docs / llm

AI Text

Run prompts through a hosted language model.

base /llm/v12 endpoints

post/llm/v1/complete3 credits

prompt -> generated text. params: prompt (str, required) + optional system (str), max_tokens (int, default 512, cap 4096), temperature (float 0-2, default 0.7), top_p, think (bool, default false — Qwen3 reasoning mode; off = direct answer, faster/cheaper, best for overlay), model. Returns text + finish_reason + usage.

Parameter	Required
prompt	required
system	optional
max_tokens	optional
temperature	optional
top_p	optional
think	optional
model	optional

Try in playground →

post/llm/v1/chat3 credits

OpenAI-style chat. params: messages (list of {role, content}, required) + optional max_tokens, temperature, top_p, think (bool, default false), model. Returns message {role, content} + text + usage.

Parameter	Required
messages	required
max_tokens	optional
temperature	optional
top_p	optional
think	optional
model	optional

Try in playground →

Example request · complete

curl -X POST https://api.reefapi.com/llm/v1/complete \
  -H "x-api-key: $REEF_KEY" \
  -H "content-type: application/json" \
  -d '{"prompt":"Reply with exactly the single word: PONG","max_tokens":8,"temperature":0}'

Response shape

{
  "ok": true,
  "data": { /* the result */ },
  "meta": {
    "latency_ms": 240,
    "record_count": 12,
    "completeness_pct": 100
  },
  "error": null
}