> ## Documentation Index
> Fetch the complete documentation index at: https://rimelabs-docs-coda-websocket-reference.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Remove Arcana model IDs

> Arcana is retired from the Rime cloud API. Replace Arcana model IDs with Coda and verify each voice and language combination.

Arcana retired from the Rime cloud API on August 15, 2026. Cloud requests that set `modelId` to `arcana`, `arcanav2`, or `arcanav3` are now served by [Coda](/docs/models#coda). Replace those identifiers with `coda` so your configuration names the model that serves the request.

<Warning>**Check every voice before changing production traffic.** Coda has a different voice catalog. A request can return audio even when the `speaker`, `modelId`, and `lang` combination is not listed, so a successful response does not prove that the pairing is supported.</Warning>

## Find every Arcana identifier

Search your code, configuration files, environment variables, and deployed environments for all three retired identifiers:

```text theme={null}
arcana
arcanav2
arcanav3
```

Requests that omit `modelId` are served by Mist v3. They do not use Coda automatically.

No response header names the model that served a request. Record the environments and call sites you inspect because the response cannot confirm that you removed every Arcana identifier.

## Check the replacement voice and language

Search the [Coda catalog](/docs/voices-coda) for each speaker you use.

* If the same speaker exists on Coda, keep the name and test it with your production text. Coda and Arcana were trained separately, so the audio will differ.
* If the speaker is absent, choose a replacement in the same language. Compare candidates on domain terms, numbers, and long utterances rather than a generic greeting.
* Set `lang` to the language listed for the Coda voice. Each Coda voice serves one language.

Coda serves English, Arabic, French, German, Hindi, Italian, Japanese, Portuguese, and Spanish through the cloud API. Word timestamps are available for English and Spanish only.

## Update and verify the request

Change `modelId` to `coda`. The endpoint, authentication, transports, and remaining request fields do not change.

```json theme={null}
{
  "speaker": "astra",
  "text": "Thanks for calling. How can I help?",
  "modelId": "coda",
  "lang": "eng"
}
```

Verify every environment, language, and transport you use. HTTP, binary WebSockets, and JSON WebSockets are separate request paths, so one successful request does not cover the others.

## Replace unsupported on-prem images

Existing Arcana on-prem images remain available to pull, but Rime no longer updates or supports them. This includes security patches and text-normalization fixes. Use a current Coda image for supported on-prem deployments, and contact [help@rime.ai](mailto:help@rime.ai) to plan the replacement.

The Rime models hosted by Together AI use Together's lifecycle and endpoints. Follow the [Together AI integration guide](/docs/together-ai) if you run Arcana there.

## Related

* [Coda voices](/docs/voices-coda): find a supported speaker and language.
* [Models](/docs/models#coda): review Coda's features and deployment options.
* [On-prem quickstart](/docs/on-prem/quickstart): deploy the current Coda image.
* [Speed](/docs/speed): use `timeScaleFactor` for Coda speed control.