CogitareLink N3-EYE

Demo project that integrates the EYE (eyereasoner) WASM reasoner into CogitareLink’s reason_over pipeline, enabling N3 rule execution alongside SHACL and SPARQL.

Prerequisites

• Python 3.11 or higher
• Node.js 16 or higher

Installation

Clone the repo and enter its directory:

git clone https://github.com/LA3D/n3-eye.git
cd n3-eye

Create and activate a virtual environment, then install Python dependencies:

python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install .

Install Node dependencies for the EYE runner:

npm install

Usage

Verify that patches are applied and the Node runner is available:

python main.py

Run the demo agent (set OPENAI_API_KEY for LLM-driven flow; otherwise it falls back to a direct EYE invocation):

export OPENAI_API_KEY=<YOUR_API_KEY>
python demo_agent.py

Testing

Run the test suite:

pytest -q

Project Structure

• patch_reason_sandbox.py: monkey-patch reason_over to support N3/EYE rules
• patch_reason_tool.py: extends the OpenAI function bridge and patches reason_tool
• demo_agent.py: Cosette-based demo agent demonstrating both LLM and direct tool invocation
• main.py: CLI script to verify patch application
• js/eye-runner.mjs: Node.js runner for EYE WASM
• tests/: unit tests covering sandbox and tool patches

Next Steps

• Integrate SHACL validation and SPARQL CONSTRUCT in the unified pipeline
• Enhance natural-language summaries and system prompts
• Capture and expose proof objects in provenance for auditing
• Provide an interactive UI or REST API for end users