gurobi-optimizer-integratorlisted

Real Gurobi (gurobipy) optimization for fleet scheduling, AHLS outreach batch selection, LP/MIP, IIS infeasibility proof, and $10/mo token-budget LP. Free pip size-limited license first (AE Fabrizio Ellis 2026-08-12). Trigger: gurobi, optimize schedule, mathematical optimization, dispatch agents, outreach capacity knapsack, IIS, token budget, certified proof.
IgorGanapolsky/mac-yolo-safeguards · ★ 3 · AI & Automation · score 69

Install: claude install-skill IgorGanapolsky/mac-yolo-safeguards

# Gurobi optimizer (system-wide, real) ## Truth - **Solver:** `gurobipy` 13.x via `~/.hermes/gurobi-venv` (also system python3.14) - **License:** size-limited free pip — non-production; ≤2000 vars/constraints - **AE path:** Fabrizio Ellis / Jue Xue for full trial when models grow - **Never mock** random "optimal" values - **Pulse steal (2026-08-19):** receipts carry `proof` (`obj_val`/`obj_bound`/`mip_gap`/`certified_optimal`). Infeasible models return IIS, not "looks feasible." ## Commands ```bash bin/gurobi-fleet license --json bin/gurobi-fleet evaluate --json bin/gurobi-fleet dispatch --file jobs.json --json bin/gurobi-fleet outreach --file prospects.json --capacity 15 --json bin/gurobi-fleet iis --file model.json --json bin/gurobi-fleet token-budget --file workloads.json --budget 10 --json bin/gurobi-mcp # MCP stdio server ``` ## MCP tools - `gurobi_license_info` — cached probe (do not re-license every solve) - `gurobi_solve_lp` — LP/MIP + certified proof + auto-IIS on INFEASIBLE - `gurobi_agent_dispatch` — priority MIP; `exclusive` tokens (`adb`,`git_lock`) ≤1; optional `ram_gb` - `gurobi_outreach_batch` — knapsack; bounce_risk ≤0.05; max 2 per `vertical`; DRAFT-only - `gurobi_evaluate` — 6 hermetic cases (LP, dispatch, outreach, license, IIS, token budget) - `gurobi_diagnose_iis` — exact conflicting constraints - `gurobi_token_budget` — binary local vs frontier under `$10` monthly ceiling ## Hybrid pattern (LLM front-end, solver back-end) 1. LLM parses messy i