MultiPresetManager

The MultiPresetManager class manages multiple model presets.

Description

MultiPresetManager stores presets in a name-keyed dict and supports default-preset selection, single/parallel preset testing, and generation of PresetReport results.

Methods

• set_default_preset(preset: ModelPreset | str) -> None: Set the default preset (auto-adds if not registered)
• get_default_preset() -> ModelPreset: Get the default preset; falls back to a random preset with a warning if none set
• get_preset(name: str) -> ModelPreset: Get a preset by name; raises ValueError if not found
• add_preset(preset: ModelPreset) -> None: Add a preset; raises ValueError if the name already exists
• get_all_presets() -> list[ModelPreset]: List all presets
• async test_single_preset(preset: ModelPreset | str) -> PresetReport: Test a single preset through its protocol adapter
• async test_presets() -> AsyncGenerator[PresetReport]: Test all presets sequentially, yielding one report each

Example

from amrita_core.preset import MultiPresetManager, ModelPreset

manager = MultiPresetManager()
preset = ModelPreset(name="gpt", base_url="https://api.example.com", api_key="...")
manager.add_preset(preset)
manager.set_default_preset("gpt")

report = await manager.test_single_preset("gpt")
print(report.status, report.time_used)