Aera: Your Digital Twin
Separate Personal and Work Twins (Inspired from Severance)
Chat with your colleague's twin when they are unavailable

Aera — A Design-First Digital Twin Platform

Inspiration

In a world full of AI “integrators,” most products feel the same:
a chat interface bolted onto a pile of APIs.

They’re powerful on paper, overwhelming in practice, and difficult to trust.

Aera is different.

Aera is a design-first digital twin platform built on the belief that
if an AI represents you, it must feel calm, intentional, and safe.

Instead of exposing raw data streams from Gmail, calendars, GitHub, or Slack, Aera converts them into a human-centric interface organized around how people actually think—time, priorities, availability, and focus. UI and UX are not cosmetic layers in Aera; they are privacy and safety mechanisms.

Where other integrators blur personal and professional boundaries, Aera enforces separation through personas, scoped views, and visible trust indicators. Where others ask users to trust the system, Aera shows boundaries directly in the interface—so users always know what data is in scope and which model is acting on their behalf.

This philosophy enables something fundamentally new:
controlled, intentional sharing of a digital twin.

Colleagues can query your availability without seeing your emails.
Teammates can plan meetings without accessing private context.
AI assistance becomes useful without being invasive.

What Aera Does

Aera is an intelligent digital twin that:

Aggregates Live Digital Context
Securely connects to Gmail, Google Calendar, GitHub, and other services to construct a real-time understanding of your digital life.
Uses an Agentic AI Core
Powered by GPT-4o through Keywords.ai, Aera dynamically reasons over context, decides when tools are required, and produces grounded, explainable answers.
Supports Dual Personas
Personal and Work personas operate with separate context windows, tools, and prompt strategies, preventing accidental leakage across domains.
Delivers Smart Mini-Apps
Daily summaries, conflict detection, urgent task surfacing, and focus insights—optimized for glanceable clarity.
Enables Twin Discovery
Search and interact with other people’s digital twins using strict, permissioned visibility rules.
Prioritizes Privacy by Design
When viewing another twin, only explicitly shared context is accessible. Emails, repositories, and private memories remain invisible by default.

Architecture Overview

Aera is built with a local-first, agent-centric architecture, with Keywords.ai acting as the LLM control plane.

Core Layers

Frontend
- Vanilla JavaScript SPA
- Glassmorphism UI inspired by Apple design
- Visual indicators for persona, scope, and privacy boundaries
Backend (FastAPI)
- OAuth 2.0 orchestration for Google & GitHub
- Context assembly and tool routing
- Persona-aware request handling
Agent Layer
- Context Engine (emails, calendar, memories)
- Tool Registry (research, scheduling, GitHub queries)
- Reasoning loop powered by GPT-4o via Keywords.ai
LLM Gateway (Keywords.ai)
- Unified API across providers
- Prompt versioning and experimentation
- Model routing, fallback, and observability

Why Keywords.ai Is Central to Aera

Keywords.ai is not just an API wrapper in Aera — it is foundational infrastructure.

1. Unified LLM Gateway

Aera routes all LLM traffic through Keywords.ai, decoupling the agent logic from any single provider.

This allows:

Instant model swaps (GPT-4o → Claude → Llama)
Cost optimization without code changes
Graceful fallback during outages

This is critical for a digital twin, where availability and consistency matter more than raw benchmarks.

2. Prompt Versioning & Safety

Aera avoids hard-coded prompts entirely.

Using Keywords.ai:

System prompts are versioned and tracked
Prompt changes can be rolled back instantly
Different personas can map to different prompt strategies

This is especially important when prompts define:

what data is allowed in context
how aggressively tools are used
how privacy boundaries are enforced

In Aera, prompt changes are treated like production deploys, not experiments.

3. Observability for Agent Behavior

Digital twins must be auditable.

Keywords.ai gives Aera:

Visibility into prompt → tool → response chains
Token usage and cost tracking per persona
Debuggability when the agent makes unexpected decisions

This enables rapid iteration without compromising user trust.

4. Enterprise Readiness

Because Keywords.ai abstracts provider-specific logic, Aera can support:

On-device models
Private cloud deployments
Enterprise-specific LLM backends

This aligns with Aera’s long-term goal:
team-wide and organization-wide digital twin networks.

Agent Workflow (With Keywords in the Loop)

User asks:

“What do I need to prepare for my meeting with Sarah?”

Context Assembly
- Fetch upcoming calendar events
- Pull recent relevant emails
- Filter based on active persona (Work)
Prompt Resolution
- Keywords.ai injects the correct prompt version
- Persona-specific rules define scope and tone
Reasoning
- GPT-4o reasons over injected context
- Determines no external tool calls are required
Response
- Aera responds instantly with a grounded, contextual answer
- No unnecessary API calls, no hallucination risk

This pattern makes Aera fast, predictable, and explainable.

Accomplishments We’re Proud Of

Design as a Privacy Boundary
UI elements explicitly communicate scope, ownership, and persona.
Agentic AI Without Opacity
Tool calls, prompts, and models are observable and controlled.
Keywords-Powered Scalability
Model choice, experimentation, and deployment are infrastructure decisions—not rewrites.
Local-First Trust Model
Sensitive data is fetched on demand and never blindly shipped to the cloud.