algenta-general-c
Domain kernels from model interop to risk. Compiled Mojo, loaded in-process.
100 modules · 1012 functions · CPU
Get started
pip install kernels torch
from kernels import get_kernel
kernel = get_kernel(
"thyn-ai/algenta-general-c",
version=1,
backend="cpu",
trust_remote_code=["thyn-ai/algenta-general-c"],
)
kernel.operations_research.eoq_quantity(1000, 50, 2) # -> 223.6068 economic order quantity, units
backend="cpu" selects the CPU build. On a Mac the loader otherwise looks for a Metal build, which
this family does not ship. trust_remote_code names the repositories you allow; Hugging Face's
trusted publishers load without it.
Plain Python in, plain Python out. Lists, tuples, buffers and tensors are accepted wherever the
contract expects a list; structured results are dictionaries. Some functions take a list and the
number of elements to use from it, which may not exceed the list's length; multi-dimensional data is
passed flattened, row-major, with its dimensions. Functions that update an argument do so in place,
as help() says. Every call is checked against the published contract before it reaches native
code. An invalid call raises KernelError with a stable code, never a crash. Engine kernels
report shape and finiteness problems as a status; the wrapper raises KernelError named after it.
Any function can also be called by name, with args as a list or a dict of parameter names:
kernel.execute("optimization.assignment", "assign_status_name", [...])
What's inside
| Module | Functions | What it does |
|---|---|---|
model_interop |
10 | Model file math: safetensors offsets, GGUF metadata size, ONNX FLOPs, RoPE scaling, perplexity |
model_parallel |
10 | Parallel training math: pipeline bubble, ring all-reduce bytes, ZeRO-3 memory, DDP overhead |
model_router |
10 | Model routing: request fit, endpoint load, fallback triggers, ensemble weights, latency tiers |
model_selection |
5 | Cross-validation splits: k-fold, stratified k-fold, walk-forward, purged/embargoed windows |
moe_production |
10 | Mixture-of-experts serving: token drop rate, expert load imbalance, routing entropy, capacity |
moe_routing |
10 | Mixture-of-experts gating: noisy top-k, Switch top-1, smooth top-2, expert-choice, cosine |
multi_objective |
7 | Multi-objective ranking: Pareto dominance and fronts, weighted sums, TOPSIS, ideal and nadir |
multimodal |
10 | Vision and audio math: ViT patches, image tokens, mel scale, STFT magnitude, CLIP similarity |
museum |
12 | Collection care: lux-hours, UV exposure, preservation index, climate stability, valuation |
music_theory |
12 | Music theory: MIDI-to-Hz, cents and semitones, scale membership, just intonation, tempo |
mycology |
10 | Mycology: spore dispersal, mycelial growth, Vanderplank epidemic rate, germination probability |
nanotechnology |
12 | Nanoscale physics: surface-to-volume, quantum confinement, mean free path, Stokes-Einstein |
nephrology |
10 | Kidney function: CKD-EPI eGFR, Cockcroft-Gault, FENa, TTKG, osmolar and free-water clearance |
neuropharmacology |
10 | CNS drug action: brain receptor occupancy, Cheng-Prusoff Ki, brain-to-plasma ratios, Emax dose |
neurophysics |
10 | Neuron biophysics: Nernst and Goldman potentials, leaky integrate-and-fire, 1/f spectral power |
nlp |
11 | Text basics: tokenization, n-grams, TF-IDF, cosine and Jaccard similarity, word counts |
nuclear |
12 | Nuclear physics: decay and half-life, activity, mass defect, Q-value, dose fall-off, shielding |
numerical_stability |
12 | Stable numerics: log-sum-exp, softmax, Kahan summation, Welford mean and variance, softplus |
numismatics |
12 | Coin collecting: melt value and premium, troy-ounce bullion, scarcity, survival, annual return |
numpy_ops |
10 | Element-wise math: clip, stable log(1+x) and exp(x)-1, sign, Heaviside step, sinc, hypotenuse |
nutrition |
12 | Nutrition: BMI, Mifflin-St Jeor BMR, TDEE, caloric deficit, macro split, ideal weight, VO2max |
observability |
10 | Serving telemetry: p99 latency, throughput EMA, cache miss rate, drift alert, health score |
ode |
8 | ODE solvers: Euler and RK4 integration, plus exact exponential decay and logistic growth |
online_learning |
10 | Streaming updates: EMA, online SGD, perceptron, passive-aggressive, AdaGrad, FTRL, DDM drift |
operations_research |
13 | Operations research: EOQ, newsvendor critical ratio, Little's law, M/M/1 queue length and wait |
ophthalmology |
10 | Eye optics: Snellen acuity, lensmaker's equation, Goldmann pressure, Ferry-Porter flicker |
optics |
10 | Optics: Snell's law, critical and Brewster angles, thin lens, Fresnel reflectance, Airy limit |
optimization.assignment |
7 | Assignment problem: Hungarian shortest augmenting paths with dual proofs, forbidden pairs |
optimization.explain |
12 | Linear program diagnostics: binding rows, shadow prices, infeasibility cores, cost breakdown |
optimization.flow |
10 | Network flow: Dinic max flow with min cut, min-cost flow by successive shortest paths |
optimization.integer |
6 | Mixed-integer linear programs: branch and bound on LP relaxations, 0/1 knapsack |
optimization.linear |
9 | Linear programming: tableau simplex with Bland's rule, duals, reduced costs, RHS ranging |
optimization.model |
37 | Optimization models: bounded continuous, integer and binary variables, expression trees |
optimization.multiobjective |
14 | Multi-objective: nondominated sort, crowding, NSGA-II, Monte Carlo hypervolume, lexicographic |
optimization.quadratic |
8 | Convex quadratic programming: active-set solve, KKT verification, convexity check, bounds |
optimization.robust |
9 | Robust linear programs: scenario sampling, wait-and-see optima with CVaR, min-max, chance rates |
optimization.routing |
15 | Vehicle routing: TSP 2-opt, Or-opt and iterated local search, capacity and time-window VRP |
optimization.scheduling |
10 | Scheduling: job-shop tabu search, RCPSP serial schedule generation, critical-path bounds |
optimization_meta |
7 | Simulated annealing, particle swarm, hill climbing and random search on built-in test functions |
optimize |
8 | Scalar optimization: golden section, Brent, bisection, Newton, gradient descent, line fit |
optmodel.formulation |
5 | LP and MILP formulations: dimension and bound checks, lowering to nonnegative variables, audits |
optmodel.judge |
3 | Benchmark judging: solve a formulation and compare its objective to a known optimum |
optmodel.solve |
2 | Formulation solving: simplex or branch and bound, optimality certificates, infeasibility cores |
orthodontics |
10 | Orthodontics: tooth movement rate, wire beam forces, Bolton ratio, Pont's index, force decay |
packaging |
12 | Packaging: box geometry, McKee compression, stacking load, cushion G, Arrhenius shelf life |
paleontology |
12 | Paleontology: C-14, K-Ar and U-Pb dating, Margalef diversity, extinction rates, femur body mass |
palynology |
10 | Palynology: exotic-spike pollen concentration, accumulation rate, tree pollen ratio, R-values |
pandas_groupby |
10 | Group aggregation: Welford mean and variance updates, running min, max, sum, approximate median |
paper_science |
12 | Paper and pulp: basis weight, bulk, tear, tensile and burst indices, kappa lignin, pulp yield |
pattern |
12 | String utilities: glob matching, template substitution, split, join, trim, pad, replace, count |
pde |
10 | Heat and wave equations in 1D by explicit finite differences, 2D Laplace by Jacobi, CFL limits |
pedology |
10 | Soil formation indices: base saturation, clay illuviation, Munsell redness, profile development |
pension |
12 | Retirement planning: replacement ratio, defined benefit, defined contribution, annuity factor |
perfumery |
12 | Fragrance formulation: concentration, dilution, note durations, accord balance, cost per ml |
persistence |
11 | In-memory key-value store, integer keys to numeric values: put, get, delete, snapshot, restore |
pest_control |
12 | Integrated pest management: economic threshold, injury level, trap catch rate, residual decay |
pharmacology |
10 | Pharmacokinetics and dose-response: Hill equation, therapeutic index, clearance, loading dose |
photonics |
12 | Optics: Snell's law, critical and Brewster angles, Fresnel reflectance, Beer-Lambert, Rayleigh |
photovoltaics |
10 | Solar PV: cell efficiency, fill factor, temperature correction, energy yield, performance ratio |
physics |
12 | Classical mechanics: kinematics, energy, momentum, projectile motion, work, elastic collisions |
phytochemistry |
10 | Plant assays: Folin-Ciocalteu phenolics, AlCl3 flavonoids, pH-differential anthocyanins, DPPH |
pipeline |
12 | List processing primitives: map, filter, reduce, batch, sliding window, flatten, zip, unique |
pipelines |
5 | Preprocessing chains: standardize, min-max, log1p, clip, median imputation, inverse transform |
plasma_physics |
10 | Plasma physics: Debye length, plasma frequency, Larmor radius, Alfvén speed, Lawson criterion |
pls |
4 | Partial least squares regression (PLS1) by NIPALS: fit, predict, R-squared |
plumbing |
12 | Pipe hydraulics: Reynolds number, Darcy-Weisbach head loss, pump power, sizing, water hammer |
pneumatics |
12 | Compressed air: Boyle's law, orifice flow, compressor power, cylinder force, receiver volume |
polymer |
12 | Polymer science: molecular weight averages, PDI, Mark-Houwink, Fox equation, Flory-Huggins |
polynomial |
10 | Polynomials: Horner evaluation, multiply, derivative, integral, composition, quadratic roots |
population |
10 | Population dynamics: exponential, logistic, Lotka-Volterra predator-prey and competition models |
positional_encoding |
10 | Transformer position encodings: sinusoidal, RoPE, ALiBi, T5 relative buckets, xPos scaling |
power_systems |
12 | AC power: active, reactive and apparent power, power factor, transformer ratio, line loss |
printing |
12 | Print production: DPI to LPI, pixel count, bleed area, spine width, ink coverage, CIE76 delta E |
privacy |
10 | Differential privacy: Laplace and Gaussian noise, exponential mechanism, randomized response |
procurement |
12 | Procurement: EOQ, reorder point, safety stock, inventory turnover, cost of ownership, scoring |
progexec.contract |
1 | Operation codes, operand kinds and status names for the FinQA arithmetic program executor |
progexec.exec |
1 | FinQA program execution: arithmetic, greater-than, table max/min/sum/average, step references |
progexec.roundexact |
1 | Exact half-to-even decimal quantization at a power-of-ten scale, matching Python bit for bit |
project_management |
12 | Earned value management (SV, CV, SPI, CPI, EAC, ETC, VAC, TCPI), PERT estimate, schedule float |
proteomics |
10 | Protein mass spectrometry: molecular weight, pI, extinction, m/z, coverage, PAI, emPAI, FDR |
protocol |
10 | Protocol Buffers wire format: base-128 varints, ZigZag integers, tagged fields, message codec |
psychometrics |
12 | Test theory and IRT: item difficulty, KR-20, Cronbach's alpha, SEM, z, T and stanine, 1PL, 2PL |
pulmonology |
10 | Lung function: minute and alveolar ventilation, FEV1/FVC, alveolar gas equation, compliance |
quantum |
11 | Qubit state vectors: Hadamard, Pauli X and Z gates, measurement, Bloch sphere, Bell state |
quantum_computing |
12 | Quantum computing estimates: gate error, surface-code qubits, quantum volume, Grover, Shor |
queue_ds |
12 | Min-heap priority queue and fixed-capacity circular buffer with push, pop and peek |
queueing |
11 | Queueing theory: M/M/1 and M/M/c metrics, Erlang B blocking, Erlang C waiting, Little's law |
radar |
12 | Radar: range equation, Doppler shift, unambiguous range and velocity, resolution, gain, noise |
radiometry |
10 | Radiometry: Planck spectral radiance, Stefan-Boltzmann, photon energy and flux, solid angle |
random_utils |
7 | Random sampling: weighted choice, Fisher-Yates shuffle, permutation, bootstrap, stratified |
real_estate |
12 | Real estate metrics: cap rate, GRM, cash-on-cash, DSCR, LTV, NOI, mortgage payment, break-even |
recommender |
9 | Collaborative filtering: user cosine similarity, k-NN rating prediction, popularity, RMSE |
refrigeration |
12 | Refrigeration cycles: COP and Carnot limits, superheat, subcooling, compressor work, EER, SEER |
regex_engine |
9 | Regex matching with wildcards, quantifiers, anchors and character classes; search, find, split |
reinforcement |
8 | Reinforcement learning: Q-learning, SARSA, value iteration, policy evaluation, epsilon-greedy |
reliability |
11 | Reliability: exponential and Weibull survival, MTBF, availability, series, parallel, k-of-n |
rerank_eval |
9 | Ranking metrics: precision, recall, hit rate, MRR, MAP and NDCG at k, pairwise margin accuracy |
reservoir |
7 | Streaming samplers: Vitter's Algorithm R, Efraimidis-Spirakis weighted reservoir, bootstrap |
rheology |
10 | Rheology: shear stress and rate, power-law and Bingham fluids, Maxwell relaxation, creep |
risk |
11 | Portfolio risk: VaR, CVaR, drawdown, Sharpe, Sortino, Calmar, information ratio, beta, alpha |
kernel.CONTRACT holds every signature, including the length rules for list arguments;
help(kernel.model_interop) documents each function.
Requirements
- Apple silicon: macOS 15 or later for the CPU build.
- Linux arm64 and x86-64, glibc 2.35 or later.
kernels0.17 or later and PyTorch 2.5 to 2.14. PyTorch has to be installed: the loader picks the build for your PyTorch version. The kernel itself never imports it.
Windows is not supported.
Notes
Calls into one kernel instance run one at a time; use processes for parallelism. Runtime state
does not survive fork(); start worker processes with spawn.
License
Algenta Community License 1.1 (LICENSE). Free for personal, research and open-source use, and
for internal use at organizations with fewer than 50 employees and under $5M in annual revenue.
Beyond that, a commercial license is required: https://algenta.ai/pricing.
Enforced in the compiled library, not just in this text: one concurrent native worker per device (ABI §9). A second process, family or thread waits its turn rather than running in parallel. That is the Community licence's worker floor made real; parallel execution comes with a commercial license.
Support
Generally Available on the platforms listed under Requirements. Within v1, functions are only added;
removals or signature changes ship as v2. Platforms, accelerators and PyTorch releases not listed are not
supported. Documentation: https://docs.algenta.ai (the kernels guide:
https://docs.algenta.ai/guides/kernels-on-hugging-face). Community: https://discord.gg/w8NDsph9an or this
repository's Community tab. Commercial licences and support: https://algenta.ai/pricing.
- Downloads last month
- 2
algenta
mojo
cpu
other
- Torch
- 2.14
- OS
- macoslinux
- Arch
- x86_64aarch64