Model and engine selection

Choose a backend engine, pick a model, use live model lists, and save favorite model choices for new sessions.

Happier uses engine in the app UI for the backend choice that will run a session.

An engine can be:

a built-in provider backend, such as Claude, Codex, Gemini, OpenCode, or Pi
a configured Custom ACP backend
a backend variant exposed by a provider integration

The Model section inside an engine controls which provider model should run the session.

Where to find it

Open the new session flow and use the engine picker in the bottom bar.

The left side of the picker shows engine choices. The right side shows the selected engine's details, including:

model choices
provider or ACP configuration options
refresh status when Happier is probing the selected machine
model-specific controls, when the provider exposes them

Running sessions can also show a model/config picker in the input settings for the current session. Favorites are part of the new session engine picker.

Inside a running session, that same picker can also change the Agent itself — see Continue with another Agent.

Choose an engine

Pick the engine that should launch the session.

The list is capability-aware:

disabled providers or hidden backends do not appear
configured ACP backends appear as normal engine choices
profiles can filter which engines are compatible
engine details can vary by machine, working directory, profile, and connected-service state

If only one engine is available, Happier can skip most of the picker chrome and use that engine directly.

Choose a model

Most model pickers start with Use CLI settings.

Use this when you want Happier to avoid setting a model override and let the provider CLI/runtime decide from its own settings.

When you choose a specific model, Happier stores that selection for the session launch. The exact apply behavior depends on the provider and runtime. Some providers can change models on the next prompt or live through ACP; others only apply the model when a new provider process starts.

For running-session apply timing, see Permissions.

Live model lists

Some engines expose models dynamically. Happier probes the selected machine/backend and shows the models that backend currently reports.

This matters because the available list can depend on:

the installed provider CLI version
whether the provider is logged in
account or subscription entitlements
selected connected-service profile
Custom ACP backend args and environment
the selected working directory, when the provider uses project-local config

Use the refresh button in the model section after changing provider login state, CLI config, Custom ACP settings, or working directory.

Static model lists

Not every provider has a dynamic model API.

For static-only providers, Happier uses its built-in provider catalog instead. Claude is the main example: Claude models can still be selected and favorited even though the model list is static rather than dynamically probed.

Static model lists can still change when Happier updates its provider catalog.

Custom model ids

Some engines support entering a custom model id.

Use this for providers or profiles where you know the exact model id but it is not shown in the list.

Important details:

Happier sends the id to the provider; the provider still decides whether it is valid.
Custom ids are useful for provider-compatible gateways, custom endpoints, or newly released models.
A custom id is not automatically a favorite candidate. Favorites use models Happier can verify from the current live list or static catalog.

Model-specific options

Some models expose extra controls, such as thinking, reasoning, effort, or a provider-specific mode.

When those controls exist, Happier shows them inside the selected model card. They are part of the selected engine's configuration surface, not a separate global settings page.

Not every provider exposes these options, and not every model under the same provider exposes the same options.

Favorite models

Favorite models help you keep a short list of model choices you use often.

To favorite a model:

Open the new-session engine picker.
Choose an engine.
Select a concrete model.
Use the star button on the selected model card.

The filled star means the model is already a favorite. The outline star means it can be added.

The Use CLI settings default is not favoritable because it is not a concrete model choice.

Favorites section

After you save at least one favorite model, the engine picker can show a Favorites section in the left rail.

Use it to jump directly to saved engine/model combinations without scanning a long model list.

Selecting a favorite:

switches to the favorite's engine
applies the favorite's model
keeps other launch settings, such as profile, permissions, modes, and config options, governed by the normal new-session controls

Favorites are account settings. They are saved per backend identity and model id, so the same model id can be a separate favorite under two different engines or Custom ACP backends.

Removing favorites

You can remove a favorite by clicking the filled star:

from the normal model list, when the favorited model is visible
from the Favorites section

Already-favorited models keep showing the filled star so you can remove them without selecting them first.

Availability rules

Favorites are saved, but selection still respects the current backend state.

For dynamically probed engines:

a favorite is selectable only when the current probe reports that model as available
if the model disappears because auth changed, the CLI changed, or the provider removed it, Happier does not start a session with that unavailable model
refresh the probe after fixing login/configuration

For static-only engines:

favorites are checked against Happier's static provider catalog
Claude models can be favorited and selected from the catalog without a live model probe

If a saved favorite no longer appears in the current live/static model list, Happier keeps it in the Favorites section with a setup hint so you can remove it. It will not be selectable until the backend reports it again.

Profile and backend compatibility

Favorites respect the current engine list.

If a selected profile filters out an engine, favorites for that incompatible engine are hidden from the Favorites section for that picker state.

This keeps the Favorites section focused on choices that make sense for the current launch configuration.

What favorites do not store

A favorite stores the engine/model pair. It does not replace profiles or backend settings.

Favorites do not store:

API keys or connected-service credentials
environment variables
working directory
permission mode
session mode
model-specific controls such as reasoning effort

Use Profiles for reusable environment, secret, endpoint, and backend-default setup.

Common use cases

Use favorite models when you want to:

keep your normal Claude, Codex, or Gemini model one click away
quickly switch between a fast model and a deeper reasoning model
save favorite models for several Custom ACP backends
avoid scrolling through a long live model list
preserve a favorite while a provider login or entitlement issue is temporarily hiding it