa-kws-zipformer-zh-en-3m-f32.gguf"
      ],
      "strip_prefix": "Sherpa-KWS-GGUF"
    }
  ],
  "dependencies": [],
  "ui": {
    "tags": [
      "Wake Word",
      "GGUF",
      "Stream"
    ],
    "docs": [
      "docs/models/sherpa_kws.md"
    ],
    "recommended_package": "sherpa_kws_zipformer_zh_en_3m_f32"
  },
  "sources": [
    {
      "format": "gguf",
      "roots": {
        "model": ".",
        "weights": "$gguf"
      },
      "files": {
        "config": "model:config.json",
        "tokens": "model:tokens.txt",
        "keywords": "model:keywords.txt"
      },
      "tensors": {
        "weights": {
          "source": "weights:",
          "prefix": "weights"
        }
      }
    },
    {
      "format": "safetensors",
      "roots": {
        "model": "."
      },
      "files": {
        "config": "model:config.json",
        "tokens": "model:tokens.txt",
        "keywords": "model:keywords.txt"
      },
      "tensors": {
        "weights": "model:model.safetensors"
      }
    }
  ]
}