Expose GitHub Copilot through an OpenAI and Anthropic-compatible API, enabling direct usage with Claude Code.

A real-time network traffic and system log analysis pipeline powered by the DeepSeek-Code-V2-Lite-Instruct local LLM, integrated with NVIDIA Morpheus and deployable via Triton Inference Server. Built for lightweight systems (Ryzen 7, RTX 3050 4GB), this project classifies network traffic and logs as malicious or benign using a locally hosted LLM,

Local AI Chat Agent using Microsoft Agent Framework integrated with Docker Model Runner (DMR) locally

SkillOpt replaces manual prompt tweaking with a mathematical feedback loop. A base model runs your tasks against a benchmark dataset while an optimizer model evaluates failures and rewrites the `## Instructions` section — automatically rejecting changes that don't improve the score.

Optimize frozen LLM agents for HotpotQA with reusable natural language skills trained through iterative validation cycles.