Self-learning LLM runtime — TurboQuant KV-cache (6-8x compression), SONA adaptive learning, FlashAttention, speculative decoding, GGUF inference

Inference acceleration with speculative decoding

Research-backed Multi-LLM Router with parallel execution, learned routing (RouteLLM), prefix caching (RadixAttention), speculative decoding (Medusa/EAGLE), token compression (ISON), local LLM support (Ollama/vLLM/LM Studio), batch processing. Python bindi

Self-learning LLM runtime — TurboQuant KV-cache (6-8x compression), SONA adaptive learning, FlashAttention, speculative decoding, GGUF inference

Multi-agent orchestration framework for AI coding CLIs (Claude Code, Gemini CLI, Codex CLI). Routes tasks to specialized cheap/fast subagents while the frontier model orchestrates — inspired by speculative decoding. ~50% API cost reduction.

Postgres bytea parser

Contains parsers and serializers for ASN.1 (currently BER only)

Secure, audited & 0-dep implementation of base64, bech32, base58, base32 & base16

socket.io protocol parser

A fast Base64 decoder with a low level API. If you want a high level API, look at [base64-js](https://github.com/beatgammit/base64-js).

[fork] TextEncoder and TextDecoder (Polyfill for the Encoding Living Standard's API) For Node.JS.

Base64 encoding/decoding in pure JS

Base32 plugin for otplib using @scure/base

Minimal 0-dep QR code generator & reader. Supports ascii, term, gif and svg formats

A pure typescript BMP encoder and decoder

KTX 2.0 (.ktx2) parser and serializer.

Encode & decode XML and HTML entities with ease & speed

a low-level, lightweight protocol buffers implementation in JavaScript

asn1js is a pure JavaScript library implementing this standard. ASN.1 is the basis of all X.509 related data structures and numerous other protocols used on the web

Generate, and decode, base64 VLQ mappings for source maps and other uses