Disability-Assist

Disability-Assist is an open-source, cloud-first assistive communication and environmental-control system intended to help people with speech, cognitive, sensory, developmental, and age-related communication barriers communicate more effectively and control parts of their environment with less frustration.

This repository is structured around a pilot-first approach for small deployments in private homes and assisted-home settings. The initial target is a 10–20 unit pilot that prioritizes real-world usability, dignity, caregiver coordination, and strong privacy boundaries over unnecessary feature sprawl.

What the project does

Disability-Assist is intended to:

• interpret incomplete, degraded, non-standard, or substituted speech;
• infer likely intent using cloud speech, language, and retrieval systems;
• provide intelligible voice output tuned to user hearing and sensory preferences;
• help caregivers understand likely meaning when direct speech is difficult to interpret;
• control supported smart-home and media devices through a constrained action layer;
• learn user-specific routines, vocabulary, and preferences over time;
• optionally support de-identified community learning and research participation.

What the project is not

This project is not positioned as:

• a diagnostic product;
• a treatment device;
• a clinical monitoring platform;
• an emergency response system;
• a substitute for trained medical or professional judgment.

Those boundaries are intentional. The system is designed as an assistive communication and accessibility platform, not as a clinical product.

Delivery model

The project is:

• open source at the code and documentation level;
• intended for public-benefit, community, research, and pilot use;
• capable of cost-recovery hosted services where needed for cloud processing, caregiver connectivity, and governance functions.

The project is not written around a universal free-service assumption. Cloud speech, language, storage, and mobile access introduce real operating costs. The PRDs are written to support a sustainable pilot and a realistic long-term operating model.

Version model

v1 Pilot

v1 is the first fully scoped pilot release and includes:

• cloud-first STT, LLM, and TTS;
• local device action routing and fail-safe behaviour;
• caregiver mobile app;
• consent management;
• smart-home and media control;
• local personalization with optional cloud-linked services;
• role-based access controls;
• private-home and assisted-home deployment modes;
• de-identified community-learning and research opt-in pathways.

v2 Shared Intelligence

v2 expands the pilot foundation with:

• cross-user assistive pattern seeding;
• stronger retrieval and condition-scoped priors;
• facility and fleet management capabilities;
• advanced routines and automation templates;
• researcher workflow tooling;
• stronger governance, analytics, and update controls.

Guiding principles

• dignity first;
• cloud-first accuracy in v1;
• minimum necessary disclosure;
• predictable behaviour for sensory-sensitive users;
• separate consent lanes for separate uses;
• strong caregiver and facility role boundaries;
• no false claims;
• public documentation and open governance where practical.

Repo reading order

Start here:

1. docs/PRD-00-Index.md
2. docs/PRD-SUMMARY.md
3. docs/PRD-01-Overview-and-Vision.md
4. roadmap.md

Then read the supporting PRDs in numerical order.

Core documents

• docs/PRD-00-Index.md — package entry point
• docs/PRD-SUMMARY.md — executive summary of the full PRD suite
• docs/PRD-01-Overview-and-Vision.md — product framing, scope, and non-goals
• docs/PRD-02-User-Personas-and-Journeys.md — user groups and accessibility scenarios
• docs/PRD-03-System-Architecture.md — system design and deployment topology
• docs/PRD-04-Voice-Input-and-STT.md — cloud speech input path and controls
• docs/PRD-05-LLM-Intent-Engine.md — intent inference and action safety
• docs/PRD-06-Adaptive-Voice-Output.md — voice output tuning
• docs/PRD-07-Smart-Home-Integration.md — device and service integration
• docs/PRD-08-Personalization-and-Memory.md — user memory and assistive profile handling
• docs/PRD-09-Caregiver-Communication-Bridge.md — caregiver-facing interpretation support
• docs/PRD-10-Hardware-Reference-Design.md — pilot hardware guidance
• docs/PRD-11-Privacy-Security-and-Ethics.md — governance and privacy boundaries
• docs/PRD-12-Open-Source-Contribution-Guide.md — contribution posture and operating rules
• docs/PRD-13-Cloud-RAG-and-Cross-User-Learning.md — shared retrieval and assistive priors
• docs/PRD-14-Caregiver-Mobile-App.md — mobile app requirements
• roadmap.md — implementation roadmap by phase

Recommended implementation posture

Build the project in this order:

• establish governance and consent flows first;
• stand up the cloud speech and intent path;
• build local action routing and smart-home control;
• add caregiver app and role-based access;
• add personalization and retrieval;
• add de-identified contribution workflows only after the core pilot is stable.

Pilot success criteria

The pilot should prove five things:

• the system improves real-world understanding for target users;
• caregivers receive useful help without unacceptable surveillance;
• the cloud architecture is accurate enough to justify cost;
• deployment in home and assisted-home settings is operationally manageable;
• the project can scale governance before it scales user count.

Contribution expectations

This project should not become a generic AI assistant repo. Contributions must support the documented assistive goals, privacy model, consent boundaries, safety rules, and deployment reality defined in the PRDs.