Open Source
Tools
Six open-source tools for production AI systems. RAGLint checks whether your data is ready. knowledge-ci fails your build when answer quality drops. ISNAD grades the chain a claim travelled through. Agent-Surface makes your system legible to the agents consuming it.
ISNAD
Claim-level provenance for multi-agent knowledge systems. Grades every agent, scraper, and model in a claim's transmission chain. 157 tests. Published paper.
pip install isnad
GitHub → · PyPI → · Full page →
Agent-Surface
Make your website usable by AI agents. Converts a web app into an agent-friendly interface — scans the app, generates a machine-readable manifest, exposes structured actions and workflows.
npm install agent-surface
GitHub →
knowledge-ci
Repository-native RAG with built-in evaluation and continuous testing. Ingests docs, answers grounded questions with citations, and fails CI when answer quality degrades.
pip install knowledge-ci
GitHub →
RAGLint
CLI-first tool that scans arbitrary data and emits an AI Data Readiness Report for RAG systems. Tells you whether your data is ready before you build.
pip install raglint
GitHub →
Magnific
Config-driven, resumable, multi-stage media pipeline. A creative idea plus two reference images becomes per-scene MP4 clips. Isolated jobs on disk; stages communicate through manifests and shared config.
pip install magnific
GitHub →
29 public repositories total. Older projects worth surfacing: isharay (Ulster project), IslamAI (Imagine Cup entry), Vehicle-Speed-NumberPlate-Detection (the viral project), hand-paint, Advent-of-code.