Message Feedback

The Chat API provides an endpoint for submitting quality feedback on individual agent messages. This data is used to improve agent behavior and can be queried for analytics.

Endpoint

PUT /chat/message-feedback?tenant={tenant}

Request Body

{
  "message_id": "uuid",
  "judgement": "relevant" | "somewhat_relevant" | "not_relevant",
  "agent_identifier": "string",
  "comment": "string",
  "conversation_id": "uuid"
}

Field	Type	Required	Description
`message_id`	UUID	Yes	The `message_id` from the agent's `ChatMessage` response
`judgement`	enum	Yes	Rating: `"relevant"`, `"somewhat_relevant"`, or `"not_relevant"`
`agent_identifier`	string	Yes	The agent that produced the message
`comment`	string	Yes	Free-text feedback or structured reasons
`conversation_id`	UUID	No	Links feedback to a conversation — use the `session_id` carried by the agent's message (see Conversations)

Example

import os
import requests

TENANT = "zetaalpha"
headers = {
    "Content-Type": "application/json",
    "X-Auth": os.getenv("ZETA_ALPHA_API_KEY"),
}

requests.put(
    f"https://api.zeta-alpha.com/v0/service/chat/message-feedback?tenant={TENANT}",
    headers=headers,
    json={
        "message_id": "the-message-id-from-response",
        "judgement": "relevant",
        "agent_identifier": "custom_agent",
        "comment": "Accurate and well-sourced answer",
    },
)

Getting the `message_id`

The message_id is returned in the agent's ChatMessage:

Non-streaming: response.json()["conversation"][-1]["message_id"]
Streaming: In the final SSE event's data, message["message_id"]

Judgement Values

Value	Meaning	Typical UI
`relevant`	The answer was helpful and accurate	👍 Thumbs up
`somewhat_relevant`	Partially useful but not fully satisfying	Neutral / partial
`not_relevant`	The answer was unhelpful or incorrect	👎 Thumbs down

Using the `comment` Field

The comment field accepts free-text. Common patterns:

User-typed feedback: "The answer was missing information about X"
Structured reasons: "factual_error,missing_citation"
Empty acknowledgment: "User clicked thumbs up"

Use whatever format suits your application — the field is stored as-is for later analysis.

Endpoint​

Request Body​

Example​

Getting the message_id​

Judgement Values​

Using the comment Field​