Crate agnosai 

AgnosAI — provider-agnostic AI orchestration framework.

This crate unifies core types, LLM providers, task orchestration, tool execution, adaptive learning, and an HTTP API server into a single library.

Feature flags

All features are opt-in (no defaults):

• sandbox — WASM (wasmtime) and Python subprocess tool sandboxing.
• fleet — Distributed multi-node crew execution and GPU scheduling.
• definitions — YAML/JSON agent preset loading and .agpkg packaging.
• hwaccel — Hardware accelerator detection via ai-hwaccel.
• otel — OpenTelemetry OTLP trace export via [hoosh::telemetry].
• full — Enables all of the above.

Modules

core

Core types, traits, and error handling for AgnosAI.

learning

Adaptive learning and reinforcement learning for agent optimization.

llm

LLM inference via hoosh.

orchestrator

Task scheduling, agent coordination, and crew execution.

server

HTTP API server for AgnosAI — REST endpoints for crew execution and management.

tools

Tool registry and execution — native Rust, WASM, and Python bridge.