Twinlix Backend Platform
What is Twinlix?
Twinlix is a cloud platform for creating and managing AI-powered voice and chat agents for businesses. Companies connect their knowledge base, configure an AI assistant's personality and behavior, and deploy it across multiple communication channels — phone calls, web chat, Instagram, Facebook Messenger, WhatsApp, and Telegram.
The backend is a RESTful API that powers the entire lifecycle: from onboarding and agent configuration to real-time conversations, billing, and analytics.
Core Capabilities
1. AI Voice & Chat Agents
Create intelligent assistants that can hold natural conversations with customers across voice and text channels.
- Configurable personality — name, greeting style, sign-off messages, language preferences, response length and tone
- Knowledge base integration — agents answer questions based on uploaded documents, web pages, and custom text
- Multi-language support — configure language-specific instructions for global businesses
- Starter questions — pre-defined conversation starters to guide customers
- FAQ system — curated question-answer pairs for precise responses
- Escalation rules — define when and how to hand off to a human operator
2. Omnichannel Deployment
Deploy a single agent across multiple channels simultaneously:
| Channel | Type | Description |
|---|---|---|
| Voice | Phone call | Inbound/outbound calls via phone numbers (Twilio) |
| Web Voice | Browser | Voice widget embedded on websites |
| Web Chat | Browser | Text chat widget for websites |
| Social | Instagram Direct Messages | |
| Messenger | Social | Facebook Messenger |
| Messaging | WhatsApp Business | |
| Telegram | Messaging | Telegram Bot |
Each channel can be independently configured, paused, resumed, or disconnected. Widget appearance (theme color, position, branding) is customizable per channel.
3. Knowledge Base Management
Build the agent's knowledge from multiple source types:
- Document upload — PDF, TXT, DOC, DOCX files (up to 10MB) are parsed, chunked, and vectorized for semantic search
- Web pages — provide a URL, the system scrapes and indexes the content; supports re-scraping on demand
- Text entries — manually authored knowledge articles
- FAQ pairs — structured question-answer entries with categories
Documents go through an automated pipeline: upload → parsing → vectorization → ready. The status is tracked and visible at every stage.
Bulk operations are supported: enable/disable documents in batches, bulk delete, search and filter by type.
4. Outbound Calling & Campaigns
Proactively reach customers at scale:
- Single outbound calls — initiate a call to any number with a custom first message and context
- Scheduled calls — book calls for a future date/time with timezone support
- Bulk campaigns — mass dialing with up to 50,000 contacts per campaign:
- Configurable concurrency (up to 50 simultaneous calls)
- Automatic retries on no-answer, busy, failed, or voicemail
- Start, pause, and cancel campaigns in real-time
- Per-contact status tracking
- Do Not Call (DNC) lists — maintain compliance with bulk import (up to 10,000 numbers) and per-number management
5. Conversation Management
Full visibility into every customer interaction:
- Conversation list — filter by status (new, important, archived), search by content, paginate
- Message history — complete transcript with timestamps, direction (inbound/outbound), and message type (text, image, video, audio, file)
- Labels & tagging — mark conversations as important, archive them, or disable AI for manual handling
- Bulk actions — batch delete conversations
- Real-time messaging — send messages to customers directly from the admin panel
6. Embeddable Widgets
Drop-in widgets for any website with minimal integration effort:
- Voice call button — a single JavaScript snippet adds a "Call" button to any page
- Web chat widget — embed a full chat interface with customizable appearance
- Widget configuration — theme color, position, placeholder text, banner, greeting message
- Public API — no authentication required for end-user interactions; session management is handled automatically
7. Team Management
Multi-user access with role-based permissions:
- Roles: Admin (full access) and Member (read access to most resources)
- Invitations — invite team members by email; they receive an activation link
- Multi-tenant architecture — each business operates in an isolated tenant with its own data, agents, channels, and billing
8. Billing & Subscriptions
Flexible pricing with usage-based metering:
- Pricing tiers — multiple plans with different quotas (call minutes, messages, storage, channels, API calls)
- Trial support — configurable trial period with automatic expiration
- Stripe integration — checkout sessions, payment methods (card), invoices with PDF download
- Usage tracking — real-time dashboard showing consumed vs. available quota per metric
- Usage history — monthly trends over the last 12 months
- Overage billing — automatic calculation of overage costs when quotas are exceeded
- Subscription lifecycle — upgrade, downgrade, cancel (immediate or end-of-period)
9. Integrations
The platform connects to external services:
| Integration | Purpose |
|---|---|
| Twilio | Phone number provisioning, inbound/outbound voice calls |
| ElevenLabs | AI voice synthesis and real-time voice streaming |
| Meta (Facebook/Instagram) | OAuth-based channel connection for Messenger and Instagram DMs |
| Stripe | Payment processing, subscriptions, invoices |
| OpenAI | LLM-powered conversation intelligence |
| n8n | Workflow automation for document processing and scraping |
10. Interactive Setup
A guided onboarding experience:
The platform includes an AI-powered Setup Agent that walks new users through configuration via a conversational interface. Instead of filling forms, users chat with the setup agent to define their business context, pain points, and goals — the system configures the AI assistant accordingly.
Security & Compliance
- JWT authentication with RS256 signing — 1-hour access tokens, 30-day refresh tokens
- Role-based access control (RBAC) — admin vs. member permissions on every endpoint
- Rate limiting — per-endpoint limits to prevent abuse (auth: 10/min, uploads: 5/min, general: 100/min)
- Webhook signature verification — HMAC-SHA256 for Stripe, Meta; HMAC-SHA1 for Twilio
- Quota enforcement — subscription tier limits checked before every resource-consuming operation
- Data retention policies — configurable per-tenant (1-365 days)
- Audit logging — every create/update/delete operation is recorded with user, IP, timestamp, and changes
- Tenant isolation — strict data separation between businesses
API Architecture
- RESTful JSON API with consistent response format:
{ status, data }for success,{ status, message, errors }for failures - Real-time voice via WebSocket connections for low-latency bidirectional audio streaming
- Async processing — document vectorization, web scraping, and call initiation are handled asynchronously with status tracking
- Zod schema validation on all request bodies
- Prometheus metrics endpoint for monitoring
- Health checks with database connectivity verification
Summary
| Metric | Value |
|---|---|
| API endpoints | 120+ |
| Communication channels | 7 types |
| Knowledge base sources | 4 types (documents, web, text, FAQ) |
| Max campaign contacts | 50,000 |
| Max concurrent calls | 50 per campaign |
| DNC bulk import | 10,000 numbers |
| Webhook integrations | 6 (Twilio, Stripe, Meta, ElevenLabs, n8n, PDF parser) |
| WebSocket streams | 2 (voice, ElevenLabs proxy) |
Twinlix provides everything a business needs to deploy, manage, and scale AI-powered customer communication — from a single voice agent handling support calls to enterprise campaigns reaching thousands of contacts across every major messaging platform.