Live proxy status

4My Model Router Proxy

Providers

Local audio

OpenAI-compatible entrypoints

Endpoint guide

Query builder and control reference

Route filter Learned route Image key Parameter name Value

Model

Direct URI

Account console

Workspace

Account

Router overview

Client configuration

Base URL

Models

Cloud proxy

Use the Tank router key as the OpenAI API key. A saved model profile supplies its provider account, encrypted key reference, provider model, endpoint, field mapping, and defaults; no query string is required. URI parameters only override that saved profile for one request.

Retain images for inline log previews Sanitized provider-to-proxy and proxy-to-client details are kept for 30 days. Image files are off by default and are stored encrypted only when this is enabled.

Provider playground

Test any saved image, video, audio, or text profile with its exact JSON payload before sending it from a client. The playground uses the same router key and provider route.

Active model IDs

Credentials

Provider accounts

Discovery and routing

Model profiles

Provider account Public model ID Display name Provider model Capability Endpoint path Image key

Provider -> proxy Provider input Client output format Provider output URI equivalents: provider_input=auto|b64|url and provider_output=auto|b64|url. The provider adapter uses its saved schema when set to automatic.

Proxy -> client Client response Match the client return preference to the provider output preference This preference is saved in the model profile; clients only need the model ID. A URI option overrides it for one call. Both formats are sent together by default so clients can display the image and use its inline pixels for follow-up vision.

Cloud proxy execution Execution mode Use the saved profile through /cloud/v1 or /proxy/v1. You can override one request with ?proxy_mode=sync, ?proxy_mode=async, or the equivalent sync=true/async=true.

Query controls JSON Provider parameters JSONModel-specific schema fields belong here. For one-off URI values, use provider.<schema_field>=value; WaveSpeed sync mode stays off unless you explicitly set that provider field. Enabled

Saved profiles

Model IDs exposed through this account's router key.

Learned routes

Successful provider routes discovered by the proxy.

Authentication

Router API key

API key

Live model test

Playground

Choose a saved model, enter what you want to make, and run it. Provider mapping and JSON are only needed for a new or unusual model.

What are you testing? Model

Choose any successful route. The router fills its working model, capability, image field, and known parameters automatically.

Prompt Target size (optional) How should it return? How many versions? Source file for edits, transcription, or video Save this new model once it returns a completed result

Advanced provider setup and JSON

Use this only to add a new model or override a model-specific setting. Query parameters still override saved defaults.

Provider Saved API-key account Request-only provider key Installed providers are listed even before you save an account. A request-only key is never saved; use Workspace → Accounts to save one for reuse. Provider model OpenAI endpoint Provider image field Query controls JSON Provider parameters JSON Advanced OpenAI-compatible JSON payload A new draft uses a saved provider account or a one-time provider key. A completed run can save one reusable profile; it never creates a duplicate each time you run it.

Response

Returned media, with the full response available to copy.

Choose a saved profile or draft a provider route.

Request telemetry

Logs

Log access

Instant updates

Live request progress

Connecting

This view updates while a provider request is queued, processing, completing, or failing. It does not include credentials or raw image data.

Retain images for inline previewsEncrypted output files are kept for 30 days and can be opened from the matching log.

Administration

System control

Accounts

Users

Routing

Provider definitions

Local inference

Audio engines

Hot reload

Provider extensions

Reserved assistance layer

Internal LLM

Base URL means the endpoint of an already-running OpenAI-compatible text model, for example a local service at http://127.0.0.1:1234/v1 or a remote text endpoint. It is not the image provider URL. Image, audio, and video routing is configured through provider accounts and model profiles; this field is only for the optional assistance model.

Security

Audit

API request records are automatically removed after 7 days.