"""LM Studio provider profile (local desktop app with an OpenAI-compatible server). Request quirk: reasoning goes out as top-level ``reasoning_effort``, gated on and clamped to the per-model ``allowed_options`` LM Studio publishes in ``/api/v1/models``. The agent probes and caches those options and passes them as ``lmstudio_reasoning_options``. Everything else stays where it lives today: endpoint, ``LM_API_KEY`` / ``LM_BASE_URL`` and the no-auth placeholder (``hermes_cli/auth.py`` registry row), the chat-model picker probe (``hermes_cli/models_local.py``). Hence ``base_url`` is left empty: a loopback ``base_url`` would map every ``127.0.0.1`` / ``localhost`` endpoint to LM Studio in URL->provider inference and disable local-server detection for Ollama, llama.cpp and vLLM. """ from typing import Any from agent.lmstudio_reasoning import resolve_lmstudio_effort from agent.reasoning_effort import generic_nested_reasoning from providers import register_provider from providers.base import ProviderProfile class LMStudioProfile(ProviderProfile): """LM Studio: top-level ``reasoning_effort`` from the model's published options.""" def build_api_kwargs_extras( self, *, reasoning_config: dict | None = None, supports_reasoning: bool = False, lmstudio_reasoning_options: list[str] | None = None, **context: Any, ) -> tuple[dict[str, Any], dict[str, Any]]: """``supports_reasoning`` is true only when the model publishes a non-``off`` option; ``lmstudio_reasoning_options`` is the agent's cached probe of them. Auxiliary calls have no probe (``None``) and keep the generic nested ``reasoning`` they always sent.""" if lmstudio_reasoning_options is None: return generic_nested_reasoning(reasoning_config), {} if not supports_reasoning: return {}, {} effort = resolve_lmstudio_effort(reasoning_config, lmstudio_reasoning_options) return {}, ({"reasoning_effort": effort} if effort is not None else {}) register_provider(LMStudioProfile( name="lmstudio", display_name="LM Studio", supports_model_listing=False, supports_health_check=False, env_vars=("LM_API_KEY", "LM_BASE_URL"), ))