RAMZ / DOCS

API specifications.

Interface design for speech, language and knowledge workflows. Tell us about your application and technical requirements.

01

Create and protect your API key

  1. Sign in to the platform and verify your email.
  2. Open API keys and create a key. Copy the secret when it is displayed; it is shown once.
  3. Store the secret on your server as RAMZ_API_KEY. Never put it in browser code, a shared screenshot or a repository.
  4. Revoke a key from API keys if it is exposed or no longer needed.
Manage API keys
02

Base URL and authentication

https://api.ramzai.org/v1

Send the Authorization header with every API request. A missing, revoked, expired or invalid key receives HTTP 401.

Authorization: Bearer YOUR_API_KEY
{"error":"A valid API key is required."}
03

Model identifiers

Put the request identifier ramz0 in the model field, not the product or company name. Retrieve the account’s model catalog with GET /v1/models; use its data[].id values when choosing from the catalog.

curl https://api.ramzai.org/v1/models \
  -H "Authorization: Bearer $RAMZ_API_KEY"

The response is a JSON object with object: list and a data array. A catalog response is not an inference response.

04

Send a chat request

POST /v1/chat/completions

Use Content-Type: application/json. model identifies the requested model; messages contains the conversation in order. Each message has role and content. Do not send sensitive data unnecessarily.

curl https://api.ramzai.org/v1/chat/completions \
  -H "Authorization: Bearer $RAMZ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "ramz0",
    "messages": [
      {"role": "system", "content": "Answer clearly and concisely."},
      {"role": "user", "content": "Explain speech recognition."}
    ]
  }'

Read the answer from choices[0].message.content. usage reports prompt_tokens, completion_tokens and total_tokens. The response model is ramz0. The service sets a maximum of 500 output tokens; do not send max_tokens. Requests that try to change this limit receive HTTP 400. temperature may be 0 to 1. Streaming and tool calls are not supported by this endpoint. Messages must contain 1 to 32 text messages, with at most 32 KB of content.

05

Credits, errors and request tracking

Each account receives ten free successful chat requests in total, with up to 500 output tokens per request. API and playground share this allowance across all keys; creating another key does not reset it. Failed generation restores the allowance. Accounts begin with $0.00 credit; after the free allowance is exhausted, requests without sufficient credit return HTTP 402. Payments and automatic billing are not enabled. View remaining free requests in Credits and usage in Usage.

{
  "error": {
    "type": "insufficient_credit",
    "code": "insufficient_credit",
    "message": "Your account has insufficient credits."
  },
  "request_id": "REQUEST_ID"
}

At most two requests may run concurrently per account. The request rate limit is 120 per hour per client address, including API and playground attempts. HTTP 400 means invalid parameters, 429 means a rate or concurrency limit, and 502 means generation failed. Do not automatically retry a generation without checking its outcome.

The actual request_id is also returned in the x-request-id response header. Keep it when contacting support; never share your API key. Handle 401 by checking your key, and 402 by checking your account credit. Do not repeatedly retry either error unchanged.