nlpmed_engine.utils package

Submodules

nlpmed_engine.utils.utils module

class nlpmed_engine.utils.utils.ModelSpec

Bases: TypedDict

device: str

max_length: int

model_path: str

tokenizer_path: str

nlpmed_engine.utils.utils.build_initial_config() → dict[str, Any]

Build the full pipeline config dict, including ml_inference models block. The first model name in API_ML_MODEL_NAMES will be treated as the default at runtime.

Returns:

dict[str, Any]: The configuration dictionary for the engine.

nlpmed_engine.utils.utils.get_effective_param(instance_value: Any, provided_value: Any, *, required: bool = True) → Any

nlpmed_engine.utils.utils.read_models_from_env() → dict[str, ModelSpec]

Read model configurations from environment variables.

Raises:

RuntimeError – If a configured model is missing its model or tokenizer path.

Returns:

Model specifications keyed by configured model name.

Return type:

dict[str, ModelSpec]

Example return structure:

{
    "modelA": {
        "device": "cpu",
        "model_path": "...",
        "tokenizer_path": "...",
        "max_length": 512,
    },
    "modelB": {
        "device": "cpu",
        "model_path": "...",
        "tokenizer_path": "...",
        "max_length": 1024,
    },
}

Module contents