What is Liatir

Liatir is a local-first bioinformatics desktop application. It runs native command-line tools (FastQC, Samtools, BCFtools, fastp, ...), custom JavaScript plugins (.lia files), WebAssembly tools, and local AI Models used by AI Tools - all from a single UI, entirely on your own machine.

Why local-first matters

Most modern bioinformatics platforms are cloud-based: you upload files to a remote server, pay per compute hour, and trust a third party with patient or proprietary data. Liatir takes the opposite approach.

• Your data never leaves your machine. No uploads. No API calls. No telemetry. Files are referenced by path — Liatir reads them directly.
• Works offline. Air-gapped HPC nodes, restricted institutional networks, no WiFi on a train — Liatir runs without any connectivity.
• No subscriptions or rate limits. Process as many samples as your local hardware allows.
• Reproducible by default. Every run stores its inputs, tool version, and parsed results locally. No external job IDs to chase, no logs that expire.

The Rust advantage

Liatir is built on Tauri 2 — a framework that pairs a Rust backend with a web-technology frontend. This matters for bioinformatics because genomic files are large.

• Streaming a gzip-compressed FASTQ, reading the first N lines of a 50 GB BAM, or processing a dense multi-sample VCF — these are synchronous file I/O operations in the Rust backend. They never block the JavaScript UI thread.
• Spawning samtools, bcftools, or fastp as a child process and capturing streamed stdout is handled natively without shelling out through a Node.js proxy.
• Binary formats (BAM, BCF, CRAM) are safely identified and flagged before any attempt is made to read them as text.

Extensibility: three layers

Liatir offers four ways to add new analysis capability:

1 — Native tools

Any binary installed in your system PATH can be wrapped as a native tool. Liatir checks availability, surfaces install instructions when the binary is missing, and runs the command with the parameters you choose. Run history and parsed results are stored automatically.

2 — .lia plugins

A .lia file is a self-contained extension bundle. Node plugins declare their inputs and outputs in code with definePlugin({ inputs, outputs }), and liatir build generates the bundle manifest from that contract. WASM plugins use a manifest file and a sandboxed plugin.wasm payload.

3 — WASM plugins

For performance-critical or cross-language logic, Liatir supports WebAssembly plugins. Compile from Rust, C, or any WASM-compatible language and expose it as a normal tool in the UI and in pipelines.

4 — AI Models and AI Tools

AI Models are locally installed model runtimes managed by Liatir. AI Tools are the pipeline capabilities that use those models for tasks such as single-cell annotation, sequence embedding, protein structure prediction, and genomic variant scoring.

Core concepts

Concept	Description
Data library	A registry of file paths on disk. Files are never copied — Liatir tracks references and detects when files move or disappear.
Native tool	A system binary wrapped with a Liatir UI, dependency check, and run history.
.lia plugin	A self-contained extension bundle that adds custom steps through the shared input/output contract.
WASM plugin	A WebAssembly-backed tool for performance-critical logic. FastQC is the primary example.
AI Model	A locally managed model runtime installed only when needed.
AI Tool	A pipeline-ready capability that runs a compatible local AI Model through the shared I/O contract.
Pipeline	A directed graph of steps whose typed outputs connect to typed inputs of subsequent steps.
Analysis run	A single execution: recorded inputs, stdout, parsed sections, and any output files.

Getting started

1. Add files to your Data library — import by path, no copying.
2. Run a native tool — start with FastQC or fastp on FASTQ files.
3. Read the AI guide — learn what AI Models and AI Tools do before interpreting results.
4. Install an AI Model — try local annotation, embedding, or structure workflows.
5. Build or import a .lia plugin — for custom logic or pipeline orchestration.
6. Connect steps in a pipeline — automatic type-matched data flow.