Skip to content

v0.1.0 — Cognitive Bridge

Latest
Latest

Choose a tag to compare

View all tags
@JosephOIbrahim JosephOIbrahim released this 30 Jun 21:05
· 11 commits to main since this release
v0.1.0
510cc1a

Cognitive Bridge v0.1.0

First tagged release. An MCP server that gives AI a compositional mind — persistent memory that catches contradictions, built on USD LIVRPS composition semantics and mechanically verified against them.

Highlights

  • 8 MCP tools spanning the full argumentation lifecycle: assert → detect → steelman → resolve → cascade → decide.
  • Schema-enforced critical thinking — falsifiability, steelman, and decision trade-offs are Pydantic validators, not prompt suggestions. The model cannot skip them.
  • Four-layer conflict detection — structural, semantic (vectorised embeddings), delegated, and cascading through a dependency DAG.
  • Non-destructive composition — nothing is ever deleted; winners are computed dynamically by resolve().
  • Mechanical USD bridge — exports reasoning state to valid .usda files; SQL and USDA resolution are verified identical (check_consistency() returns zero discrepancies).
  • 1317 tests across 49 files, run in CI on every push.

What's inside

  • 8 tools, 8 read-only resources, 3 prompts
  • 5 injection profiles (none → classical) for tuning exploration intensity
  • USD composition bridge: 6 arc sublayers + root stage.usda
  • Runnable examples: basic walkthrough, conflict scenario, MongoDB cascade showcase
  • GitHub Actions CI with a coverage gate

Install

git clone https://github.com/JosephOIbrahim/Cognitive-Bridge.git
cd Cognitive-Bridge
pip install -e ".[all]"
python -m cognitive_bridge.server

Then register it with Claude Desktop or Claude Code — see the README for config.

Requirements

  • Python 3.11+
  • fastmcp, sqlmodel, pydantic (sentence-transformers + numpy optional, for semantic detection)

The README now includes architecture and flow diagrams (Mermaid), the LIVRPS model, and the full tool reference.

Assets 2
Loading