Client recipes

Nyeti does not require a client SDK. Call the authenticated HTTP endpoint from your backend with the standard client for your language.

Node / TypeScript

TypeScript

export async function analyze(message: string) {
  const res = await fetch("https://api.bactlabs.africa/v1/analyze", {
    method: "POST",
    headers: {
      Authorization: `Bearer ${process.env.NYETI_API_KEY}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({ message }),
  });

  if (!res.ok) throw new Error(`Nyeti request failed: ${res.status}`);
  return res.json();
}

Python

import os
import requests

def analyze(message: str) -> dict:
    response = requests.post(
        "https://api.bactlabs.africa/v1/analyze",
        headers={"Authorization": f"Bearer {os.environ['NYETI_API_KEY']}"},
        json={"message": message},
        timeout=10,
    )
    response.raise_for_status()
    return response.json()

Correlate a conversation

Keep the ordinary one-message request shape. Send turns in order with the same conversation_id and identify whether each turn came from the customer or agent.

TypeScript

async function analyzeTurn(
  message: string,
  conversationId: string,
  speaker: "customer" | "agent",
) {
  const res = await fetch("https://api.bactlabs.africa/v1/analyze", {
    method: "POST",
    headers: {
      Authorization: `Bearer ${process.env.NYETI_API_KEY}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      message,
      conversation_id: conversationId,
      speaker,
    }),
  });

  if (!res.ok) throw new Error(`Nyeti request failed: ${res.status}`);
  return res.json();
}

await analyzeTurn("How can I help?", "ticket_123", "agent");
await analyzeTurn("My order still has not arrived.", "ticket_123", "customer");
await analyzeTurn("It is on its way, so this is resolved.", "ticket_123", "agent");
const result = await analyzeTurn(
  "You said that last time and nobody has helped me.",
  "ticket_123",
  "customer",
);

if (result.conversation?.review_status === "needs_review") {
  queueForHumanReview(result.conversation);
}

Safety routing

Treat CRISIS_SUPPORT as a signal to enter your established human safety workflow. Do not treat Nyeti as an emergency service or diagnosis tool.

TypeScript

const result = await analyze(userMessage);

if (result.routing.response_posture === "CRISIS_SUPPORT") {
  return handOffToTrainedResponder({ messageId, result });
}

return continueNormalFlow(result.routing);

Completed voice calls

For completed conversations, use a supported signed integration channel instead of writing a transcript replay loop yourself.