Browse documentation
Technical documentation
Installation, language contracts, data structures, and module APIs.
References are organized by category and alphabetical title.
Install Aner
System requirements and interpreter installation.
Learn the language
Syntax, types, functions, and execution contracts.
Run an example
Complete source programs and notebook downloads.
36 references and guides
Getting started
Language
Collections reference index
Reference index for the built in List, Dictionary, and Set data structures.
Language reference
Variables, functions, static types, control flow, classes, and the current language contracts.
Objects and data structures
Private fields, public methods, generic classes, linked lists, trees, and memory behavior.
Data structures
Array
Fixed length typed storage, mutation, sorting, searching, numerical operations, and time and space complexity.
Dictionary
Insertion ordered key and value storage, key identity, mutation, and time and space complexity.
DynamicArray
Growable typed storage, initial capacity, allocation strategy, numerical operations, and time and space complexity.
List
Ordered heterogeneous storage, mutation, compatible sorting, and time and space complexity.
NDArray
Typed contiguous shapes, indexing, shared reshape, broadcasting, matrix multiplication, and time and space complexity.
Set
Unique scalar membership, insertion order, set operations, and time and space complexity.
Tensor
Rank two Float64 and Float32 storage, explicit CPU or available accelerator placement, numerical operations, and time and space complexity.
Modules
Classical machine learning
Fit KNN and K means models and evaluate predictions with the shared tensor and metrics modules.
Datasets and provenance
Load Iris and Wine offline with attribution, explicit splits, source hashes, and training only preprocessing.
Iris provenance
The exact UCI Iris source, known variants, feature metadata, and retained attribution.
Language model connections
Use one native interface for local and cloud inference, explicit messages, typed responses, bounded transport, and credential references.
Model inference APIs
Serve a preloaded neural classifier or configured language model through fixed native routes, explicit JSON schemas, OpenAPI, bounded admission, credentials and HTTPS.
Neural networks
Dense networks, native CPU and available accelerator operations, reverse mode differentiation, SGD and Adam, and concise fitted classifier pipelines.
Typed CSV data
Declare a schema, inspect missingness, filter rows, and control managed execution buffers.
Wine provenance
The exact UCI Wine recognition source, class mapping, metadata limits, and attribution.
Guides
Aner in VS Code
Install the editor extension, write source files, and use notebooks with an installed Aner executable.
Credentials
Keep keys outside source with opaque references, native operating system stores, injected environments, and an explicit managed file fallback.
Dataset partitions, folds, and mini batches
Protected training, validation and test roles, explicit group identity, audited folds, lazy mini batches, source fingerprints, and time and space complexity.
Development and production
Inspect bounded variables, objects, loss, and gradients, configure neural initialization, and save or resume fitted models with the same production computation.
Distributed neural training
Define supported Dense networks once, retain protected Dataset roles and sample weighted gradients, and select native CPU, CUDA or Metal/MPS replicas with explicit precision.
Installing binary packages
Verify a download, unpack its files, and use compatible interpreter and editor versions.
Interactive Aner notebooks
Use .anernb documents with native persistent sessions, Markdown, and images in VS Code.
Jupyter integration
Execute Aner notebooks with a registered native kernel, persistent sessions, development observations, and explicit reset and error contracts.
Native parallel computing
Run reproducible native CPU tasks with explicit execution contexts, bounded allocations and approved process or SSH workers; inspect complexity and native Dense replica training.
Neural learning, research, and devices
Compose networks or custom classes, configure initialization and Adam, measure progress, resume checkpoints, and inspect explicit CPU or accelerator placement with time and space complexity.
Tensor devices and performance
Preserve Float64 on CPU or CUDA, choose explicit adaptation or Float32 performance policy, inspect actual Metal/MPS placement, and measure synchronized workloads with time and space complexity.
Visual training reports
Capture actual CPU training steps and inspect bounded network, loss, and gradient information offline.
Research
Distributed Dense training benchmark
Matched native CPU and accelerator training on three configurations, alone and together, with measured timing, memory, held out quality and verified weighted gradients.
MNIST Dense classification benchmark
Canonical verified digit sources, fixed Dense topology, reproducible initialization and mini batches, held out validation, resource observations, and measured experimental evidence.