Connect JanitorAI to Unrestricted AI

Set up JanitorAI

1. Open API Settings

   Open any chat and click the API settings button at the top of the chat.

2. Choose “Proxy”

3. Click “+ Add” under Proxy Configurations

   Give it a Configuration Name, for example Unrestricted AI.

4. Type the model ID into “Model Name”

   Copy it exactly, including dashes.

   venice-uncensored-1-2

5. Paste the full URL into “Proxy URL”

   JanitorAI needs the complete endpoint ending in /v1/chat/completions, not the base URL.

   https://tryunrestrictedai.com/v1/chat/completions

6. Paste your key into “API Key”

   Get your key

7. Click “Apply” and make sure the configuration is Active

   If the old settings still answer, save again and refresh the page.

Recommended settings

• Output limit: 4,096 tokens. Context window: 32,768 tokens.
• In Generation Settings, set a max new tokens value at or below the output limit above.
• If a message fails, open API Settings again and check that this configuration, not an older one, is marked Active.
• Text chat with streaming is supported. Image input, tool calling and multiple completions per request (n > 1) are not.

Troubleshooting

Your app shows the HTTP status and usually our error code. Match it here.

401invalid_api_key

The key is missing, mistyped or revoked. Paste the whole key (it starts with ua_) without “Bearer”, or create a new key in the console.

402insufficient_credit · key_budget_exceeded · plan_limit

No usage left. Choose a plan or add credit in Billing, create a key with a higher spending limit, or lower the max response length.

429rate_limit

More than 30 requests a minute or 3 at once. Turn off multiple swipes, auto-continue and parallel generations, then wait a minute.

503inference_unavailable

The model is not accepting requests right now, or the model ID does not match a live model. Check the model ID and the status page, then retry.

400 / 413unknown_model · invalid_request · context_limit

Wrong model ID, a response length above the output limit, or a context that is too long. Copy the model ID again and lower context or response size.