juxhinr / bindwidth

Evidence-aware on-prem LLM inference sizing and TCO calculator

398894496-arch / TDHarness-coding

Self-hosted AI workbench for a company LAN. One command on a Windows server sets up login, roster, model gateway, company disk and knowledge base; each employee's agent works on their own PC while model keys stay on the server. 一键安装的局域网 AI 工作台。