--- title: "Liatir - Powerful bioinformatics on your machine" description: "Local-first bioinformatics desktop app. Tools, AI models, visual pipelines and your existing Nextflow workflows run on your own computer — the models install in one click, with no conda or Docker." source: https://liatir.com/ --- # Liatir > Local-first bioinformatics desktop app built on Rust and Tauri. Liatir is a desktop application for bioinformatics. Native tools, visual pipelines, `.lia` plugins and scientific AI Models all run on the user's own machine, and the app keeps working offline. The pages below are the product documentation. - Full text of every page in this index: https://liatir.com/llms-full.txt - Single-page Markdown: append .md to a page URL, for example https://liatir.com/getting-started/install.md; or request the page with Accept: text/markdown. - Chunked retrieval corpus (JSON): https://liatir.com/quenta-docs.json - Curated bioinformatics background (JSON): https://liatir.com/quenta-knowledge.json - Sitemap: https://liatir.com/sitemap.xml - Generated: 2026-10-12T13:43:24.958Z ## Getting started - [Install Liatir Beta](https://liatir.com/getting-started/install): Liatir is a local-first desktop application. The application, its interface, your workspace index and your analysis history stay available without a network connection. - [Your first analysis](https://liatir.com/getting-started/first-analysis): This path creates a local single-cell embedding, opens its preview, and keeps the produced AnnData available for another tool or pipeline. - [Beta support and troubleshooting](https://liatir.com/getting-started/support): Liatir Beta focuses on one coherent local laboratory path: inspect data, run a single-cell AI workflow, inspect and reuse its outputs, and run a saved Nextflow workflow either directly or inside a… ## Introduction - [What is Liatir](https://liatir.com/introduction/overview): Liatir is a local-first bioinformatics desktop application. It runs bioinformatics tools (FastQC, Samtools, BCFtools, fastp, and more), custom plugins, local AI Models, and pipelines that connect… - [How Liatir Works](https://liatir.com/introduction/architecture): Liatir is a local-first desktop environment for bioinformatics. It gives you one place to manage data files, run tools, build pipelines, install local AI Models, and review results without sending… - [The name Liatir](https://liatir.com/introduction/the-name): Where the name Liatir comes from — lia "thread" and tir "to watch" in Tolkien's Elvish. - [Tauri](https://liatir.com/introduction/what-is-tauri): Liatir is a native desktop app built on Tauri, an open-source framework for lightweight, secure desktop applications. ## Scientific Showcases - [Liatir Scientific Showcases](https://liatir.com/showcases/overview): Scientific studies carried out with Liatir, with measured results, limitations, provenance and reproducibility artifacts. - [Single-cell foundation models vs established baselines](https://liatir.com/showcases/single-cell-foundation-benchmark): A two-dataset Liatir study of cell-type preservation, batch mixing and computational costs, with ten completed configurations and two documented UCE blockers. ## Data - [Data](https://liatir.com/data/overview): The Data page is the central file registry for Liatir. Every tool and plugin draws its input files from here. You add files once; everything else in the app can see them. ## Tools - [Tools](https://liatir.com/tools/overview): Liatir offers four kinds of analysis step: built-in tools that work out of the box, native tools — real bioinformatics programs, shipped with Liatir — AI Tools that use local AI Models, and External… - [BCFtools filter](https://liatir.com/tools/bcftools-filter): bcftools filter removes variant records from a VCF or BCF file that do not meet a quality expression. The output is a bgzip-compressed VCF (.vcf.gz) that can be registered in the Data library and… - [BCFtools](https://liatir.com/tools/bcftools): BCFtools is the standard toolkit for manipulating VCF and BCF variant files. Liatir uses the stats subcommand to compute summary statistics from a variant callset. - [BWA-MEM](https://liatir.com/tools/bwa-mem): BWA-MEM maps short sequencing reads to a reference genome and writes a SAM alignment file. - [External Workflows](https://liatir.com/tools/external-workflows): External Workflows let you run a saved workflow engine project from Liatir and reuse the same definition as one step in a larger Liatir pipeline. Nextflow is the first supported engine. - [fastp](https://liatir.com/tools/fastp): fastp is an all-in-one FASTQ pre-processing tool. It performs adapter trimming, quality filtering, per-read quality correction, and QC report generation in a single pass — typically faster than any… - [FastQC](https://liatir.com/tools/fastqc): FastQC performs comprehensive quality control analysis on raw sequencing reads. In Liatir it is a built-in tool — nothing to install, and noticeably faster than the original FastQC. - [Minimap2](https://liatir.com/tools/minimap2): Minimap2 is a versatile aligner for long reads, short reads, and assembled sequences. In Liatir it outputs SAM format. - [Samtools faidx](https://liatir.com/tools/samtools-faidx): samtools faidx creates an index for a FASTA file and enables random-access extraction of subsequences by coordinate. - [Samtools](https://liatir.com/tools/samtools): Samtools is the standard toolkit for working with sequence alignment data. Liatir uses the flagstat subcommand to produce a quick summary of alignment statistics from BAM, SAM, or CRAM files. - [seqkit](https://liatir.com/tools/seqkit): seqkit is a cross-platform toolkit for FASTA/FASTQ file manipulation. Liatir uses the stats subcommand to compute sequence statistics — read/sequence counts, length distribution, GC content, N50, and… - [Single-cell Reference Index](https://liatir.com/tools/single-cell-index): Builds the reference a single-cell experiment is measured against, from a genome and its annotation. - [Single-cell Quantification](https://liatir.com/tools/single-cell-quant): Turns raw single-cell sequencing reads into a count matrix: a table with one row per cell, one column per gene, and in each box the number of times that gene was read in that cell. - [SnpEff](https://liatir.com/tools/snpeff): SnpEff annotates VCF variants with predicted biological effects such as missense, stop gained, frameshift and splice-site changes. - [SnpSift Filter](https://liatir.com/tools/snpsift-filter): SnpSift Filter keeps only VCF variants that match a rule. SnpSift is included in the same verified, on-demand suite as SnpEff, so there is no second program to download or configure. ## AI Models and AI Tools - [AI Models](https://liatir.com/ai/models/overview): Liatir exposes only AI Models distributed as signed Runtime Boxes. Each box is an immutable, target-specific bundle containing the model, interpreter, dependencies, scientific runner, and legal… - [AI Tools](https://liatir.com/ai/tools/overview): AI Tools are built-in scientific tasks that use compatible installed AI Models. They are not .lia Plugins and they do not install their own model dependencies. - [Local AI for bioinformatics](https://liatir.com/ai/guide): Liatir runs supported scientific AI Models locally through signed Runtime Boxes. A Runtime Box contains the model weights, Python interpreter, packages, and Liatir runner that were tested together. - [Quenta](https://liatir.com/ai/quenta): Quenta is a local, read-only AI inside Liatir. It helps explain app state, bioinformatics workflows, Results, Jobs, pipeline structure, and failure logs using local context. - [scGPT Whole-human](https://liatir.com/ai/models/bowang-scgpt-whole-human): scGPT is a single-cell foundation model built for single-cell and multi-omics data. Liatir currently exposes the Whole-human checkpoint for local cell embeddings. - [Protenix base v1.0.0](https://liatir.com/ai/models/bytedance-protenix-base-v1-0-0): Protenix predicts the 3D structure of proteins and molecular complexes from their sequences. It is an independent model from Boltz-2, built by a different team, so it is useful as a second opinion on… - [Geneformer V1 10M](https://liatir.com/ai/models/ctheodoris-geneformer-v1-10m): Geneformer is a single-cell transcriptome foundation model designed to learn gene-network context from many human single-cell transcriptomes. - [Boltz-2](https://liatir.com/ai/models/jwohlwend-boltz-2): Boltz-2 predicts the 3D structure of proteins and molecular complexes from their sequences, and estimates how strongly a small molecule binds a protein. Liatir runs it entirely on your computer. - [UCE 4-layer](https://liatir.com/ai/models/snap-stanford-uce-4layer): UCE, or Universal Cell Embeddings, creates a numeric representation of each cell in a single-cell dataset. Those vectors can be used for comparison, clustering, visualization, and downstream analysis. - [Single-cell Embedding](https://liatir.com/ai/tools/single-cell-embedding): Single-cell Embedding is an AI Tool that creates numeric vectors for cells in an AnnData .h5ad dataset. ## Visualization - [Scientific Viewers](https://liatir.com/visualization/overview): Scientific viewers make output artifacts easier to inspect. They are especially important for AI outputs because raw files alone are often hard to understand. - [Genome Track Viewer](https://liatir.com/visualization/genome-track-viewer): The Genome Track Viewer displays genome-positioned artifacts such as BED tracks. - [Single-cell Viewer](https://liatir.com/visualization/single-cell-viewer): The Single-cell Viewer opens a profiled AnnData artifact together with the bounded preview produced by Single-cell Embedding. - [3D Structure Viewer](https://liatir.com/visualization/structure-viewer): The 3D Structure Viewer displays protein or molecular structure files. ## Pipelines - [Pipelines](https://liatir.com/pipeline/overview): Pipelines let you connect analysis steps visually so that the output from one step becomes the input for the next one. ## Local MCP - [Local MCP](https://liatir.com/mcp/overview): Local MCP lets an MCP-compatible client ask Liatir to run a pipeline you have already saved. Liatir keeps control of the scientific workflow: the client can request a run, but you must approve that… ## Plugins (.lia) - [Plugins (.lia)](https://liatir.com/plugins/overview): Build anything on top of Liatir: .lia plugins can run with Node, Python, or WASM, so you can choose the runtime that fits the job. - [Liatir Plugin API](https://liatir.com/plugins/api/overview): Comprehensive overview of the Liatir Plugin API, its lifecycle, usage, and platform specifics. - [Liatir API bridge](https://liatir.com/plugins/api-bridge): API bridge available inside a .lia plugin main handler. - [Liatir API packages](https://liatir.com/plugins/api-packages): package: the CLI scaffolds a single managed module (src/liatir.py / src/liatir.rs) into the project and liatir build keeps it in sync — there is nothing to install or update separately. - [Define plugin](https://liatir.com/plugins/define-plugin): Declares the input/output contract and entry point of a .lia plugin, in Node, Python, or WASM. - [Declaring plugin fields](https://liatir.com/plugins/field): Field builders for .lia plugin input and output schemas, in Node, Python, and WASM. - [Plugin structure](https://liatir.com/plugins/format): Under the hood, a .lia file is a ZIP archive with a fixed structure. Liatir validates the bundle signature before reading the manifest or executing any runtime payload. - [Getting started](https://liatir.com/plugins/getting-started): In order to start developing a Liatir plugin, you have to install the Liatir CLI from @liatir/cli, the Liatir command-line package for scaffolding, developing, and building .lia plugins. - [Plugin context](https://liatir.com/plugins/plugin-context): How the plugin handler receives its validated input in Node, Python, and WASM. - [Invoke](https://liatir.com/plugins/api/invoke): Low-level command invocation for advanced API usage. - [Log](https://liatir.com/plugins/api/log): Structured logging for plugins — stream log entries to the Jobs UI. - [Progress](https://liatir.com/plugins/api/progress): Progress tracking for plugins — display real-time progress in the Jobs UI. - [deps.checkMany](https://liatir.com/plugins/api/deps/check-many): Checks several command-line dependencies at once. - [deps.check](https://liatir.com/plugins/api/deps/check): Checks whether one command-line dependency is available. - [App API](https://liatir.com/plugins/api/desktop/app): Interface for retrieving system information and controlling the Liatir app. - [File System API](https://liatir.com/plugins/api/desktop/file-system): Comprehensive documentation for the File System API area. - [jobs.clearDone](https://liatir.com/plugins/api/jobs/clear-done): Clears completed jobs from the job registry. - [jobs.kill](https://liatir.com/plugins/api/jobs/kill): Stops a running async process job. - [jobs.list](https://liatir.com/plugins/api/jobs/list): Lists async process jobs tracked by Liatir. - [jobs.spawn](https://liatir.com/plugins/api/jobs/spawn): Starts an async native process from the Liatir API. - [jobs.status](https://liatir.com/plugins/api/jobs/status): Reads one async process job status. ## Other - [Download Liatir](https://liatir.com/download): Download Liatir for macOS (Apple silicon), Windows from the Microsoft Store, or Linux as an AppImage, .deb or .rpm. ## Optional - [Liatir Branding](https://liatir.com/branding): Official Liatir brand assets, colors, and basic usage guidelines. - [Donate](https://liatir.com/donate): Support Liatir’s Development, Maintenance, and Independence - [Privacy Policy](https://liatir.com/privacy): Privacy Policy for Liatir - [Terms and Conditions](https://liatir.com/terms): Terms and Conditions for Liatir