SPL-FRAMEWORK

The Subsumption Pattern Learning (SPL) Framework is a hierarchical multi-agent system that converts a collection of autonomous agents into a self-distilling swarm via a shared collective memory. It implements a three-layer architecture (Reactive, Tactical, Deliberative) where simple pattern matches in Layer 1 can suppress more expensive reasoning in Layer 2, and successful patterns are distilled into a Shared State that all agents can leverage. When you run the server, you enable agents to interact with a centralized memory of learned patterns, confidence scores, and provenance, enabling cross-agent learning and more efficient problem solving over time. The system supports demonstrations and experiments through a published framework, including a live demo and a formal description of inhibition signals and pattern distillation, which you can explore via the SPL server’s tooling.

To use the server, start the MCP server process (via the module entry point) and interact with the provided interfaces or APIs that read from and write to the Shared State. Agents will automatically participate in the three-layer processing flow: quick validation in Layer 0, pattern-based matching in Layer 1, and optional foundation-model reasoning in Layer 2 when needed. The shared state exposes the global pattern library, their confidence scores, and provenance, enabling cross-agent learning and progressive performance improvements as patterns are reinforced or decay over time.

Prerequisites:\n- Python 3.8 or newer (recommended 3.8+ as indicated in the README)\n- Git\n- Internet access to install dependencies\n\nInstallation steps:\n1) Clone the repository:\n git clone https://github.com/your-organization/spl_framework.git\n cd spl_framework\n\n2) (Optional) Create a virtual environment:\n python -m venv venv\n # Windows: venv\Scripts\activate \ # macOS/Linux: source venv/bin/activate\n\n3) Install dependencies:\n pip install -r requirements.txt\n # If a setup.py exists, you can also install: pip install -e .\n\n4) Run the SPL Framework server:\n python -m spl_framework\n\n5) Verify startup by visiting the demo endpoint or logs to confirm the server is listening and the Shared State is initialized.\n\nNotes:\n- If you use a containerized deployment, ensure you expose the appropriate ports for your MCP client to connect.\n- Review any environment-specific configuration in a config file or environment variables as described in additional_notes.

Environment and configuration tips:\n- If your deployment uses multiple agents, ensure the Shared State store (P_shared) is accessible with proper synchronization across processes.\n- Patterns are reinforced in the Shared State via M(p) and C(p). Consider tuning η (reinforcement rate) and δ (decay rate) to balance learning speed and stability.\n- Inhibition signals (I1) can suppress the expensive Layer 2 reasoning; adjust θ and α to control when Layer 2 is engaged.\n- Common issues include: missing dependencies in requirements.txt, module import errors (ensure spl_framework is the correct module name), and port binding conflicts.\n- If you plan to run in production, consider process managers (e.g., systemd, gunicorn) and proper logging/monitoring for health checks and performance metrics.