Skip to main content
POST
Create evaluation
Create a new evaluation configuration. eval_name and call_source are required. Updating an existing one: see Update Evaluation.

Headers

token
string
required

API token for authentication

Body

application/json
eval_name
string
required

Evaluation name

call_source
string
required

agent, app, recording, api or webhook

agent_id
string

Agent(s) to evaluate (agent source)

app_id
string

App to evaluate (app source)

source_config
object

Source-specific config (api / webhook)

system_prompt
string

Evaluation system prompt

llm_provider
string

LLM provider name, e.g. openai. See the list of providers and models below.

llm_model
string

LLM model name of the chosen provider, e.g. gpt-4o-mini. See the list of providers and models below.

fallback_provider
string

Fallback LLM provider name. See the list of providers and models below.

fallback_model
string

Fallback model name of the chosen fallback provider. See the list of providers and models below.

transcription_provider
string

Transcription provider name (recording source), e.g. deepgram. See the list of providers and models below.

transcription_model
string

Transcription model name of the chosen provider, e.g. nova-2. See the list of providers and models below.

temperature
integer

Sampling temperature

skip_shorter_than
integer

Skip calls shorter than this many seconds

run_on_all_past_and_future_calls
boolean

Evaluate past calls as well as future ones

ground_truth_sources
string[]

Ground-truth sources

cron_expression
string

Schedule for recurring evaluation

status
integer

Evaluation status

Response

200 - application/json

Success

The response is of type object.