export const article = {
  author: 'Yongrui Su',
  date: '2025-10-29',
  title: `Protect Sensitive Data with PII Anonymization`,
  avatar: '/images/yongrui_su_profile.jpeg',
  hashtags: ['Data Masking', 'PII Protection', 'AI Data Privacy', 'GDPR Compliance'],
  thumbnail,
  preview: `
Automatically protect your customers' personal information before it reaches your AI models, while keeping conversation context intact for excellent support.

**Highlights**

🔒 **Flexible PII Protection** - Choose which personal data to anonymize per workspace or integration—from emails and phone numbers to credit cards and national IDs

✅ **Intelligent Pre-Processing** - Automatically mask sensitive data before it reaches AI models or gets stored, with no workflow changes needed

🌍 **Privacy by Default** - Meet GDPR and data minimization requirements while maintaining full conversation context for your team
`,
};


We're excited to launch **PII Anonymization**—automatically protect your customers' personal information before it reaches your AI models, while keeping conversation context intact so your team can deliver excellent support.

---

## What is PII Anonymization?

PII Anonymization is your **automatic privacy shield** that sits between your customers and your AI systems. Think of it as a smart filter that identifies and masks personal information—like email addresses, phone numbers, credit cards, and national IDs—before that data reaches your AI models or gets stored in Chat Data.

🔒 **Works Automatically** - Once enabled, every customer message is scanned and sensitive data is replaced with consistent placeholders like EMAIL_1 or PHONE_1.

💬 **Preserves Context** - Your AI and support team still see the full conversation flow and can reference "the email address" or "that phone number" without seeing the actual values.

🎯 **Granular Control** - Choose exactly which types of personal data to protect, and apply different rules per workspace or integration.

---

## Why This Matters for Your Business

When customers share personal details with your chatbot, you want to provide great service while keeping their information secure. PII Anonymization gives you both:

**For E-commerce & Retail**
Customers naturally share order numbers, shipping addresses, and payment details. Anonymization lets your AI help with order tracking and support without exposing sensitive transaction data to third-party models.

**For Healthcare & Wellness**
Patient inquiries often include medical IDs, insurance numbers, and personal health details. Keep your AI assistance HIPAA-friendly by masking identifiers before they reach your LLM provider.

**For Financial Services**
Banking and fintech conversations involve account numbers, tax IDs, and transaction details. Anonymization protects customer data while your AI provides account support and answers questions.

**For Everyone**
🛡️ **Customer Trust** - Show customers you take their privacy seriously by protecting their data by default.
📊 **Safe AI Training** - Use conversation data to improve your chatbot without worrying about sensitive information leaking into training sets.
✅ **Simple Compliance** - Meet GDPR, CCPA, and SOC 2 data minimization requirements without building custom middleware or rewriting your workflows.

---

## What You Can Protect

PII Anonymization gives you fine-grained control over exactly which types of personal information get masked. Here's what you can configure:

### Built-in Protection Categories

Toggle anonymization for any of these common personal identifiers:

- 👤 **Name - Person Identification** - Full names and personal identifiers
- 📧 **Email Address** - Email addresses in any format
- 📱 **Phone Number** - Phone numbers with international format support
- 💳 **Credit Card Number** - Payment card numbers and related tokens
- 🌐 **IP Address** - IPv4 and IPv6 addresses
- 📍 **Location - Address Information** - Physical addresses and location details
- 🆔 **Tax Identification Number** - Social security numbers, VAT IDs, and tax identifiers
- 🏥 **Medical Information** - Medical license numbers and health identifiers
- 🏦 **Banking Details** - Account numbers, IBAN codes, and routing numbers
- 🚗 **Government IDs** - Passport numbers, driver's licenses, and national ID cards

### Smart Placeholders

When data is anonymized, it's replaced with **consistent, descriptive placeholders** like EMAIL_1, PHONE_2, or CREDIT_CARD_1. This means:
- Your AI can still reference "the email address" in conversation naturally
- Your support team sees the full conversation flow and context
- Automations and workflows continue working without modification
- The same email address always gets the same placeholder ID within a conversation

---

## How It Works

PII Anonymization operates automatically in the background with zero workflow changes required. When a customer shares sensitive information, the system instantly detects and replaces it with consistent placeholders like EMAIL_1, PHONE_1, or CREDIT_CARD_1 before the data reaches your AI models or gets stored.

Your AI and team still see full conversation context and can naturally reference "the email address" or "that phone number" without ever seeing the actual values. This means you can analyze patterns, train models, and deliver excellent support while keeping customer data private by default.

---

## Getting Started

Ready to protect your customer data? Head to **Settings > Privacy** in your Chat Data workspace to enable protection categories and start anonymizing sensitive information automatically.

Protect Sensitive Data with PII Anonymization

export const article = {
  author: 'Yongrui Su',
  date: '2025-10-16',
  title: `Launch: Build Multi-Step AI Agent Workflows with Auto Prompt`,
  avatar: '/images/yongrui_su_profile.jpeg',
  hashtags: ['Workflow', 'Automation', 'AI', 'AI Agents'],
  thumbnail: workflowBuilderExample,
  preview: `
Transform your chatbot into an intelligent AI agent with **Workflows**—multi-step conversation paths for lead qualification, onboarding, appointment booking, and complex customer journeys.

**Highlights**

🎯 **Workflows as AI Actions** - Trigger complex, multi-step processes with isolated session management

🤖 **AI-Powered Generation** - Build workflows in natural language with Auto Prompt, or use manual drag-and-drop

🧪 **Visual Testing & Debugging** - Real-time node transitions with full debug info and AI-powered test simulations

🌐 **Multi-Platform Ready** - Deploy across website widgets and third-party integrations (WhatsApp, Messenger, Instagram, Telegram, Slack, Discord)

📦 **Easy Management** - Centralized variables, import/export workflows, and collaborate with your team
`,
};


We're excited to launch the **Workflow Builder**—a powerful new way to create sophisticated, multi-step chatbot experiences that go beyond simple conversations.


---

## What Are Workflows?

Workflows are **specific AI actions** that guide your chatbot through complex, multi-step processes. Think of them as automated conversation paths for lead qualification, customer onboarding, appointment booking, or any process that requires multiple steps and decision points.

🎯 **Triggered Like AI Actions** - Each workflow operates as a dedicated AI action that can be triggered just like your other chatbot actions.

🔄 **Isolated Sessions** - When a workflow starts, your chatbot enters a focused session following your defined logic step by step. Once complete, it returns to normal conversation and can trigger another workflow if needed.

🎭 **Flexible AI Behavior** - AI Conversation nodes within workflows can use completely different base prompts tailored to specific tasks, letting you optimize responses for each step.

⚡ **Clean Execution** - Only one workflow runs per conversation at a time, preventing conflicts and ensuring predictable paths.

---

## How to Build Workflows

Create workflows in two ways—whichever fits your needs:

### AI-Powered Generation (Recommended)

🤖 **Two-Phase Process** - Describe your goal in natural language, review the generated blueprint, give feedback, and regenerate until it's right. Works for both new workflows and updating existing ones.

✅ **Instant Build** - Once approved, AI builds the complete workflow with all nodes, variables, and connections.

🔄 **Easy Updates** - Use Auto Prompt to modify existing workflows too. Simply describe what you want to change, and AI will update your workflow to meet your new business goals.

🎯 **Smart Configuration** - AI automatically ensures proper data capture, branching logic, and integrations are configured correctly.

📊 **Safe Review** - Every AI-generated change is marked as unsaved so you can review before publishing.

💡 **Best Practice**: After AI generates your workflow, audit variable types—some boolean flags may be created as strings and should be corrected to ensure reliable condition checks (e.g., true/false instead of 'yes'/'true').

### Manual Drag-and-Drop

🎨 **Visual Canvas** - Drag-and-drop pre-built blocks for messaging, UI elements, data capture, conditional logic, automation, code execution, and variable management.

🎛️ **Node Editors** - Configure each step in the right panel with easy access to email content, API payloads, form fields, validation rules, and wait timers.

---

## Testing Your Workflows

Before going live, thoroughly test your workflows with built-in tools:

🧪 **Visual Testing Panel** - Watch real-time node transitions with full debug information showing exactly why the workflow moved from one step to another, making validation and troubleshooting effortless.

⚙️ **Test Scenarios** - Configure visitor presets, seed conversations with scripted openings, and let the simulator detect loops or stalled edges.

💾 **Test Set Management** - Save test scenarios and run AI-powered automatic simulations to verify your workflow behaves correctly across different user paths.

💬 **Live Debug Info** - Conversation transcripts, transition breadcrumbs, and variable snapshots update in real-time so you can verify your logic.

---

## Where Workflows Work

Deploy your workflows across all your customer touchpoints:

🌐 **Multi-Platform Support** - Workflows work seamlessly across website widgets and third-party integrations (WhatsApp, Messenger, Instagram, Telegram, Slack, Discord).

🎴 **Smart Adaptation** - Rich UI elements like carousel cards adapt intelligently to each platform. When a platform can't render interactive components, we automatically send the content as files.

🔌 **API Coming Soon** - API integration compatibility is in development for our next release.

---

## Managing Your Workflows

Keep your workflows organized and shareable:

🔐 **Centralized Variables** - Manage all your workflow variables in one place with easy search, rename, and organization capabilities.

📦 **Import/Export** - Share workflows across environments, collaborate with team members, or create backup copies with secure import and export functionality.

---

## What's Next

We're already working on the next wave of workflow enhancements:

🔌 **API Compatibility** - Enable workflows to work with API integrations

🌍 **Static Message Internationalization** - Automatically adapt static messages to match your visitors' browser language

📚 **More Templates** - Ready-to-use workflow templates you can customize for common use cases

🎨 **Additional UI Elements** - New interactive components based on your feedback to create richer user experiences

---

 Check out our [Workflow Builder documentation](https://cookbook.chat-data.com/docs/workflow) for detailed guides.

Launch: Build Multi-Step AI Agent Workflows with Auto Prompt

export const article = {
  author: 'Samuel Su',
  date: '2025-10-07',
  title: `New API Endpoint: Attach Lead & Enhanced Customer Filtering`,
  avatar: '/images/samuel_su_profile.png',
  hashtags: ['API', 'Lead-Management', 'Filtering'],
  thumbnail,
  preview: `We've added a new **/attach-lead** endpoint and enhanced the /get-customers endpoint with filtering capabilities.

**What's New:**
- New POST **/attach-lead** endpoint to programmatically attach or update leads to conversations
- **/get-customers** now supports filtering by leadId, email, phone, name, psid, ig_username, and username
- All customer records now include leadId field for consistent reference
`,
};


We've released two API enhancements to improve lead management and customer data integration.

---

## New Endpoint: POST /attach-lead

Attach or update lead information for a specific conversation.

**Request Example:**

```bash
POST /api/v2/attach-lead
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

{
  "conversationId": "conv_123456",
  "chatbotId": "chatbot_789",
  "email": "customer@example.com",
  "phone": "+1234567890",
  "name": "John Doe"
}
```

**Response:**

```json
{
  "status": "success",
  "lead": {
    "leadId": "550e8400-e29b-41d4-a716-446655440000",
    "name": "John Doe",
    "email": "customer@example.com",
    "phone": "+1234567890",
    "source": "api"
  }
}
```

**Common Use Cases:**

- Sync customer data from your CRM (Salesforce, HubSpot, etc.) to Chat Data conversations
- Update lead information after purchases, form submissions, or user actions
- Link conversations across channels (web, WhatsApp, Messenger) to the same customer record

**Key Features:**

- Searches for existing leads using all provided fields (best-effort matching)
- Creates new leads automatically if no match is found
- UUID cannot be changed once a lead exists
- Supports fields: email, phone, name, psid, ig_username, username, uuid

---

## Enhanced: GET /get-customers

The /get-customers endpoint now supports filtering and includes leadId in responses.

**New Query Parameters:**

- leadId - Filter by lead identifier
- email - Filter by email address
- phone - Filter by phone number
- name - Filter by customer name
- psid - Filter by Page Scoped User ID
- ig_username - Filter by Instagram username
- username - Filter by username

**Example Request:**

```bash
GET /api/v2/get-customers/chatbot_789?email=customer@example.com
Authorization: Bearer YOUR_API_KEY
```

**Example Response:**

```json
{
  "status": "success",
  "customers": [
    {
      "leadId": "550e8400-e29b-41d4-a716-446655440000",
      "name": "John Doe",
      "email": "customer@example.com",
      "phone": "+1234567890",
      "source": "widget",
      "createdAt": 1696118400000
    }
  ],
  "total": 1
}
```

**Common Use Cases:**

- Look up specific customers by email or phone number
- Filter customers by source and other attributes for reporting
- Use leadId to cross-reference customers across different API endpoints

---

## Getting Started

**Step 1:** Get your API key from Account Settings → API Keys or Chatbot Settings → API Keys

**Step 2:** Try the endpoints

```bash
# Attach a lead to a conversation
curl -X POST https://api.chat-data.com/api/v2/attach-lead \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "conversationId": "your_conversation_id",
    "chatbotId": "your_chatbot_id",
    "email": "customer@example.com",
    "name": "John Doe"
  }'

# Filter customers by email
curl -X GET "https://api.chat-data.com/api/v2/get-customers/your_chatbot_id?email=customer@example.com" \
  -H "Authorization: Bearer YOUR_API_KEY"
```

---

## Documentation

For complete API specifications, request/response schemas, and error handling details, see:

- [API Reference - /attach-lead](/api-reference#tag/Conversation-Operations/operation/AttachLead)
- [API Reference - /get-customers](/api-reference#tag/Chatbot-Analytics/operation/GetCustomers)

Available now for all plans with API access. See [PR #1281](https://github.com/chat-data/chat-data/pull/1281) for technical implementation details.

New API Endpoint: Attach Lead & Enhanced Customer Filtering

export const article = {
  author: 'Samuel Su',
  date: '2025-09-30',
  title: `Claude 4.5 Sonnet Now Available`,
  avatar: '/images/samuel_su_profile.png',
  hashtags: ['Claude 4.5 Sonnet', 'AI Models', 'Anthropic', 'Technical Support', 'Finance', 'Cybersecurity'],
  thumbnail,
  preview: `We're thrilled to announce Claude 4.5 Sonnet on Chat Data. Anthropic's most advanced Sonnet model brings specialized domain expertise for technical, financial, and security-focused chatbots.

**Enhanced Domain Knowledge:**
- **Coding & Technical Support**: Developer documentation, troubleshooting, and technical guidance
- **Finance & Banking**: Complex financial inquiries and investment support
- **Cybersecurity**: Security advisory, compliance information, and best practices

Perfect for chatbots serving developers, fintech, cybersecurity, and complex technical support. Available now at 4 credits per message.`,
};


We're thrilled to announce the availability of Anthropic's Claude 4.5 Sonnet on Chat Data. This model brings specialized domain expertise and enhanced accuracy to your chatbots, making it ideal for technical, financial, and security-focused customer support.

## Enhanced Domain Expertise

Claude 4.5 Sonnet excels in specialized domains with deep knowledge in:

### Coding & Technical Support
- **Developer Documentation**: Accurately explain complex APIs, SDKs, and technical concepts
- **Code Troubleshooting**: Help customers debug issues and understand error messages
- **Technical Product Support**: Provide detailed technical guidance and best practices

### Finance & Banking
- **Financial Services Support**: Handle complex financial inquiries with accuracy
- **Investment Queries**: Provide informed responses about financial products and markets
- **Banking Assistance**: Support customers with account-related technical questions

### Cybersecurity
- **Security Advisory**: Answer security-related questions with expertise
- **Compliance Information**: Help customers understand security compliance requirements
- **Threat Awareness**: Provide informed guidance on security best practices

## Key Improvements

- **More Accurate Domain Responses**: Highest precision for specialized technical and financial queries
- **Better Complex Query Handling**: Superior performance on multi-step customer questions
- **Enhanced Problem Solving**: Deeper reasoning for detailed troubleshooting scenarios
- **Longer Conversation Context**: Maintains accuracy and detail across extended support sessions

## Ideal For

Claude 4.5 Sonnet is perfect for chatbots serving:
- Developer communities and technical documentation
- Financial services and fintech companies
- Cybersecurity products and services
- Complex technical product support

Claude 4.5 Sonnet is now available in your Chat Data model selector at 4 credits per message.

Claude 4.5 Sonnet Now Available

export const article = {
  author: 'Samuel Su',
  date: '2025-08-29',
  title: `OpenAI GPT-Realtime Now Available with 25% Lower Pricing`,
  avatar: '/images/samuel_su_profile.png',
  hashtags: ['GPT-Realtime', 'Voice AI', 'OpenAI', 'Price Reduction'],
  thumbnail,
  preview: `OpenAI's production-ready GPT-Realtime model is now available on Chat Data with 25% credit cost reduction, 82.8% accuracy (26% improvement), and new natural-sounding voices. All agents using gpt-4o-realtime-preview have been automatically upgraded - no action required.`,
};


We're thrilled to announce the integration of OpenAI's production-ready GPT-Realtime model into Chat Data! This advanced voice AI model delivers superior performance, more natural conversations, and comes with a 25% price reduction.

## What's New

OpenAI's GPT-Realtime represents a significant upgrade from the preview version with enhanced accuracy, stability, and voice quality. The model achieves **82.8% accuracy** on benchmarks compared to 65.6% for the previous version.

## Key Improvements

- **More Natural Voices**: Two new voices - Cedar and Marin - deliver more expressive and natural-sounding speech
- **Better Accuracy**: 26% improvement in benchmark performance for more reliable responses  
- **Enhanced Stability**: Production-ready reliability with better instruction following
- **Multilingual Support**: Improved handling of multiple languages and accents

## 25% Price Reduction

We're passing on the savings! GPT-Realtime is 25% cheaper than the previous model:
- Text input: 0.75 credits per 1k tokens (was 1 credit)
- Text output: 3 credits per 1k tokens (was 4 credits)
- Audio input: 6 credits per 1k tokens (was 8 credits)
- Audio output: 12 credits per 1k tokens (was 16 credits)

## Automatic Migration

**No action required!** All AI agents previously using gpt-4o-realtime-preview have been automatically upgraded to GPT-Realtime. Agents using gpt-4o-mini-realtime-preview will remain unchanged as there's no mini version available yet.

Changelog

Create Chatbots with your data

In just a few minutes, you can craft a customized AI representative tailored to yourself or your company.