Dashboard Models

The Models page controls where Anyy sends model work. It covers the main conversational model, ordered text fallbacks, and dedicated image, video, text-to-speech, and speech-to-text routes.

Open Models

Start Dashboard and select Models:

anyy dashboard

The page groups routes into Text, Image, Video, and Voice. Each card shows its current provider/model and whether the route is primary, enabled, disabled, or unavailable.

Configure the Main Model

Open Main model. The form exposes only the choices supported by the selected provider contract:

1. Choose a provider type.
2. Choose the service and protocol when more than one is available.
3. Choose an authentication method.
4. Enter an API key or complete the subscription login when required.
5. Keep the default base URL, or enter a trusted compatible endpoint.
6. Refresh the model catalog or switch to manual model entry.
7. Choose the model and reasoning defaults.
8. Select Test connection to check catalog access and model presence, then Save.
9. Open a new Chat session, confirm its selected provider/model, send a short prompt such as “Reply with Ready,” and check that a real reply returns. Temporarily disable configured fallback routes for this check and restore their previous settings afterward, or verify that the actual producing provider/model is the tested target and no fallback was used. Selection plus a successful reply alone is insufficient.

The configuration preview shows the non-secret settings that will be written. Saved API keys are not displayed in full.

Check discovery and generation separately

After changing provider, endpoint, credential, or model, Test connection uses the pending settings or saved credential to refresh/check the model catalog. It does not send a conversation-generation request. A connected status confirms that catalog check; use the new-session prompt above to verify generation.

API keys and subscription login

Provider-specific authentication appears in the same detail page. Depending on the provider, you can:

• enter or replace an API key;
• clear a stored API key;
• start or repeat an OAuth/subscription login; or
• log out of a connected subscription profile.

Authentication choices are bound to the selected endpoint. Changing service or protocol can change the credential realm and model catalog. Codex uses subscription device login; the Anthropic Claude Code subscription option accepts an externally obtained setup token, which does not establish generic OAuth support. Catalog/login success still requires a separate real model-request test.

See Provider Protocols and Model Interfaces for the five current conversation families and additional native protocols that are not implemented.

Add Fallback Models

On the Text tab, select Add fallback. Configure the provider/model, check its catalog connection, and save it. Then select that target explicitly in a new conversation and send a short prompt, using the same disabled-fallback or confirmed-no-fallback check as for the main model. Restore any settings you temporarily changed.

Fallback cards are evaluated in their saved order. Each entry can be enabled or disabled and can set:

• whether a route may drop reasoning history it cannot replay;
• whether it may use a smaller context window; and
• the maximum attempts for that fallback, including its first attempt.

Remove a fallback from its detail page when it should no longer be eligible.

Fallback is availability routing

Fallback configuration does not merge answers or ask every model. It gives the runtime an ordered alternative when error policy and target compatibility permit. Tool, media, reasoning-history, or context requirements can skip an entry. Session auto-retry and provider SDK retries are separate controls.

Image and Video Routes

Image understanding and video understanding can use the main model automatically when it declares the required input capability. Otherwise, choose a dedicated provider/model.

For understanding routes, the detail page also supports:

• enabling or disabling the feature;
• auto or summary_first routing;
• a capability prompt;
• token/character limits;
• timeout; and
• thinking behavior.

summary_first requires a dedicated route. Test manually entered models: an unknown model ID is not assumed to support image or video input.

Image generation is a separate output route with its own provider/model, aspect ratio, timeout, and byte limit. The product tool currently requires OpenAI Responses; an ordinary OpenAI-compatible Chat endpoint alone is insufficient. Codemode’s openrouter-images model interface is configured separately.

Voice Routes

The Voice tab contains:

• Text to speech (TTS) — provider, model, protocol, voice, output behavior, and an audio preview; and
• Speech to text (STT) — provider, model, protocol, default language, and connection test.

A working text-chat provider does not automatically support speech endpoints. TTS uses audio-speech or chat-audio-output; STT uses audio-transcriptions or chat-audio-input. Enter the API root and match the protocol to the model. Enable a route after testing it; see Voice & Speech.

When Changes Take Effect

Dashboard saves through the running Gateway. New sessions use the saved main model. An existing Chat session can retain its own selected provider/model and reasoning state until you change it in that session or start a new one.

For a terminal-only configuration path, see Configuring Models.

Troubleshooting

Problem	Fix
Model list is empty	Check provider authentication and base URL, then select Refresh models
Test connection fails	Verify endpoint, credential, model ID, and provider account access
Main model shows unavailable	Check the catalog connection and saved route, then confirm generation in a new Chat session
Connection is shown as connected but Chat fails	Catalog access succeeded; inspect the actual generation error, model access, endpoint, and credential
Image/video route asks for a dedicated model	The main model does not declare that input capability, or summary_first is selected
A saved model is absent in Chat	Start a new session or switch the current session's model explicitly
Subscription login is stale	Use re-authenticate/login again, or log out and connect the intended profile