RisuAI custom API setup with an uncensored model

What RisuAI is

RisuAI is an open-source front end for AI roleplay and character chat. Its source is on GitHub at kwaroran/RisuAI, and it runs in the browser at risuai.net as well as in desktop builds. It has no model of its own. It sends each chat to the model you pick in its settings, and one of the options in that list is a custom endpoint that speaks the OpenAI format.

That option has changed names over the years. Older community guides call it "Reverse Proxy". In RisuAI's current source (main branch, read on 24 October 2026) the model entry is named Custom API. The English string for its URL field is still "Reverse Proxy", so you may see both words. They are the same setting.

The RisuAI custom API fields

Choosing Custom API shows four fields under the model picker. These are the values for this API, with the labels from RisuAI's Bot Settings page and its English strings.

URL

https://api.uncensoredllm.ai/v1/chat/completions

The full path, ending in /chat/completions

Key/Password

sk-ullm-live-…

Created in the console, shown once

Request Model

gemma-4-31b-fabled

Any id from the Models page, typed exactly

Format

OpenAI Compatible

The first option in the list

Why the full path: with the full /chat/completions URL, RisuAI sends the request to exactly that address. With Autofill Request URL on, RisuAI also accepts https://api.uncensoredllm.ai/v1 and appends /chat/completions itself, because the URL ends in v1. The full path works in both cases, so it is the one to paste.

Connect RisuAI to the uncensored API

1. Create a key in the console. It starts with sk-ullm- and is shown once, so paste it somewhere safe first. The balance is prepaid from $10; a key with nothing on it is accepted but the first message answers 402.
2. Pick a model id from the uncensored model list or from the table in the next section. Copy it rather than typing it.
3. In RisuAI, open Settings and go to Bot Settings.
4. Set Model to Custom API. The URL, Key/Password, Request Model and Format fields appear.
5. Paste https://api.uncensoredllm.ai/v1/chat/completions into URL.
6. Paste the key into Key/Password. The placeholder says to leave it blank if there is no password; this API needs the key on every request.
7. Type the id into Request Model, for example gemma-4-31b-fabled.
8. Leave Format on OpenAI Compatible. The other options (Anthropic Claude, Mistral, Google Cloud and the OpenAI Response API) expect a different request shape.
9. Open a chat with any character and send a message. A reply means the URL, the key and the model id are all right.

RisuAI also has a second route: a Custom Models list where each entry carries its own URL, format and key, plus an Additional Parameters box. Per RisuAI's help text, that box adds fields to the request body, and a key written as header::Name becomes a header instead. The Bot Settings route above is enough for one model.

Which uncensored build to pick for RisuAI

All four builds below carry the roleplay tag in the catalog, and all four take the OpenAI wire format that the Custom API entry sends. The prices and windows are read from the live catalog when the page renders.

Context windows read from the live catalog on 2026-10-27. The start value is the window capped at 32,768 tokens.

RisuAI's strings label its context setting Max Context Size. It decides how much card, lorebook and history goes out with each message, and every token of it is billed as input. Keep it at or under the model's window. The two Gemma 4 tunes belong to a larger family on the uncensored Gemma page, and the Qwen 3.8 tune to the uncensored Qwen page.

Samplers RisuAI sends with a custom API

RisuAI's model list gives the Custom API entry a fixed set of sampler parameters: temperature, top_p, frequency_penalty, presence_penalty, repetition_penalty, min_p, top_a, top_k and a thinking-token budget. It also marks the entry as streaming, so replies appear as they are written. These are the values we start a roleplay preset from; each one is a starting point to adjust, not a setting the model requires.

A reply that stops mid-sentence is not an error. It means the response hit Max Response Size; raise it and regenerate. The charge is per token either way, so a longer limit only costs more when the model writes more.

What a RisuAI session costs

The worked example: 200 messages in one sitting, each reply about 250 tokens. The short column assumes about 4,000 tokens of card and history per message; the long column assumes 16,000, a lorebook and a long chat.

Computed at render from the catalog prices on 2026-10-27: messages × prompt tokens × input price, plus messages × 250 × output price.

Input dominates: 200 × 4,000 is 800,000 input tokens against 50,000 output tokens. A shorter card, a trimmed lorebook and a lower Max Context Size move the bill more than the choice of model. Every per-token rate is on the pricing page.

Check the key and the model id before RisuAI

Before you open RisuAI, this request lists the model ids your key can reach. It costs nothing and takes a second. Swap in your key and run it from a terminal.

A 200 with a list of ids means the key is good. If the id you put in Request Model is in that list, the only field left to get wrong is the URL. A 401 means the key, and the message says which way:

Both bodies as the API returned them on 24 October 2026.

Troubleshooting RisuAI custom API errors

401: the key

"No API key was sent." means Key/Password is empty. "The API key is not valid." means it is wrong, cut short or revoked. Create a fresh key in the console and paste it whole.

404: the model id or the URL

A 404 with the code model_not_found means the id in Request Model is not in the catalog: a typo, a display name instead of an id, or a build that has since left. Run the request above and copy the id from its output. A 404 without that code means the URL is wrong. The usual cause is a doubled path such as /chat/completions/chat/completions, or a missing /v1. Paste the URL from this guide again.

402: the balance

The prepaid balance is at zero. Top up in the console; a request refused with 402 is not charged.

Browser errors (CORS)

The web build of RisuAI calls the API straight from the browser. The API allows calls from any origin, including the Authorization header (checked on 24 October 2026), so no proxy in between is needed and the key goes only to the API.

The daily real-model check sends a fixed prompt to every model each day, so the model behind an id is the one on its label. Base URL, key and model id for other clients are on the docs page.

Questions