Private AI documentation

Quickstart

Make your first call in three steps.

1. Create a free account and verify your email.
2. Make a sandbox key in API, then Keys. It is shown once; keep it in an environment variable.
3. Run your first call against the base URL below. https://api.aideamaker.com/v1

curl https://api.aideamaker.com/v1/chat/completions \
  -H "Authorization: Bearer $AIDEAMAKER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "Laguna-S-2.1",
       "messages": [{"role": "user", "content": "Say hello"}]}'

from openai import OpenAI
import os

client = OpenAI(base_url="https://api.aideamaker.com/v1", api_key=os.environ["AIDEAMAKER_API_KEY"])
reply = client.chat.completions.create(
    model="Laguna-S-2.1",
    messages=[{"role": "user", "content": "Say hello"}],
)
print(reply.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://api.aideamaker.com/v1", apiKey: process.env.AIDEAMAKER_API_KEY });
const reply = await client.chat.completions.create({
  model: "Laguna-S-2.1",
  messages: [{ role: "user", content: "Say hello" }],
});
console.log(reply.choices[0].message.content);

Switch from the OpenAI SDK

Change the base URL and the key. Nothing else changes.

from openai import OpenAI
import os

client = OpenAI(
    base_url="https://api.aideamaker.com/v1",
    api_key=os.environ["AIDEAMAKER_API_KEY"],
)
# nothing else changes

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.aideamaker.com/v1",
  apiKey: process.env.AIDEAMAKER_API_KEY,
});
// nothing else changes

Call the API from your own server. It does not accept requests from web pages (no cross-origin access), and a key placed in a web page is a key anyone can copy.

Authentication

Send your key as a Bearer token in the Authorization header. Any other way of sending it is refused with a 401.

Authorization: Bearer YOUR_KEY

A key starts with sk-am- and is shown once, when you create it. Keep it secret. You can replace or revoke it any time in API, then Keys. A sandbox key expires after 30 days and can be renewed.

Models

Put the id shown for a model in the table below in the model field of each request. GET /v1/models lists the models your key may use. More ids (aliases) for these models may be added later; the ids below keep working.

curl https://api.aideamaker.com/v1/models \
  -H "Authorization: Bearer $AIDEAMAKER_API_KEY"

for model in client.models.list():
    print(model.id)

for await (const model of client.models.list()) {
  console.log(model.id);
}

A model that another customer has rented is not available to your key while the rental lasts. The Models page shows until when, in CST.

Chat completions

POST /v1/chat/completions takes the standard Chat Completions request and returns the standard response, including the usage counts.

Parameters that work

Output is limited by your key's cap and by the model's cap. A request for more is cut to the cap without an error. A call that is not streamed ends after 50 seconds and is capped lower than a streamed one, so use streaming for anything long.

Qwen3.8 Flash Next reasons before it answers, and that reasoning counts toward the completion tokens in the usage counts.

Accepted and ignored

These fields do not cause an error and do not change the answer: seed, user, presence_penalty, frequency_penalty, and the name field of a message.

Not supported yet

The call still succeeds, so your code will not see an error, but these have no effect. Check your own requests against this list.

These endpoints are not available and answer 404 in the OpenAI error format: /v1/completions, /v1/embeddings, /v1/responses, /v1/files, the image and audio endpoints, and GET /v1/models/{id}.

Streaming

Set stream to true to receive server-sent events: data chunks whose first delta carries the role, a final chunk with the finish reason, then data: [DONE]. Add stream_options with include_usage to get the token counts in a last chunk.

stream = client.chat.completions.create(
    model="Laguna-S-2.1",
    messages=[{"role": "user", "content": "Write one sentence about autumn."}],
    stream=True,
)
for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="")

const stream = await client.chat.completions.create({
  model: "Laguna-S-2.1",
  messages: [{ role: "user", content: "Write one sentence about autumn." }],
  stream: true,
});
for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}

curl -N https://api.aideamaker.com/v1/chat/completions \
  -H "Authorization: Bearer $AIDEAMAKER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "Laguna-S-2.1", "stream": true,
       "messages": [{"role": "user", "content": "Write one sentence about autumn."}]}'

While a stream waits, the service sends a comment line (starting with a colon) every 15 seconds to keep the connection open; the OpenAI SDKs ignore it. A stream can run up to 9 minutes. An error inside a stream arrives as a data chunk holding the error, and the stream then ends without [DONE].

Errors

Errors use the OpenAI shape: an error object with message, type, param and code. Every response carries an x-request-id header; quote it if you contact us.

{"error": {"message": "Invalid API key.", "type": "invalid_request_error", "param": null, "code": "invalid_api_key"}}

Treat any 5xx answer, including a plain web-server page if the service is restarting, as a reason to retry after a pause.

Rate limits and tiers

Limits apply to your whole account, not to each key. A free account has the sandbox limits; paid tiers upgrade automatically as your total credit purchases reach each level.

Past a limit the call is refused with a 429 and a Retry-After header: wait that long and send it again. The OpenAI SDKs do this for you.

Billing and credit

Per-token use is paid from prepaid credit, bought from $5 and used oldest first. Credit expires 12 months after the purchase, and when it runs out, calls stop with a clear error.?Prepaid credit stops being usable this long after the purchase. Credit from the oldest purchase is used first.

The usage counts in each response are the tokens we count for you. A small overrun after your balance reaches zero is taken from your next purchase. The sandbox is never billed.

Pricing

Rentals

A rental gives you exclusive use of one model for a 30-day prepaid term after approval, and your rental key works until the term ends.

How rentals work

Data handling

Your prompts and answers pass through our API gateway without being written to our databases, logs or backups. We keep usage records only (which key, which model, token counts, timing, result, and the calling address and software name) to bill, limit and run the service, for up to 400 days.

Calls travel over HTTPS, which encrypts them in transit over the internet.

Do not send health records, payment details or other regulated data unless we have agreed it in writing first.

Software built with us that uses Private AI keeps only what its own page says it keeps.

Privacy Policy

FAQ