Liatir

Liatir is an open-source desktop application from the Liatir organization that brings bioinformatics tools, locally managed AI models, plugins and visual pipelines into one local-first environment. The homepage lists version 0.2.1 under the GNU GPL v3 and states that nothing opened in the app leaves the machine.

What It Is

Liatir is a desktop environment for bioinformatics built with Rust, Tauri 2 and SvelteKit. Users run native tools, AI models, .lia plugins, API connectors and external workflows such as Nextflow from one interface. Workspaces retain inputs, execution logs, results and provenance. Local analyses keep working offline once their dependencies are installed.

Tools, models and pipelines

Bundled tools include FastQC, fastp, seqkit, samtools, bcftools, BWA-MEM, minimap2 and SnpEff. Local AI models cover single-cell embeddings and annotation (for example Geneformer and scGPT) as well as molecular relaxation and dynamics, delivered as signed packages that install in one click and can use a GPU when available. Tools, models, plugins and API connectors are connected as nodes in visual pipelines. Built-in viewers show genome tracks, protein and molecular structures, and single-cell data next to the step that produced them.

Extensibility and assistant control

Plugins in the .lia format can run with Node, Python or WASM, either as pipeline nodes or standalone. A controlled local MCP interface lets an AI assistant on the same computer start a saved pipeline, using only allowed files and only after the user approves the run.

Scientific Showcases

The project publishes showcases documenting scientific questions, methods, results and limitations. One compares pretrained Geneformer and scGPT representations with PCA, Harmony and scVI on PBMC and Pancreas data. It reports that the pretrained models did not show a uniform advantage, and the study notes that it covers two datasets and one seed.