Header-only generic optimizer for manifold-based nonlinear least squares

• Python support through JIT compilation
• SE3/SO3 manifold optimization with Lie algebra
• Automatic differentiation (no manual Jacobians needed)
• Sparse Schur complement solver
• Marginalization support via Sparse Gaussian Prior
• Cache-friendly SlotMap containers (O(1) operations)
• Compile-time type safety

Installation, requirements, and a quick start guide with code examples.

Deep dives into the core building blocks:

• Variables — Defining optimizable parameters on manifolds
• Error Terms — Creating constraints between variables
• Autodiff — How automatic differentiation works

Complete class and function documentation.

Tangent was originally developed as part of QDVO (Quasi-Direct Visual Odometry).