Create Metric | Cartesia

1 import requests 2 3 url = "https://api.cartesia.ai/agents/metrics" 4 5 payload = { 6 "name": "evaluate-user-satisfaction", 7 "prompt": "Task: 8 Evaluate how engaged and satisfied the user is with the conversation. Engagement may be shown through active interest in the agent’s products/services, expressing that the agent was helpful, or indicating they would want to interact again. 9 10 Decision Logic: 11 - If the user shows strong engagement (asks detailed follow-up questions, expresses high interest, compliments the agent, or states they would use the service/agent again) → classify as HIGH_SATISFACTION 12 - If the user shows some engagement (asks a few relevant questions, shows mild interest, or gives neutral feedback) → classify as MEDIUM_SATISFACTION 13 - If the user shows little or no engagement (short answers, off-topic responses, disinterest, no signs of satisfaction) → classify as LOW_SATISFACTION 14 15 Notes: 16 - Engagement can be verbal (explicit statements of interest) or behavioral (asking more about features, prices, benefits, or next steps). 17 - Expressions of satisfaction, gratitude, or willingness to call again count as positive engagement. 18 - Ignore scripted greetings or polite closings unless they contain genuine feedback. 19 20 Return: 21 Only output the exact category name as a string: HIGH_SATISFACTION, MEDIUM_SATISFACTION, or LOW_SATISFACTION. 22 ", 23 "display_name": "Evaluate User Satisfaction" 24 } 25 headers = { 26 "Cartesia-Version": "2025-04-16", 27 "Authorization": "Bearer <token>", 28 "Content-Type": "application/json" 29 } 30 31 response = requests.post(url, json=payload, headers=headers) 32 33 print(response.json())

1 { 2 "id": "am_abc123", 3 "name": "evaluate-user-satisfaction", 4 "prompt": "Task:\nEvaluate how engaged and satisfied the user is with the conversation. Engagement may be shown through active interest in the agent’s products/services, expressing that the agent was helpful, or indicating they would want to interact again.\n\nDecision Logic:\n- If the user shows strong engagement (asks detailed follow-up questions, expresses high interest, compliments the agent, or states they would use the service/agent again) → classify as HIGH_SATISFACTION\n- If the user shows some engagement (asks a few relevant questions, shows mild interest, or gives neutral feedback) → classify as MEDIUM_SATISFACTION\n- If the user shows little or no engagement (short answers, off-topic responses, disinterest, no signs of satisfaction) → classify as LOW_SATISFACTION\n\nNotes:\n- Engagement can be verbal (explicit statements of interest) or behavioral (asking more about features, prices, benefits, or next steps).\n- Expressions of satisfaction, gratitude, or willingness to call again count as positive engagement.\n- Ignore scripted greetings or polite closings unless they contain genuine feedback.\n\nReturn:\nOnly output the exact category name as a string: HIGH_SATISFACTION, MEDIUM_SATISFACTION, or LOW_SATISFACTION.\n", 5 "created_at": "2025-08-13T16:24:17.457Z", 6 "display_name": "Evaluate User Satisfaction" 7 }

Create a new metric.

Request

This endpoint expects an object.

namestringRequired

The name of the metric. This must be a unique name that only allows lower case letters, numbers, and the characters _, -, and .

promptstringRequired

The prompt associated with the metric, detailing the task and evaluation criteria.

display_namestringOptional

The display name of the metric.

Response

This endpoint returns an object.

idstring

The unique identifier for the metric.

namestring

The name of the metric. This is a unique name that you can use to identify the metric in the CLI.

promptstring

The prompt associated with the metric, detailing the task and evaluation criteria.

created_atdatetime

The timestamp when the metric was created.

display_namestring or null

The display name of the metric, if available. This is the name that is displayed in the Playground.

Headers

Request

Response