| title | The PCAS Whitepaper: A New Foundation for Personal AI | |||||
|---|---|---|---|---|---|---|
| description | An in-depth whitepaper detailing the philosophy, architecture, risk mitigation, and roadmap of the PCAS project, a local-first intelligent decision-making engine. | |||||
| tags |
|
|||||
| version | 0.1.2 |
The modern digital experience is a paradox: while technology offers unprecedented power, it comes at the cost of fragmentation and the loss of data sovereignty. PCAS (Personal Central AI System) is an open-source project designed to resolve this conflict.
PCAS is not a user-facing application, but a deployable, local-first, intelligent decision-making engine that you run in your private environment. It is the foundational technology powering the DreamHub Ecosystem—a broader vision for a user-centric AI future.
Built on the principle of "Absolute Data Sovereignty, Flexible Compute Scheduling," PCAS acts as a secure "decision center" for your digital life. It orchestrates various applications (D-Apps) via an Intelligent Event Bus and allows you to choose between the privacy of local AI computation and the power of cloud APIs. Our ultimate goal is for PCAS to enable users to build unique, private datasets from their daily interactions, which can then be used to train truly personal AI models.
This whitepaper details the philosophy, architecture, risk mitigation, and roadmap of the PCAS project.
Our digital lives are scattered across countless application silos. More critically, the implicit agreement we've made with technology—exchanging data for convenience—is broken. We've ceded control of our most valuable asset, our data, to opaque cloud services, exposing ourselves to privacy risks and vendor lock-in. Most current AI solutions exacerbate this dependency, they do not solve it.
PCAS offers a new foundation. Instead of a service you subscribe to, PCAS is a software engine you own and control.
The most accurate analogy for PCAS is the "UKVI Decision-Making Centre" for the United Kingdom's visa and immigration services. It is a central hub that ingests information from numerous, disparate sources (D-Apps), uses an intelligent engine to understand context and make complex decisions, and then issues commands to other D-Apps to execute tasks. It is a system designed for reasoning and orchestration within a trusted, private environment.
PCAS's architecture is a mesh-like, event-driven network. This "Intelligent Event Bus" model provides extreme flexibility and scalability, allowing any application or service to be integrated as a D-App.
To capture and correlate all interactions, decisions, and feedback, PCAS employs a unified Graph-based data model. Every key entity (event, decision, command) is a Node, and their causal relationships are Edges. This structure natively supports data lineage ("receipt-based UI") and counterfactual logging, forming the technical core of our "Data Crucible" vision.
Under the core principle of "Absolute Data Sovereignty," PCAS's built-in "Policy Engine" provides unprecedented flexibility in computation:
- Local Mode: For maximum privacy. All AI computations are performed on your local device using frameworks like Ollama.
- Hybrid Mode: For the perfect balance. You set the rules (e.g., based on data sensitivity tags), and PCAS intelligently decides whether to use local models or to send anonymized compute tasks to privacy-conscious cloud APIs (like OpenAI).
- Cloud Mode: For maximum power. Defaults to using cloud APIs for all AI computations.
A mature architecture anticipates risk. We have designed PCAS with the following mitigation strategies:
- Protocol Stability: Adopting Protobuf for schema definition, mandating
versionfields, and ensuring compatibility with CloudEvents v1.0 to guarantee long-term ecosystem stability. - Decision Engine Complexity: Employing a layered decision architecture (Rules Engine + LLM Engine) and mandating explainability logs for all AI-driven decisions to ensure control and auditability.
- Security & Permissions: Implementing a Zero-Trust model where the event bus acts as a "secure bus," enforcing permissions via Capability Tokens for every D-App.
- State Persistence: Offering a pluggable
StorageProviderinterface, with a default SQLite implementation for ease of use and an advanced PostgreSQL option for power users. - Multi-Device Sync: Focusing on a robust single-node experience in V1, with a clear roadmap to introduce dedicated sync services like NATS JetStream in V2.
PCAS is more than software; it's a mission to create an open, thriving ecosystem. We are committed to building a global community and establishing the PCAS architecture and protocols as an open standard for a new generation of user-centric AI.
For V2 and beyond, we are exploring:
- PCAS Federation: Decentralized, secure communication between independent PCAS instances.
- Counterfactual Logs: Enabling the system to learn from decisions it didn't make.
- Local Model Orchestration: Coordinating multiple local AI models for complex, offline, multi-modal tasks.
- "Receipt-based" Data UI: Making data sovereignty a tangible, verifiable user experience.
Following the principle of "Run one chain, then go upstairs," our roadmap will deliver a Preview Release within months, moving through key milestones from a minimal event bus to a fully explainable decision engine with an SDK.
A new era of personal AI, one that places the user back in control, is dawning. PCAS is at the forefront of this movement. We invite you, whether a developer, investor, or future user, to join us in building this exciting future.