Chat completions

The OpenAI-compatible generation endpoint. Same request shape, any model.

POST/v1/chat/completions

Send a model and a non-empty messages array. Fields not listed below are passed through to the underlying model.

REQUEST BODY

Parameter	Type	Description
`model`required	`string`	Model ID from List models, or `auto` to let the gateway choose. Accepts residency tags.
`messages`required	`object[]`	Conversation so far. Must contain at least one message. See the message object below.
`stream`	`boolean`	Return the response as server-sent events. Defaults to `false`. See Streaming.
`temperature`	`number`	Sampling temperature. Higher values give more varied output; lower values are more deterministic.
`top_p`	`number`	Nucleus sampling cutoff. Usually set this or temperature, not both.
`max_tokens`	`integer`	Maximum tokens to generate. Also used to estimate quota, so set it to what you need.
`max_completion_tokens`	`integer`	OpenAI-style alias for the output token limit, including reasoning tokens.
`stop`	`string \| string[]`	Sequence(s) at which generation stops.
`seed`	`integer`	Best-effort deterministic sampling.
`n`	`integer`	Number of completions to generate.
`presence_penalty`	`number`	Penalizes tokens that have already appeared, encouraging new topics.
`frequency_penalty`	`number`	Penalizes tokens by how often they have appeared, reducing repetition.
`response_format`	`object`	Output format, for example `{"type":"json_object"}`. See Tools & JSON mode.
`tools`	`object[]`	Functions the model may call, in OpenAI tool format.
`tool_choice`	`string \| object`	Controls tool use: `auto`, `none`, or a specific function.
`stream_options`	`object`	Streaming options. Usage is included in the final chunk by default; set `include_usage` to `false` to opt out.
`reasoning_effort`	`string`	`low`, `medium` or `high`. See Reasoning.
`thinking`	`object`	Provider-style thinking configuration, mapped to the model's native format.
`user`	`string`	Your end-user identifier, for attribution.

MESSAGE OBJECT

Parameter	Type	Description
`role`required	`string`	`system`, `user`, `assistant` or `tool`.
`content`	`string \| object[] \| null`	Message text, or an array of content parts (for example text and images). May be null on assistant messages that only carry tool calls.
`name`	`string`	Optional participant name.
`tool_calls`	`object[]`	Tool calls requested by an assistant message.
`tool_call_id`	`string`	On `tool` messages, the ID of the call being answered.

curl https://api.eggsplain.com/v1/chat/completions \
  -H "Authorization: Bearer $EGGSPLAIN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "sovereign/gpt-6-luna",
  "messages": [
    { "role": "system", "content": "You are concise." },
    { "role": "user", "content": "Hello, eggsplain!" }
  ],
  "temperature": 0.3,
  "max_tokens": 200
}'

Good to know

Identical cacheable requests may be served from cache. Check x-egg-cache.
Audio models are not accepted here; use speech to text.
Edit the request body in the tester to try any parameter.