LAJBI (Layered Architecture for Justification-Based Intelligence) is an advanced framework for interacting with and controlling AI. Think of it less like a chat and more like an operating system for AI agents.
It uses a structured, command-based language to move beyond simple conversation and enable complex, reliable, and transparent operations with single or multiple AI agents.
Standard conversational prompts are often ambiguous and lack fine-grained control. LAJBI solves this by introducing a new paradigm built on three core principles:
- 🤖 From Conversation to Control: Replace ambiguous natural language with a clear, declarative command syntax (
<command> KEYWORD:[parameter]) for precise, repeatable results. - 🤝 From a Single Assistant to a Team of Specialists: Define and load a team of specialized AI agents (Personas) who collaborate and debate to solve complex problems.
- 🧠 From a Black Box to a Transparent Mind: Every agent operates on the PABIAM Cognitive Architecture, making its reasoning process transparent and auditable.
- Declarative Control Language: A simple but powerful syntax for controlling AI actions.
- Multi-Agent Collaboration: Orchestrate teams of specialized AI agents who debate and reach a consensus.
- Cognitive Transparency: Inspect any agent's internal "mind" (its beliefs, intentions, and focus) using the PABIAM model.
- Dynamic & Extensible: Load new commands and behaviors in-session with extension modules.
- Systematic Refinement: Built-in tools to evaluate, interpret, and optimize your prompts.
Start a session with the LAJBI framework. The system will load its core modules and signal readiness.
SYS_AI:
LAJBI core ready.
Use the <join> command to add your first specialized agent.
USER:
<join> TITLE: Python Coder NAME: Coder PROMPT: **Role:** You are an expert Python developer...SYS_AI:
Persona 'Coder' has joined the team as 'Python Coder'.
Address the agent directly using the @ prefix to assign a task.
USER:
@Coder write a Python function to find the factorial of a number.Coder:
Of course. Here is a Python function...
This README provides a brief overview. For a complete guide, including the full command reference, architectural breakdown, and advanced examples, please visit our Official Project Wiki.
We welcome contributions to the LAJBI framework! Please see our CONTRIBUTING.md file for guidelines on how to submit issues, propose features, and contribute to the code.
If you find this framework useful, you can show your support by including a "Powered by LAJBI" badge in your project. While not required, it is greatly appreciated! You can find instructions and different badge formats in the ATTRIBUTION.md file.
This project is licensed under the MIT License - see the file for details.