Service Guide

Service Guide

This service supports two API-key modes. The mode is stored on the API key itself and controls how requests are handled.

1. Create an account

Open Register, create a username and password, then sign in.

2. Create an API key

In the dashboard open Create API Key and choose one of the modes below.

Mode	Use it for	Main endpoint
emulator	Codex and GPT-5.4 emulation	/v1/responses
bypass	Raw Ollama-compatible chat requests	/v1/chat/completions

3. Optional: add your own Ollama Cloud keys

If you want the service to use your own Ollama Cloud account, add one or more Ollama site API keys in Ollama Cloud Keys.

Then choose:

• which Ollama key is currently preferred
• which Ollama cloud model should be used with it

If no Ollama key/model is selected, the service falls back to the shared upstream configured on the server.

4. Resource display and failover

For every Ollama key the dashboard shows:

• Spent in 7 days
• Spent in 5 hours
• Last checked
• Status

Those values are cached and are not refreshed more often than once per 10 minutes per key.

If the preferred Ollama key is exhausted, the service automatically tries the next non-exhausted key belonging to the same user.

Using emulator mode

Use an emulator key when your client expects GPT-5.4-style Responses API behavior.

curl -sS https://llm.chat-artin.ru/v1/responses \
  -H 'Authorization: Bearer <your-emulator-key>' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "gpt-5.4",
    "input": [
      {
        "type": "message",
        "role": "user",
        "content": [{"type": "input_text", "text": "Reply with hello"}]
      }
    ],
    "stream": true
  }'

Using bypass mode

Use a bypass key when you want raw Ollama-compatible chat/completions responses without GPT-5.4 emulation.

curl -sS https://llm.chat-artin.ru/v1/chat/completions \
  -H 'Authorization: Bearer <your-bypass-key>' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "qwen3-coder-next:cloud",
    "messages": [
      {"role": "user", "content": "Reply with bypass-ok"}
    ]
  }'

Mode rules

• emulator keys are meant for /v1/responses
• bypass keys are meant for /v1/chat/completions
• If you use the wrong endpoint for the key mode, the service returns a descriptive error

Available endpoints

• GET /healthz - service health check
• GET /v1/models - models list for the current key mode
• POST /v1/responses - GPT-5.4 emulator path
• POST /v1/chat/completions - raw bypass path