Dream
DocumentationDownload
Configuration and help

Models and defaults

Enable models, choose defaults for chats and Git drafting, and change a conversation's model.

Dream separates the models available to use from the model selected for a particular task. Settings → Providers controls availability. Settings → General, in the Models group, controls defaults. The selector beneath each chat's message box controls that conversation.

Enable models#

  1. Install and authenticate a supported provider CLI.
  2. Open Settings → Providers and enable the provider.
  3. Refresh the provider to detect its CLI and available models.
  4. Enable the models you want to appear in Dream's selectors.

You can enable models from several providers. If a model is missing from a selector, check that it is enabled here first. Refreshing detects availability; it is separate from choosing your default.

Set the default for new chats#

  1. Open Settings → General → Models.
  2. Under Default chat model, select an enabled model.
  3. Choose Effort and Speed if the selected model offers those controls.
  4. Return to a project and start a new chat to use the selection.

The default applies to newly created chats across projects. Existing conversations keep their own model settings. If New chat brings back an empty draft you already opened, check its composer and change the model there if necessary.

Chat defaults and the text generation model for commit and PR drafts have separate controls.

Set a model for commit and PR drafting#

In the same Models group, choose Text generation model and its effort and speed when supported. Dream uses these selections when drafting commit messages and pull-request text.

For example, you can use one model for implementation chats and another for Git summaries. Changing the Git drafting model does not change your chat models. See Git workflows for reviewing and editing generated text before publishing it.

Change a chat's model#

  1. Open the conversation and wait for its current response to finish.
  2. Click the model name beneath the message box.
  3. Select an enabled model, including a model from another provider if needed.
  4. Check the effort and speed controls before sending the next message.

The conversation remains visible in Dream. Changing the model resets its effort and speed selections, so choose them again if you need specific values. A model change also starts a new provider session for subsequent work; it does not change messages already generated.

A chat's composer shows its own model, effort, and speed selections.

Effort and speed#

ControlPurpose
EffortSelects a reasoning-effort level offered by that model
SpeedSelects a supported speed tier, such as Standard or Fast

These controls appear only when the selected model supports them. Available effort levels vary by model. Speed tiers can have different usage implications; use the option descriptions and your provider's usage information when choosing a tier.

Default effort and speed apply to new chats. Changing either in a chat affects that conversation. Changing the speed also resets its provider session for subsequent work.

If a default becomes unavailable#

Disabling a selected default or refreshing a changed model list can cause Dream to select another enabled model. Recheck both default selectors after changing your enabled models. If no models are enabled, enable one before starting a new chat.

For missing controls or detection failures, see Troubleshooting.