admiral-solver 1.0.0

Python SDK for Admiral combinatorial optimization cloud

Admiral

Combinatorial optimization cloud platform
Solve QUBO, Ising, CQM, MILP, Max-SAT and 6 more problem types
with specialized algorithms via cloud API, Python SDK, or CLI.

Docs · Pricing · Discord · Self-Host

---

Quick Start

pip install admiral-solver

from admiral import Solver
import numpy as np

solver = Solver(api_key="adm_sk_...")
Q = np.array([[-5, 2, 4], [0, -3, 1], [0, 0, -8]])
result = solver.solve_qubo(Q, timeout=30)
print(result.energy)     # -13.0
print(result.solution)   # [1, 0, 1]

Local mode (no server):

from admiral import LocalSolver
result = LocalSolver().solve_qubo(Q, solver="tabu", seed=42)

11 Problem Types

Type	Formulation	Variables	Solvers
QUBO	min x^TQx	binary	SB, PT, Tabu, Roof duality
Ising	min Σhᵢsᵢ + ΣJᵢⱼsᵢsⱼ	spin	SB, PT, Tabu
HUBO	min Σcₛ∏xᵢ	binary	Native SA, Quadratize→QUBO
DQM	min Σaᵢ(xᵢ)+Σbᵢⱼ	discrete	Discrete SA/Tabu, Bucket elim
CQM	min x^TQx+c^Tx s.t. Ax≤b	mixed	ADMM, Feasibility pump, B&C
MILP	min c^Tx s.t. Ax≤b	int+real	Branch & cut, RINS
MIQP	min x^TQx+c^Tx Ax≤b	mixed	QP B&C, ADMM
Max-SAT	max Σwⱼ·sat(Cⱼ)	boolean	WalkSAT, Core-guided RC2
PBO	min Σcⱼ∏lᵢ s.t. ≥b	binary	SAT cutting planes
WCSP	min Σf(x)	discrete	VAC+B&B, LNS, BP
Potts	min −ΣJδ(sᵢ,sⱼ)	discrete	Swendsen-Wang, Discrete PT

Architecture

Clients (SDK / REST / CLI)
        │
        ▼
   API Server (FastAPI)  ──▶  PostgreSQL
        │
        ▼
      Redis (job queue + rate limits)
        │
   ┌────┴────┐
   ▼         ▼
CPU Workers  GPU Workers
Tabu, B&C    SB, PT
ADMM, RC2

Self-Hosting

git clone https://github.com/admiral-dev/admiral-solver.git
cd admiral-solver
cp .env.example .env   # edit secrets
docker-compose up -d   # starts everything

Scale: docker-compose up -d --scale worker-cpu=8 --scale worker-gpu=4

K8s: kubectl apply -f deploy/k8s-api.yaml -f deploy/k8s-workers.yaml

CLI

admiral solve --type qubo --input problem.json --local
admiral convert --from qubo --to ising --input q.json
admiral bench --sizes 10,100,1000 --solver sb
admiral server start --port 8000
admiral worker start --redis redis://localhost:6379

API

curl -X POST https://api.admiral.dev/v1/solve \
  -H "Authorization: Bearer adm_sk_..." \
  -d '{"problem_type":"qubo","data":{"Q":[[-5,2,4],[0,-3,1],[0,0,-8]]}}'

Method	Endpoint	Description
POST	/v1/solve	Submit problem
GET	/v1/jobs/{id}	Status + result
GET	/v1/jobs/{id}/pool	Solution pool
POST	/v1/analyze	Complexity analysis
POST	/v1/convert	Format conversion
DELETE	/v1/jobs/{id}	Cancel
GET	/v1/solvers	List engines
GET	/v1/health	Health check

Development

pip install -e ".[dev]"
pytest tests/ -v
ruff check admiral/

License

Apache 2.0 — see LICENSE

---

Built by Admiral Technologies