Artificial Intelligence

Client library and tooling for the Hugging Face Xet data storage system.

Client library for communicating with Hugging Face Xet storage servers. Use through the hf-xet crate.

Async runtime, configuration, logging, and utility infrastructure for the Hugging Face Xet client tools.

Data processing pipeline for chunking, deduplication, and file reconstruction; used in the Hugging Face Xet client tools. Intended to be used through the API in the hf-xet package.

Core data structures including MerkleHash, metadata shards, and Xorb objects.

Ergonomic top-level Agent SDK facade for the Bamboo agent framework

Core agent abstractions and execution primitives for the Bamboo agent framework

Compatibility umbrella crate re-exporting awaken

On-device inference engine for Apple silicon.

Independent / Multi-Agent Proximal Policy Optimization (PPO) Algorithms for the RelayRL framework

Conversational elicitation of strongly-typed Rust values via MCP

A fully self-contained AI agent backend framework with built-in web services, multi-LLM provider support, and comprehensive tool execution

A safe, fast, flexible code analysis and parsing library built in Rust. High-level entry point for the Thread ecosystem.

Runtime management CLI.

Procedural macros for stand-in MCP server framework.

A2A v1.0 protocol wire types and helpers for Awaken

Zero-trust sandbox for local inference and secure AI coding agent runtimes.

Glob and regex pattern matching for tool IDs in Awaken

Compatibility facade for Awaken runtime and server contracts

Phase-based execution engine, plugin system, and agent loop for Awaken

Storage backends (memory, file, PostgreSQL, SQLite mailbox) for Awaken agent state

Encrypted-first memory engine for AI agents, built on the Citadel database

A Rust framework for building AI agents from YAML specifications with trait-based extensibility

Decapod is a Rust-built governance runtime for AI agents: repo-native state, enforced workflow, proof gates, safe coordination.

Core AST engine for Thread - parsing, matching, and transforming code using AST patterns. Forked from ast-grep-core.

floDl — a flow-graph deep learning framework built on libtorch

Bounded session management for MCP servers — max sessions, FIFO eviction, rate limiting

Permission plugin with allow/deny/ask policies for Awaken tool execution

Server-driven UI component plugin (A2UI) for Awaken

Cloud fine-tune APIs (OpenAI, Anthropic, Together, Fireworks, Anyscale, Bedrock, Vertex AI) for the rullama agent framework. Dataset pipelines live in `rullama-datasets`; local LoRA/QLoRA/DoRA fine-tuning lives in `rullama-lora`.

Model Context Protocol (MCP) server exposing Citadel encrypted memory as tools

Cross-crate workflow plugins for elicitation — code recovery, multi-crate compositions, and server-side orchestration

Reverse engineering assistant that uses a locally running LLM to aid with pseudocode analysis.

High-performance, local AI memory with AAAK v3.2 protocol and temporal Knowledge Graph

Extensions and support for linear algebra in DataFusion

A lightweight, configurable, local-first full-text search engine for Markdown knowledge bases with Japanese text support

Model Context Protocol (MCP) client plugin for Awaken

OpenTelemetry observability plugin with GenAI semantic conventions for Awaken

Autonomous agent runtime for Citadel: ReAct + Reflexion + memory-backed plan caching

MCP server for semantic memory — pure-Rust embeddings, vector search, git-backed storage

Low-level, generic LLM provider abstraction for Rust

NER, coreference resolution, relation extraction, PII detection, and zero-shot entity types

SDK for mini-chat: policy plugin traits, models, and errors

Skill package discovery and activation plugin for Awaken

A composable, deterministic text data pipeline for ML. Ingest, denoise, chunk, split, and sample multi-source corpora into reproducible training triplets.

Persistent, git-friendly memory storage for AI agents with hybrid semantic + full-text search

Ling - The Omniglot Systems Language

Rule-based context message injection plugin for Awaken

Deferred tool loading with ToolSearch and probability-based deferral for Awaken

Training data pipelines for rullama: JSONL I/O, tokenization, dedup, format conversion (Alpaca / ChatML / OpenAI / ShareGPT / Together), and quality stats. Shared by cloud (`rullama-finetune`) and local (`rullama-lora`, in the rullama workspace) fine-tuning.