PCAS (Personal Central AI System)

PCAS is an open-source, local-first, intelligent decision-making engine designed to power a new generation of personal AI operating systems.

It serves as the core technical heart of the DreamHub Ecosystem, a broader vision for a user-centric AI future built on the principle of "Absolute Data Sovereignty, Flexible Compute Scheduling."

This project is currently facing some challenges, likely due to considerations regarding product positioning and practicality. Development has been temporarily put on hold until I identify a clear path forward.

---

📖 What is PCAS?

PCAS is not a user-facing application. It is a deployable software engine that you run in your private environment (e.g., your PC or home server). Its sole purpose is to act as a secure and intelligent "decision-making center" for your digital life.

It connects to various applications and services (we call them D-Apps) through an Intelligent Event Bus, allowing you to create powerful, automated workflows while ensuring your data never leaves your control.

To delve deeper into the philosophy and technicals, please read the PCAS Whitepaper and the PCAS Architecture Overview.

✨ Core Features

• 🛡️ Absolute Data Sovereignty: PCAS and your data run in your private environment. You have full control. Period.
• 🎛️ Flexible Compute Modes: Through a built-in "Policy Engine," you decide how AI tasks are processed:
  • Local Mode: Maximum privacy with local AI models.
  • Hybrid Mode: The perfect balance of privacy and performance.
  • Cloud Mode: Maximum power using cloud AI APIs.
• 🤖 Intelligent Decision-Making: PCAS acts as your "Personal Decision Center," understanding your intent and coordinating D-Apps to get things done.
• 🧩 Open D-App Ecosystem: The event bus architecture allows any service to be integrated as a D-App.
• 🚀 Foundation for Personal AI: PCAS is designed to be a "Data Crucible," helping you build a private dataset to fine-tune your own personal AI models.
• 🌐 Open Standard & Community: We aim for PCAS to become an open standard for a new pattern of personal AI.

🏛️ Architecture

PCAS is the central hub in a mesh-like, event-driven network of D-Apps.

graph TD
    subgraph "The World"
        DApp1[D-App: Communicator]
        DApp2[D-App: Scheduler]
        DAppN[More D-Apps...]
    end

    subgraph "Your Private Environment"
        PCAS_Core[PCAS Engine]
    end

    %% Communication Flow
    DApp1 <--> |Events/Commands via Secure Bus| PCAS_Core
    DApp2 <--> |Events/Commands via Secure Bus| PCAS_Core
    DAppN <--> |Events/Commands via Secure Bus| PCAS_Core

    style PCAS_Core fill:#cde4ff,stroke:#36c,stroke-width:3px

Loading

🚀 Quick Start

Get PCAS up and running in minutes with our modern development workflow.

Prerequisites

• Go 1.24+
• Docker and Docker Compose
• An OpenAI API key (OpenAI or OpenRouter key)

1. Clone and Setup

git clone https://github.com/soaringjerry/pcas.git
cd pcas

2. Configure Environment

Create a .env file in the project root:

# Required: Your OpenAI API key
OPENAI_API_KEY=sk-your-api-key-here

# Optional: Customize ports if needed
PCAS_PORT=50051
POSTGRES_PORT=5432

3. Start Everything

Use our all-in-one command:

make dev-up

This will:

• Start PostgreSQL with pgvector extension
• Initialize the database schema
• Build the PCAS binaries
• Start the PCAS server with your configuration

4. Try It Out

Create a memory with user identity:

./bin/pcasctl emit --type pcas.memory.create.v1 \
  --user-id alice \
  --subject "My favorite programming language is Go"

Ask a question (with RAG-enhanced response):

./bin/pcasctl emit --type pcas.user.prompt.v1 \
  --user-id alice \
  --data '{"prompt": "What is my favorite programming language?"}'

Search memories semantically:

./bin/pcasctl search "programming preferences"

5. Explore Advanced Examples

Check out our Multi-AI Chatbot example to see PCAS's multi-identity capabilities in action:

cd examples/multi-ai-chatbot
go run main.go --user-id alice

🛑 Stopping PCAS

make dev-down

This cleanly shuts down all services and preserves your data.

Alternative: CLI Guided Setup

Prefer a lightweight, interactive flow without Compose?

• Run guided setup: pcasctl setup
  • Generates/updates .env and policy.yaml, and can start the server.
• Start server manually: pcas serve --host 0.0.0.0 --port 50051
• Send a test event: pcasctl emit --type pcas.test.echo.v1 --data "{\"prompt\":\"hello\"}"
• One‑click Docker install/update pulls newest build by default (edge channel). Use --channel stable to pin releases.

Using OpenRouter instead of OpenAI:

• Set env: OPENAI_BASE_URL=https://openrouter.ai/api/v1 and use an OpenRouter API key in OPENAI_API_KEY.
• Optional headers (recommended by OpenRouter): OPENROUTER_SITE_URL=https://your.site and OPENROUTER_APP_NAME=PCAS.
• Auto-detect: if your OPENAI_API_KEY starts with sk-or-, PCAS will default to https://openrouter.ai/api/v1 unless OPENAI_BASE_URL is explicitly set.

🤝 Community & Contribution

PCAS is an open-source project driven by the community. We sincerely invite you to join us.

• Join the discussion: [Discord Link TBD]
• Contribute: Please read our Contribution Guide.
• Documentation: Explore our comprehensive Documentation Portal.
• Report issues: Please use the Issues section.

📄 License

PCAS is open-sourced under the MIT License.

简体中文

PCAS (Personal Central AI System)

PCAS is an open-source, local-first, intelligent decision-making engine designed to power a new generation of personal AI operating systems.

It serves as the core technical heart of the DreamHub Ecosystem, a broader vision for a user-centric AI future built on the principle of "Absolute Data Sovereignty, Flexible Compute Scheduling."

---

📖 What is PCAS?

PCAS is not a user-facing application. It is a deployable software engine that you run in your private environment (e.g., your PC or home server). Its sole purpose is to act as a secure and intelligent "decision-making center" for your digital life.

It connects to various applications and services (we call them D-Apps) through an Intelligent Event Bus, allowing you to create powerful, automated workflows while ensuring your data never leaves your control.

To delve deeper into the philosophy and technicals, please read the PCAS Whitepaper and the PCAS Technical Plan.

✨ Core Features

• 🛡️ Absolute Data Sovereignty: PCAS and your data run in your private environment. You have full control. Period.
• 🎛️ Flexible Compute Modes: Through a built-in "Policy Engine," you decide how AI tasks are processed:
  • Local Mode: Maximum privacy with local AI models.
  • Hybrid Mode: The perfect balance of privacy and performance.
  • Cloud Mode: Maximum power using cloud AI APIs.
• 🤖 Intelligent Decision-Making: PCAS acts as your "Personal Decision Center," understanding your intent and coordinating D-Apps to get things done.
• 🧩 Open D-App Ecosystem: The event bus architecture allows any service to be integrated as a D-App.
• 🚀 Foundation for Personal AI: PCAS is designed to be a "Data Crucible," helping you build a private dataset to fine-tune your own personal AI models.
• 🌐 Open Standard & Community: We aim for PCAS to become an open standard for a new pattern of personal AI.

🏛️ Architecture

PCAS is the central hub in a mesh-like, event-driven network of D-Apps.

graph TD
    subgraph "The World"
        DApp1[D-App: Communicator]
        DApp2[D-App: Scheduler]
        DAppN[More D-Apps...]
    end

    subgraph "Your Private Environment"
        PCAS_Core[PCAS Engine]
    end

    %% Communication Flow
    DApp1 <--> |Events/Commands via Secure Bus| PCAS_Core
    DApp2 <--> |Events/Commands via Secure Bus| PCAS_Core
    DAppN <--> |Events/Commands via Secure Bus| PCAS_Core

    style PCAS_Core fill:#cde4ff,stroke:#36c,stroke-width:3px

Loading

🚀 Getting Started (Actionable Roadmap)

The project is in active development based on the following roadmap. Our core principle: "Run one chain, then go upstairs."

ETA	Target	Key Deliverables
+2 Weeks	Minimal Event Bus & CLI	pcas serve, pcas emit commands; trace_id.
+1 Month	Policy v0 + Providers	policy.yaml static rules; OpenAIProvider & MockLocalProvider.
+2 Months	Explainable AI + Graph Store	LLM-decide() with decision logs; SQLite persistence.
+3 Months	SDK & Example D-Apps	Go/TS SDK; Scheduler, Communicator, Knowledge D-Apps.
+4 Months	Preview Release & Community	GitHub Beta Tag; Docs Site; RFC process.

🧰 Quick Start

• One-time setup: pcasctl setup
  • Guides environment (.env), generates/updates policy.yaml, and can start the server.
• Start server manually: pcas serve --host 0.0.0.0 --port 50051
• Send a test event: pcasctl emit --type pcas.test.echo.v1 --data "{\"prompt\":\"hello\"}"
• Update binaries/images: pcasctl update (rebuild from source and/or pull Docker image)

🤝 Community & Contribution

PCAS is an open-source project driven by the community. We sincerely invite you to join us.

• Join the discussion: [Discord Link TBD]
• Contribute: Please read our CONTRIBUTING.md (TBD).
• Report issues: Please use the Issues section.

📄 License

PCAS is open-sourced under the MIT License.

6c57eb8 (feat(cli): add interactive setup/update wizards (pcasctl setup, pcasctl update))