How to use eroq with SillyTavern — uncensored API in five minutes

SillyTavern is the power tool of character chat — cards, lorebooks, group chats, total control — and it's bring-your-own-API by design. That makes it a perfect match for eroq: an OpenAI-compatible endpoint whose engines are tuned for exactly what SillyTavern users do all day, with no refusal wall in the middle of your scenes.

Here's the whole setup, start to finish.

1. Get a key

Create an account — it starts with 50 free credits, no card — and mint a key on the API keys page. The key is shown once; keep it somewhere safe.

2. Point SillyTavern at eroq

In SillyTavern:

1. Open API Connections (the plug icon).
2. API: select Chat Completion.
3. Chat Completion Source: select Custom (OpenAI-compatible).
4. Custom Endpoint (Base URL): https://eroq.ai/v1
5. API Key: paste your eroq_sk_… key.
6. Model: enter eroq-rp-plus (or eroq-rp-mini — see below).
7. Hit Connect — the status dot goes green, and you're running on eroq.

That's the entire integration. Streaming works out of the box (standard SSE), so replies render token by token like any other source.

3. Pick your model like you mean it

• eroq-rp-plus (3 credits per reply) — the flagship: persona hold across long scenes, repetition damping tuned at RP temperatures, the prose texture your best cards deserve.
• eroq-rp-mini (1 credit per reply) — same persona handling, sized for speed and volume. The right default for fast back-and-forth and group chats where replies fly.

Both are uncensored within the written policy: dark, violent and intense fiction renders in character — no jailbreak prompt required, no sudden Thursday refusals. Prompts outside the hard limits return an explicit error instead of a lecture, and are never charged. And as eroq's own models, neither retains anything upstream.

4. Settings that actually matter

• Temperature: RP+ is tuned around 0.8 — start there rather than importing exotic presets. If you drag any model far from its tuning point, that's where loops begin.
• Max tokens: eroq caps completions at 1,200; a max_tokens around 400–500 fits most card styles and keeps replies snappy.
• Context size: credits are flat per reply — a huge context costs you nothing extra, but a curated one roleplays better. SillyTavern's summarize extension plus a reasonable context beats raw size.
• System prompt: your cards' system blocks replace eroq's neutral preamble entirely — full control, as SillyTavern intends.

5. What the flat pricing means for a SillyTavern habit

No token math, ever: a reply is 1 or 3 credits whether your card has a 200-token or 2,000-token context. A heavy evening — say 150 replies on RP mini — is 150 credits, about $1.50 at pack rates. The pricing page has the full table, and every response carries usage.credits_remaining so you always know where you stand. When the balance runs out, calls fail fast with a clear error and nothing is charged — top up and continue exactly where the scene stopped.

Troubleshooting

• 401 — key pasted with a space, or revoked. Mint a fresh one; it's instant.
• 402 — balance empty. Top up; the scene resumes where it stopped.
• A refusal? Lawful fiction — horror, crime and violence included — is what these engines are chosen for, so the prompt armor mainstream sources need is wasted here. The hard limits — minors, real people, illegal content — are enforced at the account level rather than as a per-message error code on chat; the explicit content_blocked error belongs to the image and video endpoints, where a provider can refuse a render.

FAQ

Does SillyTavern work with eroq?

Yes, with no extension or proxy. Choose Chat Completion as the API, Custom (OpenAI-compatible) as the source, https://eroq.ai/v1 as the base URL, and your eroq_sk_… key. Streaming works out of the box because the chunk format is standard SSE.

Which model should I select in SillyTavern?

eroq-rp-plus at 3 credits a reply for long scenes and prose texture, eroq-rp-mini at 1 credit for fast back-and-forth and group chats. You can switch between them per chat without touching anything else in the connection — it is one string in the model field.

Do I need a jailbreak prompt?

No. Roleplay renders in character within the written acceptable-use policy, dark and intense scenes included, so the prompt armor that mainstream sources need is wasted tokens here. Prompts crossing the hard limits — minors, real people, illegal content — return content_blocked instead of a lecture, and are not charged.

How far do the 50 free credits go?

50 replies on RP mini, or 16 on RP+ — enough to run your own cards through the scenes that usually break elsewhere. Credits never expire, and a heavy evening of 150 mini replies afterwards is about $1.50 at pack rates.

Five minutes, three strings, and your cards run on an engine that was tuned for them. If something in the setup fights you, support answers within a business day.