Deep dive: fraud alert verification (voice)

import { BimpeAI } from "@bimpeai/sdk";

const bimpe = new BimpeAI({ apiKey: process.env.BIMPEAI_API_KEY! });

const page = await bimpe.workflows.list({ scope: "public", search: "fraud verification" });
const workflow = page.data[0];
console.log(workflow.id, workflow.name);

// Inspect the full workflow to understand its conversation branches
const full = await bimpe.workflows.retrieve(workflow.id);
console.log("Branches:", full.flows?.map((f) => f.name));

import os
from bimpeai import BimpeAI

client = BimpeAI(api_key=os.environ["BIMPEAI_API_KEY"])

page = client.workflows.list(scope="public", search="fraud verification")
workflow = page.data[0]
print(workflow.id, workflow.name)

# Inspect the full workflow to understand its conversation branches
full = client.workflows.retrieve(workflow.id)
if full.flows:
    print("Branches:", [f.name for f in full.flows])

const agent = await bimpe.agents.create(
  {
    name: "PrimoAI Bank — fraud alert agent",
    description: "Outbound fraud verification agent for PrimoAI Bank.",
    system_prompt:
      "You are the automated fraud alert agent for PrimoAI Bank. " +
      "You have been given a flagged transaction and a customer's phone number. Your job is to call the customer, verify their identity, and determine whether the transaction is authorised. " +
      "Introduction: Greet the customer, state that you are calling from PrimoAI Bank's fraud team, and confirm you are speaking to the account holder by asking them to confirm their name and the last four digits of their card. " +
      "If they cannot confirm identity: do not proceed. Tell them to call the number on the back of their card and end the call. " +
      "Transaction verification: State the transaction amount, merchant, and date. Ask: 'Did you make this transaction?' " +
      "If yes: thank them, confirm the transaction will be marked as verified, and end the call politely. " +
      "If no or unsure: tell them to freeze their card immediately in their banking app (Settings → Cards → Freeze). " +
      "Tell them a fraud analyst will call them back within 60 minutes on this number. " +
      "Do not attempt to block the card yourself — that is handled by the card management team. " +
      "If no answer: leave a voicemail stating there is an urgent security matter and they should call the number on the back of their card. " +
      "Never read out the full card number. Never transfer to an external number. Never state the call is unrecorded — all calls are recorded.",
    agent_workflow_id: workflow.id,
    language: "en",
    timezone: "Europe/London",
  },
  { idempotencyKey: "create-primoai-fraud-agent-v1" },
);

console.log("Agent ID:", agent.id);

agent = client.agents.create(
    name="PrimoAI Bank — fraud alert agent",
    description="Outbound fraud verification agent for PrimoAI Bank.",
    system_prompt=(
        "You are the automated fraud alert agent for PrimoAI Bank. "
        "You have been given a flagged transaction and a customer's phone number. Your job is to call the customer, verify their identity, and determine whether the transaction is authorised. "
        "Introduction: Greet the customer, state that you are calling from PrimoAI Bank's fraud team, and confirm you are speaking to the account holder by asking them to confirm their name and the last four digits of their card. "
        "If they cannot confirm identity: do not proceed. Tell them to call the number on the back of their card and end the call. "
        "Transaction verification: State the transaction amount, merchant, and date. Ask: 'Did you make this transaction?' "
        "If yes: thank them, confirm the transaction will be marked as verified, and end the call politely. "
        "If no or unsure: tell them to freeze their card immediately in their banking app (Settings → Cards → Freeze). "
        "Tell them a fraud analyst will call them back within 60 minutes on this number. "
        "Do not attempt to block the card yourself — that is handled by the card management team. "
        "If no answer: leave a voicemail stating there is an urgent security matter and they should call the number on the back of their card. "
        "Never read out the full card number. Never transfer to an external number. Never state the call is unrecorded — all calls are recorded."
    ),
    agent_workflow_id=workflow.id,
    language="en",
    timezone="Europe/London",
    idempotency_key="create-primoai-fraud-agent-v1",
)

print("Agent ID:", agent.id)

// Fraud response policy and verification question sequence
await bimpe.agents.knowledgeBases.create(agent.id, {
  type: "text",
  name: "Fraud response policy",
  content:
    "A transaction is flagged when it meets one or more of these criteria:\n" +
    "- Amount exceeds 3× the customer's 30-day average spend\n" +
    "- Transaction originates from a country the customer has not previously transacted in\n" +
    "- Merchant matches a known fraud pattern in our monitoring system\n" +
    "- Transaction occurs between 02:00–05:00 local time with no prior activity in that window\n" +
    "\n" +
    "Identity verification (ask in order, stop if identity is not confirmed):\n" +
    "1. 'Can you confirm the name on the account?'\n" +
    "2. 'Can you confirm the last four digits of the card used?'\n" +
    "\n" +
    "Transaction verification question:\n" +
    "'On [date] a transaction of [amount] was made at [merchant]. Did you authorise this?'\n" +
    "\n" +
    "If transaction is denied:\n" +
    "1. Ask the customer to freeze their card immediately via the banking app.\n" +
    "2. Confirm a fraud analyst will call back within 60 minutes.\n" +
    "3. Do not attempt to block the card during the call — the card management team handles this after the call.\n" +
    "\n" +
    "Regulatory note: all calls are recorded for quality and compliance purposes. State this if asked.",
});

// Card freeze and block instructions
await bimpe.agents.knowledgeBases.create(agent.id, {
  type: "text",
  name: "Card freeze instructions",
  content:
    "Freeze via banking app (immediate effect):\n" +
    "  Open PrimoAI Bank app → tap Cards → select the card → tap Freeze card.\n" +
    "  A freeze stops new contactless and online transactions but allows ATM withdrawals with the PIN.\n" +
    "\n" +
    "Freeze via telephone banking:\n" +
    "  Call the number on the back of the card. Automated freeze takes under 2 minutes.\n" +
    "\n" +
    "Freeze vs. cancel:\n" +
    "  A freeze is temporary and reversible — useful while the investigation is active.\n" +
    "  A permanent cancel requires a replacement card (5–7 business days standard, 1–2 days express).\n" +
    "  The fraud analyst will recommend cancel vs. keep frozen during the follow-up call.\n" +
    "\n" +
    "Pending transactions at the time of freeze: these may still settle. The fraud team reviews and refunds disputed settled transactions within 5 business days.",
});

# Fraud response policy and verification question sequence
client.agents.knowledge_bases.create(agent.id, {
    "type": "text",
    "name": "Fraud response policy",
    "content": (
        "A transaction is flagged when it meets one or more of these criteria:\n"
        "- Amount exceeds 3× the customer's 30-day average spend\n"
        "- Transaction originates from a country the customer has not previously transacted in\n"
        "- Merchant matches a known fraud pattern in our monitoring system\n"
        "- Transaction occurs between 02:00–05:00 local time with no prior activity in that window\n"
        "\n"
        "Identity verification (ask in order, stop if identity is not confirmed):\n"
        "1. 'Can you confirm the name on the account?'\n"
        "2. 'Can you confirm the last four digits of the card used?'\n"
        "\n"
        "Transaction verification question:\n"
        "'On [date] a transaction of [amount] was made at [merchant]. Did you authorise this?'\n"
        "\n"
        "If transaction is denied:\n"
        "1. Ask the customer to freeze their card immediately via the banking app.\n"
        "2. Confirm a fraud analyst will call back within 60 minutes.\n"
        "3. Do not attempt to block the card during the call — the card management team handles this after the call.\n"
        "\n"
        "Regulatory note: all calls are recorded for quality and compliance purposes. State this if asked."
    ),
})

# Card freeze and block instructions
client.agents.knowledge_bases.create(agent.id, {
    "type": "text",
    "name": "Card freeze instructions",
    "content": (
        "Freeze via banking app (immediate effect):\n"
        "  Open PrimoAI Bank app → tap Cards → select the card → tap Freeze card.\n"
        "  A freeze stops new contactless and online transactions but allows ATM withdrawals with the PIN.\n"
        "\n"
        "Freeze via telephone banking:\n"
        "  Call the number on the back of the card. Automated freeze takes under 2 minutes.\n"
        "\n"
        "Freeze vs. cancel:\n"
        "  A freeze is temporary and reversible — useful while the investigation is active.\n"
        "  A permanent cancel requires a replacement card (5–7 business days standard, 1–2 days express).\n"
        "  The fraud analyst will recommend cancel vs. keep frozen during the follow-up call.\n"
        "\n"
        "Pending transactions at the time of freeze: these may still settle. The fraud team reviews and refunds disputed settled transactions within 5 business days."
    ),
})

// Verify integrations are connected (dashboard must connect these first)
const integrations = await bimpe.agents.integrations.list(agent.id);
console.log("Integrations:", integrations.map((i) => i.name));
// Expected: ["Transaction monitoring", "Customer contact DB"]

// Channels list reflects what was configured in the dashboard
const channels = await bimpe.agents.channels.list(agent.id);
console.log("Channels:", channels.map((c) => c.type));
// Note: voice channel type appears here but there is no "voice" filter in conversations.list

# Verify integrations are connected (dashboard must connect these first)
integrations = client.agents.integrations.list(agent.id)
print("Integrations:", [i.name for i in integrations])
# Expected: ['Transaction monitoring', 'Customer contact DB']

# Channels list reflects what was configured in the dashboard
channels = client.agents.channels.list(agent.id)
print("Channels:", [c.type for c in channels])
# Note: voice channel type appears here but there is no "voice" filter in conversations.list

// List all recent conversations (voice calls will appear here after they complete)
const conversations = await bimpe.conversations.list(agent.id);

for (const conv of conversations.data) {
  console.log(conv.id, conv.created_at);

  // Retrieve the full message history of a completed call
  const messages = await bimpe.conversations.messages.list(agent.id, conv.id);
  for (const msg of messages.data) {
    console.log(`[${msg.role}] ${msg.message}`);
  }
}

# List all recent conversations (voice calls will appear here after they complete)
conversations = client.conversations.list(agent.id)

for conv in conversations.data:
    print(conv.id, conv.created_at)

    # Retrieve the full message history of a completed call
    messages = client.conversations.messages.list(agent.id, conv.id)
    for msg in messages.data:
        print(f"[{msg.role}] {msg.message}")

import { BimpeAI } from "@bimpeai/sdk";

const bimpe = new BimpeAI({ apiKey: process.env.BIMPEAI_API_KEY! });

// 1. Find workflow
const page = await bimpe.workflows.list({ scope: "public", search: "fraud verification" });
const workflow = page.data[0];

// 2. Create agent
const agent = await bimpe.agents.create(
  {
    name: "PrimoAI Bank — fraud alert agent",
    system_prompt:
      "You are the automated fraud alert agent for PrimoAI Bank. " +
      "Call the cardholder, verify identity (name + last 4 digits), then ask about the flagged transaction. " +
      "If confirmed: end the call. " +
      "If denied: instruct card freeze and tell them a fraud analyst calls back within 60 minutes. " +
      "Never read the full card number. All calls are recorded.",
    agent_workflow_id: workflow.id,
    language: "en",
    timezone: "Europe/London",
  },
  { idempotencyKey: "create-primoai-fraud-agent-v1" },
);

// 3. Add knowledge bases
await bimpe.agents.knowledgeBases.create(agent.id, {
  type: "text",
  name: "Fraud response policy",
  content:
    "Identity verification: confirm name + last 4 card digits before proceeding.\n" +
    "Transaction verification question: 'On [date] a transaction of [amount] at [merchant]. Did you authorise this?'\n" +
    "If denied: instruct card freeze via app (Cards → Freeze), confirm analyst callback within 60 minutes.\n" +
    "Do not block the card during the call — card management team handles this after the call.",
});

await bimpe.agents.knowledgeBases.create(agent.id, {
  type: "text",
  name: "Card freeze instructions",
  content:
    "Freeze via app: PrimoAI Bank app → Cards → select card → Freeze card.\n" +
    "Freeze via phone: call number on back of card, automated system freezes in 2 minutes.\n" +
    "Freeze is temporary and reversible. Analyst will advise cancel vs. keep frozen.",
});

// 4. Verify integrations (all connected in the dashboard — voice channel is also dashboard-only)
const integrations = await bimpe.agents.integrations.list(agent.id);
const channels = await bimpe.agents.channels.list(agent.id);

console.log("Agent ID:", agent.id);
console.log("Integrations:", integrations.map((i) => i.name));
console.log("Channels:", channels.map((c) => c.type));

// 5. Read completed call outcomes
// Outbound calls are initiated from the dashboard or the transaction monitoring webhook.
// The SDK cannot initiate calls. After a call completes it appears in conversations.list.
const conversations = await bimpe.conversations.list(agent.id);
console.log("Completed conversations:", conversations.data.length);

import os
from bimpeai import BimpeAI

client = BimpeAI(api_key=os.environ["BIMPEAI_API_KEY"])

# 1. Find workflow
page = client.workflows.list(scope="public", search="fraud verification")
workflow = page.data[0]

# 2. Create agent
agent = client.agents.create(
    name="PrimoAI Bank — fraud alert agent",
    system_prompt=(
        "You are the automated fraud alert agent for PrimoAI Bank. "
        "Call the cardholder, verify identity (name + last 4 digits), then ask about the flagged transaction. "
        "If confirmed: end the call. "
        "If denied: instruct card freeze and tell them a fraud analyst calls back within 60 minutes. "
        "Never read the full card number. All calls are recorded."
    ),
    agent_workflow_id=workflow.id,
    language="en",
    timezone="Europe/London",
    idempotency_key="create-primoai-fraud-agent-v1",
)

# 3. Add knowledge bases
client.agents.knowledge_bases.create(agent.id, {
    "type": "text",
    "name": "Fraud response policy",
    "content": (
        "Identity verification: confirm name + last 4 card digits before proceeding.\n"
        "Transaction verification question: 'On [date] a transaction of [amount] at [merchant]. Did you authorise this?'\n"
        "If denied: instruct card freeze via app (Cards → Freeze), confirm analyst callback within 60 minutes.\n"
        "Do not block the card during the call — card management team handles this after the call."
    ),
})

client.agents.knowledge_bases.create(agent.id, {
    "type": "text",
    "name": "Card freeze instructions",
    "content": (
        "Freeze via app: PrimoAI Bank app → Cards → select card → Freeze card.\n"
        "Freeze via phone: call number on back of card, automated system freezes in 2 minutes.\n"
        "Freeze is temporary and reversible. Analyst will advise cancel vs. keep frozen."
    ),
})

# 4. Verify integrations (all connected in the dashboard — voice channel is also dashboard-only)
integrations = client.agents.integrations.list(agent.id)
channels = client.agents.channels.list(agent.id)

print("Agent ID:", agent.id)
print("Integrations:", [i.name for i in integrations])
print("Channels:", [c.type for c in channels])

# 5. Read completed call outcomes
# Outbound calls are initiated from the dashboard or the transaction monitoring webhook.
# The SDK cannot initiate calls. After a call completes it appears in conversations.list.
conversations = client.conversations.list(agent.id)
print("Completed conversations:", len(conversations.data))