Platform Overview & Architecture

Twig AI is an enterprise-grade Retrieval-Augmented Generation (RAG) platform that enables organizations to build, deploy, and manage intelligent AI agents powered by their own knowledge bases.

What is Twig AI?

Twig AI combines state-of-the-art language models with your organization's data to provide accurate, contextual answers to questions. The platform ingests data from multiple sources, processes it using advanced chunking and embedding techniques, and retrieves relevant information to augment AI responses.

Core Components

1. AI Agents

Configurable AI assistants that can be customized with specific data sources, instructions, and behavior patterns.

2. Data Sources

Connect 14+ types of data sources including Confluence, Slack, Google Drive, SharePoint, websites, and more.

3. RAG Engine

Advanced retrieval system with multiple strategies (Redwood, Cedar, Cypress) for optimal accuracy and performance.

4. Knowledge Base

Automatically generated and curated knowledge base that continuously learns from interactions.

5. Analytics & Monitoring

Comprehensive analytics dashboard to track usage, quality, and performance metrics.

Architecture Overview

┌─────────────┐
│ Data Sources│
└──────┬──────┘
       │ Ingestion
       ▼
┌─────────────────┐
│ Processing Layer│ → Chunking, Embedding, Indexing
└────────┬────────┘
         │
         ▼
┌────────────────┐
│  Vector Store  │ → Pinecone / TigrisDB
└────────┬───────┘
         │
         ▼
┌────────────────┐
│  RAG Strategies│ → Redwood, Cedar, Cypress
└────────┬───────┘
         │
         ▼
┌────────────────┐
│      LLM       │ → GPT-4, GPT-3.5, Claude
└────────┬───────┘
         │
         ▼
┌────────────────┐
│   AI Agents    │
└────────┬───────┘
         │
         ▼
┌────────────────┐
│  Deployments   │ → API, Widget, Extensions
└────────────────┘

Key Features

Enterprise-Grade Security

• SOC 2 Type II compliance

• SSO integration (SAML, OAuth)

• Role-based access control

• Data encryption at rest and in transit

Multiple RAG Strategies

• Redwood: Fast, direct retrieval (~1-2 sec)

• Cedar: Context-aware with prompt rewriting (~2-3 sec)

• Cypress: Advanced with reranking and tier-based retrieval (~3-4 sec)

Flexible Deployment

• REST API for programmatic access

• Browser extensions (Chrome, Firefox)

• Email add-ins (Outlook, Gmail)

• Embeddable widget

• Native integrations (Slack, Zendesk, Help Scout)

Continuous Learning

• Interaction feedback loop

• Automatic KB article generation

• Response quality tracking

• Evaluation framework

Technology Stack

Language Models

• OpenAI GPT-4, GPT-4o, GPT-3.5-turbo

• Anthropic Claude (coming soon)

• Custom model support

Vector Databases

• Pinecone (primary)

• TigrisDB (alternative)

Embedding Models

• OpenAI text-embedding-ada-002

• Custom embedding models

Infrastructure

• Next.js frontend

• Node.js backend

• PostgreSQL database

• AWS S3 for file storage

Use Cases

Customer Support

Empower support teams with instant access to accurate answers from your knowledge base.

Internal Knowledge Management

Enable employees to quickly find information across all company documentation.

Sales Enablement

Provide sales teams with instant access to product information, pricing, and competitive intelligence.

Developer Documentation

Help developers find answers in technical documentation and API references.

Getting Started

Ready to get started? Continue to the Quick Start Guide to create your first AI agent in 5 minutes.