optimaze 0.3.1

AI-powered optimization model tuning

Optimaze

AI-powered Gurobi optimization tuning — automatically speed up your models.

Installation

pip install optimaze

Usage

optimaze optimize model.py --repo https://github.com/you/repo --key YOUR_KEY

Options:

Flag	Description
model.py	Your Python script that builds and solves a Gurobi model
--repo	GitHub repo where improvements are committed
--key	Your Optimaze license key (or set OPTIMAZE_KEY)
--max-iterations	Max improvement iterations (default: 10)

How It Works

┌──────────────────────┐      ┌──────────────────────┐
│   Your Machine       │      │   Optimaze Cloud     │
│                      │      │                      │
│  1. Runs your model  │ ──── │  2. Analyzes logs    │
│     (your Gurobi     │ logs │     & source code    │
│      license)        │      │                      │
│                      │ ◄─── │  3. Generates code   │
│  4. Re-runs model    │ code │     improvements     │
│     with changes     │      │                      │
└──────────────────────┘      └──────────────────────┘

EULA-compliant architecture: Gurobi runs locally on your machine with your own license. The cloud service only analyzes solver logs and generates code — it never runs the solver.

Example

============================================================
  OPTIMAZE
  AI-powered optimization tuning
============================================================

Script:     /path/to/model.py
Repo:       https://github.com/you/repo

[Iteration 0] Running model...
  Runtime: 18.25s
  (baseline)

[Iteration 1] Running model...
  Runtime: 13.10s
  Improvement: 28.2% faster

RESULTS
----------------------------------------
Baseline:  18.25s
Best:      13.10s
Speedup:   +28.2%

To apply:
  git merge agent/optimize-sess_xxxxx

What It Tries

The agent analyzes your model's solver logs and applies targeted improvements:

• Solver Parameters — MIPFocus, Cuts, Presolve, Method, Heuristics, Threads, Symmetry
• Big-M Tightening — tighten large coefficients that cause numerical issues
• Symmetry Breaking — add constraints to eliminate symmetric solutions
• Valid Inequalities — strengthen the LP relaxation
• Branching Priorities — guide the branch-and-bound search
• Warm Starting — provide initial feasible solutions
• Lazy Constraints — defer rarely-binding constraints

Typical speedup: 20–50%

Requirements

• Python 3.10+
• Gurobi installed with a valid license (free academic license)
• Git configured for your repo

Environment Variables

Variable	Description
OPTIMAZE_KEY	Your license key
OPTIMAZE_SERVER	Custom server URL (optional)

License

MIT — see LICENSE