Welcome
Build AI characters that see, hear, and remember.
Estuary is a platform for creating conversational AI agents with persistent memory, real-time voice, and spatial awareness -- deployed across smartglasses, mobile, and web.
Why Estuary?
Most conversational AI treats every session as a blank slate. Estuary doesn't.
Create AI characters that remember users across conversations spanning months. They understand physical environments through world models and respond in real-time. And they scale from prototype to thousands of concurrent users without rearchitecting.
Core Capabilities
| Capability | What it means |
|---|---|
| Persistent Memory | Characters recall past conversations, user preferences, and relationships over time |
| Spatial Awareness | World models for object detection, depth estimation, and scene understanding |
| Real-time Voice | Sub-second voice interactions with Deepgram STT and ElevenLabs TTS |
| Cross-platform | Build your character once, deploy everywhere: Lens Studio, mobile, and web with no additional configuration |
| Action Triggers | AI responses can invoke in-app events, animations, or external integrations |
| Multi-tenant Scale | One character definition serves hundreds of concurrent end users with isolated sessions |
Quick Overview
┌──────────────────────────────────────────────────────────────────────┐
│ Your Application │
├──────────────┬───────────────┬─────────────┬────────────────────────┤
│ Web │ Lens Studio │ Unity │ REST API / cURL │
│ SDK │ SDK │ SDK │ │
└──────┬───────┴───────┬───────┴──────┬──────┴───────────┬────────────┘
│ │ │ │
▼ ▼ ▼ ▼
┌──────────────────────────────────────────────────────────────────────┐
│ Estuary Platform │
│ ┌──────────────┐ ┌───────────┐ ┌───────────────────┐ │
│ │ Gateway │ │ Workers │ │ Memory + Knowledge│ │
│ │ (Socket.IO + │ │ STT→LLM │ │ Graph (PostgreSQL │ │
│ │ REST API) │ │ →TTS │ │ + pgvector) │ │
│ └──────────────┘ └───────────┘ └───────────────────┘ │
└──────────────────────────────────────────────────────────────────────┘
Getting Started
Create a character and start a conversation in under 5 minutes.
Manage characters, conversations, and memory programmatically.
Build web-based voice and text experiences.
Build conversational AI for Snap Spectacles.
Build AI characters for Unity games and AR/VR apps.
Requirements
Before you begin, ensure you have:
- Estuary Account with an API key (sign up)
Architecture
The platform is organized into layers:
| Layer | Components | Purpose |
|---|---|---|
| Gateway | FastAPI + Socket.IO | WebSocket connections, authentication, audio streaming, REST API |
| Workers | STT, LLM, TTS pipeline | Stateless request processing with streaming responses |
| Memory | PostgreSQL + pgvector | Persistent memory, knowledge graph, entity extraction |
| Infrastructure | Redis, LiveKit | Job queue, pub/sub, WebRTC voice |
See Platform Overview for details.
Next Steps
- Create your first character with the quickstart guide
- Explore the REST API for programmatic access
- Choose an SDK for your target platform
- Learn how memory works across conversations