Search

Contains models and classes for use with immutable storage

Logging connector implementation using in entity storage.

Wallet connector implementation using entity storage.

Run work that exceeds your machine's memory losslessly, within a guaranteed memory ceiling: 16-bit LLM fine-tuning and generation beyond device memory, and a budgeted runtime for large arrays (alpha).

OpenAI GPT-5 model plugin for XpertAI platform using Responses API

Section-based prompt composition and slot variable management for LLM applications

Private object storage with explicit Lenso instance references. Business methods are ordinary async functions. The package root has no filesystem, AWS SDK, database or Web imports.

slices v2: file-storage

slices v2: the file-storage service on this machine's disk. No key, nothing to provision.

slices v2: the file-storage service over any S3-compatible bucket: AWS, R2, Backblaze, MinIO, Tigris.

slices v2: the transcription service, backed by OpenAI (or any OpenAI-compatible gateway) through your own route.

Oneport account & metering client — login, token balance, redeem codes, and the managed-LLM proxy every Oneport tool bills through.

AI pre-ship secret & .env gate — deterministic detection, LLM triage, git-history aware

hugpy's router: one OpenAI-compatible address in front of every box's hugpy-wrapper front door. Picks the box per call (a model already serving with the same settings first, else a box that can load it), proxies the call (streaming, uploads, binary), bounces a refused call to the next box, and logs one row per decision.

Provider-routed LLM request retry policy for the QiLin

Boruna: a deterministic, capability-safe programming language and framework for LLM-native applications