POST /v1/completions

The legacy text-completion endpoint. Same shape as OpenAI's /v1/completions. Most clients use chat completions instead, but completion-style models (older base models, code-completion models) still use this surface.

Request

curl https://api.inference.club/v1/completions \
  -H "Authorization: Bearer $INFERENCE_CLUB_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "code-llama-7b",
    "prompt": "def fizzbuzz(n):",
    "max_tokens": 100
  }'

Routing rules and authentication are identical to chat completions. The body is forwarded to <provider-callback-url>/completions unchanged, and streaming works the same way ("stream": true).

If you don't have a clear reason to use this endpoint, use chat completions — modern instruct-tuned models are trained for that format.