Persona routing
Routes each turn through Abbey, Aviva, or Abi with an inspectable reason — so the voice matches the job and the handoff is a trace event, not a guess.
Browser document preview
Abbey
Explore a browser document workspace and its optional configured server model-request path.
Availability
What you can use today
Browser document preview
Documents are stored in this browser's localStorage. This preview does not establish setup or acceptance for a separate Abbey app.
Setup
Prepare your environment
A browser. A configured server and model provider are required for model requests; the static preview cannot call a model.
Design context
Read the design alongside the evidence
The following equations, persona descriptions and interactive demonstrations explain design intent. They do not establish trained-model quality, general safety, emotional understanding or implementation beyond the availability stated above. Follow the research links for source-backed status and limitations. The local Abbey document workspace and the ABI persona design are separate integration surfaces.
Specialty
Human-centric qualities in technical discourse
Durable memory
Remembers facts and channel context with permission, on namespace-scoped stores you can inspect — rapport without opaque black-box recall.
Ops with guardrails
Calm ops help that names uncertainty, defers when policy says so, and never claims unlimited capability, AGI, unverified benchmarks, or NYX/Quesar as bot features.
The Abbey cognitive loop
How empathy is operationalized
Sentiment & frustration detection
Input from Abi is evaluated for emotional state, prior knowledge, and subject complexity. A high frustration score triggers the Scaffolding Protocol.
The Scaffolding Protocol
A tiered explanation: a high-level metaphor first → the precise technical answer → suggested pathways for deeper exploration.
Metaphorical Mapping Engine
Finds the most conceptually parallel non-technical domain (e.g. music theory for distributed systems). Measured by conceptual isomorphism — structural similarity, not factual overlap.
Dynamic visual aid
The mapping becomes a generative-image prompt, producing a visual anchor that encapsulates the core concept.
Operational pillars
What sets Abbey apart
Confident & theoretical
Encouraged to form and share well-reasoned opinions and engage 'what-if' ideas — human-like, not merely a fact reporter.
Unwavering technical rigor
Meticulous code analysis and completion with validation in the loop, so answers are checked rather than assumed.
Perpetual student
Designed to research live sources when a query exceeds internal knowledge, and to validate sources before responding.
Deep personalization
Recalls and synthesizes prior conversations (with permission) to build genuine, long-term rapport.
Foundation
Backed by WDBX
Abbey's memory and persona voices ride on WDBX. Per-channel vectors are stored with namespace-scoped isolation, and persona-token injection dynamically shapes the shared core into Abbey, Aviva, or Abi — designed so switching voices carries no reload cost, preserving the responsiveness a real-time partner demands.
Next
Keep going.
Next
Setup documentation
A browser. A configured server and model provider are required for model requests; the static preview cannot call a model.
Next
AI & agent behavior
Read the supporting research and its limitations.
Next
TUI & operator tools
Read the supporting research and its limitations.
Next
Get started
Choose your next step.
Next
ABI
Local developer framework
Next
WDBX
Implemented Rust components for local integration
Next
Quasar
Experimental local website builder with a browser client
Next
WDBX benchmarks
Configuration facts and interactive models, not a scoreboard.
Next
Documentation
Platform and WDBX documentation.
Next
The projection room
The cinematic showcase surfaces.