Technical OSINT Corpus

A local-first CLI “meeting studio” for orchestrating multi-agent discussions/debates (courtroom/workshop style), using external agents (e.g., Hermes/Codex/Cursor) and/or local model runtimes (Ollama/vLLM) with bilingual CN/EN support.

Systematic evaluation of post-hoc GNN explanation methods (Saliency Attribution, Integrated Gradients, GNNExplainer, LRP) to test whether explanations recover disease-relevant topological structure (topological signature of disease-associated hubs) in breast cancer molecular graphs built from RNA-seq projected onto a protein-protein interaction (PPI) network.

Provide a GPU-accelerated Ollama runtime packaged with a shared Docker bridge network to enable multi-application local LLM serving.

Uses Meta’s Segment Anything Model (SAM) to perform human image segmentation and generate segmentation masks.

Local platform to build vulnerability intelligence and AI-assisted security workflows for asset/API analysis, exposure modeling (including SBOM), prioritization, patch planning, and safe verification.