API Reference

Endpoint

Every request goes to:

https://api.revomail.io/v1

Revo Mail implements the OpenAI Chat Completions interface, so any library or tooling built for that format works out of the box — just swap the base URL and API key.

Authentication

Attach your Revo Mail key to every request using the Authorization header:

Authorization: Bearer YOUR_API_KEY

Keys are generated in the developer console. A missing or expired key returns 401 Unauthorized.

Compatibility Summary

Sending a Chat Request

Required Fields

{
  "model": "gpt-5.4",
  "messages": [
    { "role": "system", "content": "You are a precise technical writer." },
    { "role": "user",   "content": "Describe API rate limits in two sentences." }
  ]
}

Optional Fields

Response Structure

{
  "id": "chatcmpl-xyz789",
  "object": "chat.completion",
  "created": 1714000000,
  "model": "gpt-5.4",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Rate limits cap the number of requests per unit time..."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 30,
    "completion_tokens": 18,
    "total_tokens": 48
  }
}

The usage block reflects the exact token counts billed for this call.

Streaming

Pass "stream": true to receive a stream of server-sent events (SSE). The connection stays open and each chunk arrives as a data: line:

curl https://api.revomail.io/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.2",
    "messages": [{"role": "user", "content": "List five programming languages."}],
    "stream": true
  }'

The stream terminates with a final data: [DONE] line. Handle it in your client to know when the response is complete.

Caching

Revo Mail automatically caches eligible responses to reduce cost and latency. When a cached response is returned, the response includes the header:

X-Revo Mail-Cache: HIT

Caching applies to non-streaming requests with deterministic settings (temperature ≤ 0 and top_p = 1). Two layers are used:

• Exact match — identical model + messages return a cached response (1-hour TTL)
• Semantic match — highly similar requests (95%+ similarity) may return a cached response

See Advanced Features for full details.

Error Reference