Support model configuration for custom endpoints (context window, temperature, model parameters)

Summary

Custom model endpoints (configured via GUI) have no way to specify model metadata or parameters. There is no setting for context window size, temperature, or model-specific parameters in settings.toml, and the GUI exposes none of these either.

Problem

Warp supports connecting to custom model endpoints, but the configuration is GUI-only and exposes essentially zero model settings. This means:

• Context window size is unknown to Warp. There is no way to tell Warp how much context a custom model supports. Warp's internal context management cannot make informed decisions about truncation, summarization, or conversation length.
• No temperature control. There is no setting for temperature anywhere in the settings schema or GUI endpoint configuration.
• No model-specific parameters. Settings like chat_template_kwargs, max_output_tokens, or other provider-specific parameters are completely unsupported.
• GUI-only configuration. Custom endpoint settings are not represented in settings.toml at all, making them invisible to programmatic configuration, version control, or team sharing.

Desired workflow

1. Configure a custom endpoint with a URL and API key (this exists today).
2. Define model metadata alongside it: context window size, max output tokens, temperature, and any provider-specific parameters.
3. Warp uses the context window size for accurate context usage display and truncation behavior.
4. All settings are available in settings.toml for version control and programmatic configuration.

Warp version

v0.2026.05.27.15.44.stable_01

Operating system

Windows 11