Continuity Core (C2)

Continuity Core is a local-first cognitive memory engine and MCP server. It unifies a biomimetic, multi-store memory architecture with the Manifold Resonance Architecture (MRA) for directed curiosity.

This repository is a fresh implementation that focuses on:

• Nested learning and gated plasticity for consolidation
• Strategic forgetting via decay and pruning
• A tiered memory model (working, episodic, semantic, procedural, autobiographical, goals)
• An append-only Event Log as the source of truth
• MRA stress and void detection for targeted inquiry

Core Concepts

• Event Log: Append-only journal of all interactions and internal transitions.
• Memory Stores: Specialized stores with distinct decay and access patterns.
• Context Composer: Budgeted expected-utility selection of prompt context.
• MRA: Multi-perspective dissonance scoring + conceptual void detection.
• Credit Assignment: Reinforces memories that improve outcomes.

MVP MCP Tools

• c2.write_event: Append to the Event Log and trigger indexing.
• c2.context: Return a prompt pack from the Context Composer.
• c2.introspect: Run MRA stress/void scans and return inquiry candidates.

Quickstart (Local)

1. Start dependencies:

docker compose up -d

2. Create a virtual environment:

python -m venv .venv
.venv\Scripts\activate

3. Install the package (editable):

pip install -e .

4. Run the MCP server (stdio):

python -m continuity_core.mcp.server

5. Ingest local notes (optional):

c2-ingest . --ext .md .txt

Project Layout

continuity_core/
  config.py
  event_log.py
  context/
  graph/
  ingest/
  memory/
  mra/
  mcp/
  services/

License

MIT. See LICENSE.