A) TOOLKIT / ARXIS
Foundational modules for LLM development
Core packages for adaptive memory, cache allocation, and verified decoding.
AI Layer
PLATFORM FOR PRODUCTION.
/// AI_STACK Built across two subcategories: Arxis toolkit modules for foundational LLM development, and agent-oriented frameworks + skills for extending existing agents.
B) AGENTS ORIENTED
Frameworks and skills to extend existing agents
An execution layer for agent workflows and integrations, with additional frameworks to be announced.
Arxis Core
Three packages, one coherent AI toolkit.
adaptive-memory
Learned memory management
Predictive prefetching, adaptive compression, and intelligent caching for inference workloads.
OPEN PACKAGE →pama-cache
Learned KV-cache allocation
Neural policy routing to keep, compress, or evict context under strict memory budgets.
OPEN PACKAGE →verified-decoding
Differentiable constraints
Guides generation using learned world models and logit repair instead of rigid grammar masking.
OPEN PACKAGE →