eroq api · text
Expressive AI,
one API.
The uncensored roleplay engine behind millions of conversations, in an OpenAI-shaped chat API. Persona context, streaming, vision — and scenes that render instead of hitting a refusal wall.
Private and uncensored · a written policy, not a filter
注册即送 40 积分 · 无需绑卡
curl https://eroq.ai/v1/chat/completions \
-H "Authorization: Bearer $EROQ_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "eroq-rp-plus",
"context": "Mira: sardonic starship mechanic, dry humor, hates small talk. Scene: engine bay, mid-shift.",
"messages": [
{
"role": "user",
"content": "The reactor is making that noise again."
}
]
}'模型
3 个模型,价格固定。
eroq RP+
chateroq-rp-plus
我们在正式环境中使用的角色扮演模型:长场景中始终保持角色设定,采样经过调校以减少重复和废话,支持 SSE 流式输出。我们面向大众的网络中数百万场对话背后,正是这个模型。
3 积分 / 次回复
eroq RP mini
chateroq-rp-mini
面向高用量产品的经济档。API 结构和人设处理与 eroq RP+ 相同,针对延迟和成本做了精简。
1 积分 / 次回复
Claude Fable 5.2
chatclaude-fable-5-2
Anthropic 最新的 Fable,也是它向所有人开放的最强模型:每次回答前先思考,然后写作、讲解、规划,并推理解决难题。可在 Studio 的文本工具和 API 中使用,按每次回复统一定价,思考过程也包含在内。由其提供方在上游进行内容审核。
20 积分 / 次回复
你能得到什么
Text,为真正上线的产品而打造。
Stays in character
eroq RP+ is tuned for persona work: repetition damping, filler suppression, and sampling that survives 200-message scenes. A context field carries the character sheet; a messaging mode answers like texting when your product is DMs, not novels.
Uncensored by design
Fiction goes where your users take it. Dark and intense roleplay renders instead of refusing — the wall is a written acceptable-use policy (no minors, no real people, ever), not a filter's mood.
Streaming first
Server-sent events with the standard chunk format. First tokens reach your UI while the model is still writing, and the shape is the one your existing client already parses.
It can see
Send an image alongside the messages and the model reacts to it in character — the missing half of a companion product, where users send pictures and expect to be seen.
Flat per reply
One credit on eroq RP mini, three on eroq RP+, whatever the length of the reply or the size of the context you send. Per-user costs stay predictable as scenes get long, which is exactly when token pricing stops being.
Two tiers, one shape
eroq RP+ for the turns that carry a scene, eroq RP mini for volume. Identical API and persona handling, so routing between them is a model string and nothing else.
常见问题
长话短说。
Is the chat API OpenAI-compatible? +
The request and response shapes follow the OpenAI convention, so most client code ports by changing the base URL, the key and the model id. We are not affiliated with OpenAI.
What does a reply cost? +
One credit on eroq RP mini and three on eroq RP+, flat. The price does not move with reply length or context size.
Will the model refuse my users mid-scene? +
The engines are chosen so intense scenes render rather than hitting a refusal wall. A request outside the acceptable-use policy returns an explicit content_blocked error instead of a lecture, and is never charged.
Are there embeddings or tool calling? +
No. There is no embeddings endpoint and no tool-calling surface. Keep a mainstream API for retrieval and agents, and route the character work here.
Is there a context window limit? +
Long scenes are what these models are tuned for, but they are not infinite. Summarise older turns into the context field rather than resending an entire history, which also keeps latency flat.
