Overview
Liatir is a local-first desktop environment for bioinformatics, built with Rust, Tauri 2, and SvelteKit. It combines native scientific tools, locally managed AI models, plugins, visual pipelines, and scientific viewers in one application.
Analyses can continue offline after their dependencies are installed. Workspaces preserve inputs, execution logs, results, and provenance, helping researchers keep the context of each run alongside its outputs.
Key Features
- Runs native bioinformatics tools and locally managed AI models.
- Supports visual pipelines and scientific viewers within the desktop application.
- Extends functionality through .lia plugins and API connectors.
- Integrates with external workflows and exposes a controlled local MCP interface for other applications.
- Uses shared scientific types and input/output contracts through its core package.
Local Research Workflow
Liatir is designed for analyses where data should remain on the local machine. Its Rust-powered native architecture is intended to handle large scientific files, while installed dependencies allow workflows to operate without a network connection.
The workspace model records the materials and history associated with an analysis rather than separating inputs, logs, results, and provenance across unrelated tools. This provides a unified environment for executing and reviewing computational biology work.
Scientific Evaluation
The project includes scientific showcases that document research questions, methods, measured results, limitations, and reproducibility artifacts. One completed comparison evaluated pretrained Geneformer and scGPT representations alongside PCA, Harmony, and scVI on PBMC and Pancreas single-cell datasets.
The reported results did not show a uniform advantage for pretrained models: scGPT was competitive on PBMC, while PCA and scVI remained strong on Pancreas. The study also identifies constraints, including its two-dataset, one-seed scope and differences in execution hardware, so its timings do not establish a same-host speed ranking.