Response Structure

Standard Response (`search` / `asearch`)

llm_response

string

The main response from the language model based on search results.

response_time

float

Time taken to process the request in seconds.

input_tokens

integer

Number of input tokens used.

output_tokens

integer

Number of output tokens generated.

llmlayer_cost

integer

LLMLayer infrastructure cost for the request.

model_cost

integer

Model cost for the request, based on the model used.

sources

array

List of source documents used (when return_sources=true).

Show Source object

title

string

Title of the source document.

link

string

URL of the source.

snippet

string

Relevant excerpt from the source.

date

string

date of the source

images

array

List of relevant images (when return_images=true).

Show Image object

imageUrl

string

URL of the image.

thumbnailUrl

string

thumbnail Url of the image.

title

string

Image title or alt text.

Streaming Response (`search_stream` / `asearch_stream`)

The streaming response sends Server-Sent Events with different types:

type='llm'

event

Show Properties

content

string

A chunk of the LLM response text.

type='sources'

event

Show Properties

data

array

Array of source objects (same structure as above).

type='images'

event

Show Properties

data

array

Array of image objects (same structure as above).

type='done'

event

Show Properties

response_time

float

Total processing time in seconds.

type='usage'

event

Show Properties

input_tokens

integer

Total input tokens used.

output_tokens

integer

Total output tokens generated.

llmlayer_cost

integer

Cost for llmlayer infrastructure

model_cost

integer

The cost for the model used

Best Practices

Choose the Right Model

Optimize Token Usage

Filter for Quality

Handle Streaming Properly

JSON Responses

Get Started