Generate a text completion

This endpoint can be used to generate completions from a Large Language Model. It is a simple proxy forwarding your requests to the desired model. Any LightOn model is deployed on a vLLM-based image. Response Types:

• When stream=false (default): Returns a complete JSON response with all completion choices
• When stream=true: Returns Server-Sent Events (SSE) with incremental completion chunks Streaming Format: Each SSE event contains a JSON object with incremental text. The stream ends with data: [DONE].

POST /api/v2/completions

Try it

cURL

LightOnModelExample

curl --request POST \
  --url https://paradigm.lighton.ai/api/v2/completions \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "model": "alfred-40b-1123",
  "prompt": "Hello, "
}
'

200

LightOnModelExample

{
  "model": "alfred-40b-1123",
  "prompt": "Hello, "
}

Authorizations

Authorization · string · header · required

Bearer token authentication

Body

Request serializer for completions endpoint.

model · string · required

Model to use for generating completions, must exist and be configured from the admin

prompt · string

The prompt to generate completions for

max_tokens · integer

Maximum number of tokens to generate

temperature · number<double>

Sampling temperature between 0 and 2

top_p · number<double>

Nucleus sampling parameter

n · integer

Number of completions to generate

stream · boolean

Whether to stream back partial progress

logprobs · integer | null

Include the log probabilities on the logprobs most likely tokens

echo · boolean

Echo back the prompt in addition to the completion

stop · string

Up to 4 sequences where the API will stop generating further tokens

presence_penalty · number<double>

Penalty for new tokens based on whether they appear in the text so far

frequency_penalty · number<double>

Penalty for new tokens based on their existing frequency in the text

best_of · integer

Generates multiple completions server-side and returns the best

logit_bias · object

Modify the likelihood of specified tokens appearing in the completion

Show child attributes

logit_bias. {key} · any

user · string

A unique identifier representing your end-user

suffix · string

The suffix that comes after a completion of inserted text

Response

200

application/json

Successful response

Response serializer for completions endpoint results.

id · string · required

Unique identifier for the completion

object · string · required

Object type, always 'text_completion'

created · integer · required

Unix timestamp of when the completion was created

model · string · required

The model used for generating the completion

choices · object[] · required

List of completion choices generated by the model

Show child attributes

choices. text · string · required

The generated text completion

choices. index · integer · required

The index of this choice in the list of choices

choices. logprobs · any | null

Log probability information for the choice

choices. finish_reason · string | null

The reason the model stopped generating tokens

usage · any

Usage statistics for the completion request