admiral-solver 2.0.1

Python SDK for Admiral — combinatorial optimization platform. 11 problem types, 33 solvers, solver racing, AI advisor, NL builder.

Admiral Solver

Python SDK for the Admiral Combinatorial Optimization Platform
11 problem types · 33 proprietary algorithms · Solver racing · AI advisor · NL builder

---

Install

pip install admiral-solver

Quick Start

from admiral import Solver

solver = Solver(api_key="adm_sk_...")

# Solve a QUBO problem
result = solver.solve_qubo(
    Q=[[-5, 2, 4], [0, -3, 1], [0, 0, -8]],
    timeout=10,
    race=True
)
print(result.energy)     # -13.0
print(result.solution)   # [1, 0, 1]

Features

• 11 problem types: QUBO, Ising, HUBO, DQM, CQM, MILP, MIQP, MAX-SAT, PBO, WCSP, Potts
• 33 proprietary algorithms: Auto-selected or raced in parallel
• Solver Racing: Race up to 8 solvers — first to converge wins
• AI Advisor: Analyze problem structure, get solver recommendations
• NL Builder: Describe problems in plain English, get formulations
• Auto-Detect: Submit JSON data, Admiral identifies the problem type

New in v2.0

# AI Advisor — analyze before solving
analysis = solver.analyze("qubo", {"Q": Q})
print(analysis.complexity)        # "EASY"
print(analysis.recommendations)   # [{solver: "admiral_01", confidence: 0.95}]

# NL Builder — natural language to formulation (powered by AI)
formulation = solver.formulate("Minimize travel between 5 cities")
print(formulation.problem_type)   # "qubo"
print(formulation.explanation)    # {variables: "...", objective: "..."}

# Solver Racing — race all algorithms
result = solver.solve_qubo(Q, race=True)
print(result.race_metadata)       # {solvers_raced: 8, ...}

# Auto-Detect problem type
detected = solver.detect({"Q": Q})
print(detected.detected_type)     # "qubo"
print(detected.confidence)        # 0.95

All Problem Types

Method	Type	Variables
solve_qubo(Q)	QUBO	Binary {0,1}
solve_ising(h, J)	Ising	Spin {-1,+1}
solve_hubo(terms, n)	HUBO	Binary, higher-order
solve_dqm(domains)	DQM	Discrete multi-valued
solve_cqm(c, Q, A, b)	CQM	Mixed with constraints
solve_milp(c, A, b)	MILP	Mixed integer linear
solve_miqp(c, Q, A, b)	MIQP	Mixed integer quadratic
solve_maxsat(clauses)	MAX-SAT	Boolean
solve_pbo(obj, cons, n)	PBO	Pseudo-boolean
solve_wcsp(n, domains, fns)	WCSP	Discrete constraint
solve_potts(nodes, states, edges)	Potts	Multi-state lattice

Links

• Platform: admiral-platform.tech
• API Docs: api.admiral-platform.tech/v1/docs
• Dashboard: admiral-platform.tech/dashboard.html
• PyPI: pypi.org/project/admiral-solver