Add OpenRouter, Ollama Cloud, and custom OpenAI-compatible endpoints #41

Goal

Cover OpenAI-compatible hosted services without embedding or operating a gateway.

Depends on

• Define a provider-neutral AI client and capability contract #35, Move AI execution and credentials into a secure Electron main-process service #36, Add provider profiles and connection controls to Settings #37, and Migrate LM Studio to the provider abstraction without behavior regressions #38.

Scope

• Add presets for OpenRouter and Ollama Cloud.
• Add a custom OpenAI-compatible profile usable with a user's LiteLLM deployment or another compatible endpoint.
• Support bearer authentication first.
• Use the endpoint-security rules from Move AI execution and credentials into a secure Electron main-process service #36.
• Detect and report capability gaps instead of assuming full OpenAI parity.

Acceptance criteria

• Presets provide sensible endpoints while remaining editable where appropriate.
• Custom endpoints can fetch models when supported and allow manual model entry otherwise.
• Redirects, HTTP/HTTPS policy, timeouts, and sanitized errors are tested.
• Lucid does not host inference, embed LiteLLM, or automatically route/fail over between providers.