EN ▾

Janitor AI API settings with an OpenAI-compatible endpoint

Uncensored AI: Debunking Common Myths

Updated

To configure Janitor AI with your own endpoint, you only need three pieces of data: the URL, the API key, and the model name. In this guide to janitor ai api settings, you will see where each goes, how to test it beforehand, and what to do if it fails.

Key points

  • You need the URL https://api.apisinrestricciones.com/v1, your API key and the uncensored model.
  • Test the API key with curl before pasting it into Janitor AI: this separates API key errors from configuration errors.
  • A session of 200 messages with 4,000 context tokens costs about $0.15.
  • If you get 429 errors, reduce retries: the test key supports 2 concurrent requests.
Scenario with masks and dialogue bubbles connected to a settings panel

What you need before launching Janitor AI

Janitor AI allows you to connect an external service instead of using only its built-in model, as long as that service uses the OpenAI format. This is what this API offers. Gather these three pieces of data before entering the settings screen:

  • Base URL: https://api.apisinrestricciones.com/v1
  • Key: you obtain it on the keys page, after signing in with Google or with email and password.
  • Model name: uncensored

The exact text of Janitor AI menus changes with web versions, so I describe them generally here. If something does not match what you see, check the official Janitor AI help. The service is text-only and intended for fiction, roleplay, and companion apps for adults. It does not accept sexual content involving minors in any case.

Test the key with curl first

The most common error is tapping ten settings without knowing if the key works. With one command you know:

curl https://api.apisinrestricciones.com/v1/chat/completions \
  -H "Authorization: Bearer TU_CLAVE" \
  -H "Content-Type: application/json" \
  -d '{"model": "uncensored", "messages": [{"role": "user", "content": "Hola"}], "max_tokens": 30}'

If you receive a JSON with choices, the key and model are correct. Any subsequent error will be in how you wrote the data in Janitor AI, not in the API. If the curl fails, review the key and make sure the URL starts with https://.

The settings in Janitor AI, one by one

Inside a chat, open the API settings menu. Depending on the version, you will see a list of providers where you must choose the proxy option or your own OpenAI-compatible endpoint, instead of the platform's built-in model. The following fields will appear:

FieldWhat to enter
Model nameuncensored
Proxy URL / endpointhttps://api.apisinrestricciones.com/v1/chat/completions
API keyYour API key, without spaces or line breaks

Regarding the URL: depending on the version, the field asks for the base URL or the full endpoint. First try the full path ending in /chat/completions; if the chat does not respond, try only the base ending in /v1. This is the point that causes the most confusion, and the Janitor AI help page is the definitive source for the format it expects.

After saving the settings, reload the chat page. Some versions do not apply the change until you reload.

Generation settings worth adjusting

Janitor AI exposes generation controls that are sent to the API. These matter:

  • Temperature. Between 0.7 and 1.0 works well for roleplay and fiction: lower gives more predictable responses, higher adds variety and errors.
  • Max response tokens. Set it to a reasonable value, for example 300 to 600. Without a cap, the model can use much more than you need and you pay for every output token.
  • Context size. The model window is 100,000 tokens, between prompt and response. There is no need to go that far: the more context, the more each message costs because the entire history is forwarded. A value of 4,000 to 16,000 is practical for long chats.
  • Penalties. Penalty parameters are passed to the API. If the character repeats phrases, increase the repetition penalty slightly.

For the full list of supported parameters (top_p, stop, seed), see the documentation.

Chat session cost

At $0.15 per million input tokens and $0.50 per million output tokens, the cost depends mostly on context. Two examples for 200 character messages, with 300 output tokens:

Average contextPer message200 messages
4,000 tokens0.0006 + 0.00015 = $0.00075$0.15
16,000 tokens0.0024 + 0.00015 = $0.00255$0.51

Accounts: 4,000 × 0.15 / 1,000,000 = $0.0006 and 300 × 0.50 / 1,000,000 = $0.00015. With 16,000 context, 16,000 × 0.15 / 1,000,000 = $0.0024. The $0.50 free trial credit covers about 196 messages in the second case (0.50 / 0.00255). If you run out of balance, top up from $20 in crypto (USDT on TRC20 or USDC on Base); see prices and bonuses.

Common errors and how to fix them

  • Chat does not respond and there is no message. Open browser developer tools (F12 key), go to the network tab, and check the request response. You will see the HTTP code and error body there.
  • Authentication error. The key was pasted incorrectly or has a trailing space. Copy it again.
  • URL error or 404. The path does not match what the field expects. Toggle between the full path and the base URL, as explained above.
  • 429 with concurrency. There are more simultaneous requests than allowed: 2 with a trial key, 8 with a full key. Avoid pressing regenerate multiple times in a row, because a streaming response occupies its slot until it finishes.
  • 429 with rate_limited. More than 60 requests per minute with the trial key (300 with the full key). Wait a minute.
  • 429 with Retry-After: 5. Temporary overload. Retry after 5 seconds.
  • Model not found. Check that you typed uncensored in lowercase.

Failed requests are not charged. If none of these options fix the issue, contact Support with the value of the X-Request-Id header from the response.

More guides

Frequently asked questions

What URL do I put in Janitor AI?

Try with https://api.apisinrestricciones.com/v1/chat/completions and, if it does not respond, with the base https://api.apisinrestricciones.com/v1. The exact format depends on the Janitor AI version.

What model name do I use?

uncensored, in lowercase.

Do I need to pay to try it?

No. After signing in, you can claim $0.50 in free trial credit with one click, valid for 7 days and without a payment method.

Are my chats used to train the model?

No, prompts are not used for training.

Your key is one step away

Create an account, copy the key and change the base URL. That's it.

Get API key
Get API key