modelis-openai

Use Modelis in Aider, Cline, Continue, or any OpenAI SDK — one base_url, one key for GPT / Claude / Gemini, zero migration.

Modelis is an OpenAI-compatible gateway. With a Modelis account key, point your tool straight at it:

Setting	Value
Base URL	https://modelishub.com/v1
API key	your Modelis key (sk-…)
Model	modelis-auto (the standard tier)

modelis-auto auto-routes each request. It is the standard tier, the same router as modelis-auto:standard. Public auto ids use the colon tiers below. Call a named model by its own id (for example claude-sonnet-4-6). Do not prefix a model name with modelis-auto: and do not append :high or :low.

Model	Per call, within the limit	Input / output limit
modelis-auto:basic	$0.008	4k / 1k tokens
modelis-auto and modelis-auto:standard	$0.04	8k / 2k tokens
modelis-auto:premium	$0.10	4k / 1k tokens

Above those limits the request is billed per token. Named models (for example claude-sonnet-4-6 or deepseek-v4-flash) are billed per token at the upstream price. Top-ups add a 15% service fee ($10 or ¥72 becomes $8.70 of credit). About 95 models are on the gateway. New accounts get $1 of free credit. In the console Wallet, pay by card through Gumroad, or with Alipay/WeChat through an Afdian redemption code. Packs and the live model list: https://modelishub.com/pricing.

RapidAPI sells a separate key (X-RapidAPI-Key; the Basic plan is 100 requests a month). Coding tools send Authorization: Bearer. This package is the single-file bridge for that key: a zero-dependency local proxy that speaks the OpenAI wire protocol on 127.0.0.1, rewrites the auth, and forwards to Modelis. A Modelis account key calls https://modelishub.com/v1 directly.

your tool ──OpenAI(Bearer)──▶ modelis-openai (localhost) ──X-RapidAPI-Key──▶ Modelis on RapidAPI

---

Quickstart

Modelis account (direct): sign up at https://modelishub.com — new accounts get $1 of free credit — create a token, and use the table above. No proxy.

RapidAPI key: subscribe on the pricing page (Basic plan, 100 requests a month), then copy your key: 👉 https://modelishub.com/pricing

Run with Node (18+):

npx modelis-openai
# or:  node modelis-openai.mjs

Run with Docker (no Node needed):

git clone https://github.com/modelishub/modelis-openai
cd modelis-openai
docker build -t modelis-openai .
docker run --rm -p 8787:8787 modelis-openai

Then point your tool at the proxy:

Setting	Value
Base URL	http://127.0.0.1:8787/v1
API key	your RapidAPI key
Model	modelis-auto

The proxy rewrites the request model to modelis-auto (standard tier: $0.04 within 8k input / 2k output tokens, then per token). Set MODELIS_MODEL to another tier, or to "" to pass the tool's model through.

Verify the proxy is alive:

curl http://127.0.0.1:8787/health

---

Tool recipes

Direct gateway (Modelis account key)

from openai import OpenAI
client = OpenAI(base_url="https://modelishub.com/v1", api_key="sk-YOUR-TOKEN")
print(client.chat.completions.create(
    model="modelis-auto",
    messages=[{"role": "user", "content": "Hello"}],
).choices[0].message.content)

import OpenAI from "openai";
const client = new OpenAI({ baseURL: "https://modelishub.com/v1", apiKey: "sk-YOUR-TOKEN" });
const r = await client.chat.completions.create({
  model: "modelis-auto:basic", // or modelis-auto, modelis-auto:premium, or a named model id
  messages: [{ role: "user", content: "Hello" }],
});
console.log(r.choices[0].message.content);

Aider (RapidAPI key, via this proxy)

export OPENAI_API_BASE=http://127.0.0.1:8787/v1
export OPENAI_API_KEY=<your-rapidapi-key>
aider --model openai/modelis-auto

Cline / Roo Code (VS Code)

• API Provider: OpenAI Compatible
• Base URL: http://127.0.0.1:8787/v1 for a RapidAPI key, or https://modelishub.com/v1 for a Modelis account key
• API Key: your RapidAPI key, or your Modelis sk-… token
• Model ID: modelis-auto

Continue (~/.continue/config.yaml)

models:
  - name: Modelis
    provider: openai
    model: modelis-auto
    apiBase: http://127.0.0.1:8787/v1
    apiKey: <your-rapidapi-key>

For a Modelis account key, set apiBase to https://modelishub.com/v1 and apiKey to your sk-… token. Other tiers are modelis-auto:basic and modelis-auto:premium. A named model is its own id, such as deepseek-v4-flash.

Any OpenAI SDK (RapidAPI key, via this proxy)

from openai import OpenAI
client = OpenAI(base_url="http://127.0.0.1:8787/v1", api_key="<your-rapidapi-key>")
print(client.chat.completions.create(
    model="modelis-auto",
    messages=[{"role": "user", "content": "Hello"}],
).choices[0].message.content)

import OpenAI from "openai";
const client = new OpenAI({ baseURL: "http://127.0.0.1:8787/v1", apiKey: "<your-rapidapi-key>" });
const r = await client.chat.completions.create({
  model: "modelis-auto",
  messages: [{ role: "user", content: "Hello" }],
});
console.log(r.choices[0].message.content);

Cursor: Cursor sends requests from its own servers, so a localhost endpoint won't reach this proxy. Use a Modelis account key with base_url https://modelishub.com/v1, or expose the proxy on a public URL you control. The local setup above is for tools that call the API from your machine (Aider, Cline, Continue, scripts).

---

How it works

Streaming (stream: true) is piped straight through, so token-by-token output in your tool works exactly as with the OpenAI API. The proxy:

1. reads the key from Authorization: Bearer <key> (or MODELIS_RAPIDAPI_KEY),
2. rewrites the request model to modelis-auto (the standard tier; configurable),
3. forwards to the RapidAPI gateway with X-RapidAPI-Key / X-RapidAPI-Host,
4. streams the response back unchanged.

It also answers GET /v1/models and GET /health so tools that probe on startup don't error. GET /v1/models lists modelis-auto, the id this proxy forwards by default. The full auto-router set exported from the module is modelis-auto, modelis-auto:basic, modelis-auto:standard, and modelis-auto:premium.

---

Configuration

All optional — sensible defaults work out of the box.

Env var	Default	Purpose
MODELIS_PORT	8787	Local listen port
MODELIS_HOST	127.0.0.1	Local bind address (0.0.0.0 in Docker)
MODELIS_RAPIDAPI_KEY	(unset)	Fallback key if your tool can't send one
MODELIS_MODEL	modelis-auto	Model sent upstream (standard tier). Set modelis-auto:basic or modelis-auto:premium for another tier, or "" to pass the tool's model through
MODELIS_RAPIDAPI_HOST	modelis-auto-chat.p.rapidapi.com	RapidAPI gateway host (from your Code Snippet)
MODELIS_UPSTREAM_PATH	/v1/chat/completions	Upstream endpoint path

---

Privacy & trust

• Runs entirely on your machine; the only outbound connection is to the RapidAPI gateway over HTTPS.
• It does not log, store, or inspect message contents — it only swaps the auth header and forwards bytes.
• One file, no third-party dependencies. Read it.

Develop

node --test     # unit + in-process integration tests, zero deps

License

MIT