Rust Libraries

Overview

A growing collection of Rust crates extracted from production desktop applications, designed to solve common challenges when building AI-powered tools. Each crate is independent, well-documented, and battle-tested in real applications like Gloss, VisionForge, and Palisade.

Crates

• agent-graph: A LangGraph-inspired agent orchestration framework for Rust. Define nodes, edges, conditional routing, and state machines for multi-step AI workflows with async execution.
• LLM-Pipeline: Typed prompt pipeline for local LLMs via Ollama. Chain prompts with structured output parsing, retries, and streaming support.
• Tauri-Queue: Background job queue for Tauri 2 apps with SQLite persistence, priority scheduling, retry logic, and progress reporting to the frontend.
• ComfyUI-RS: Rust client for ComfyUI's WebSocket + REST API. Submit workflows, track progress, and retrieve generated images programmatically.
• Ollama-Vision-RS: Vision model integration for Ollama. Send images with prompts for captioning, analysis, and visual Q&A.
• AI-Batch-Queue: Batch processing queue for AI workloads with concurrency control, deduplication, and result caching.
• job-queue: Generic async job queue with SQLite backing, dead-letter handling, and configurable retry policies.
• Tauri-React-Hooks: TypeScript React hooks for Tauri 2 IPC—invoke commands, listen to events, and manage Tauri state with idiomatic React patterns.

Technical Highlights

• All crates use Tokio for async runtime compatibility
• SQLite-backed persistence where applicable (rusqlite)
• Serde-based serialization throughout
• Comprehensive error types with thiserror
• Examples and integration tests included per crate

Technology Stack

• Language: Rust 2021 edition
• Async Runtime: Tokio
• Serialization: serde, serde_json, rmp-serde (MessagePack)
• Storage: rusqlite (SQLite)
• AI Integration: Ollama HTTP API, ComfyUI WebSocket API
• Desktop: Tauri 2 plugin system

Current Status

Stable and actively maintained. All crates are used in production across multiple desktop applications. New crates added as common patterns emerge from ongoing projects.