Skip to main content

Genie documentation

A genome browser with an agent that fetches, measures and cites

Genie pairs a genome browser rendered by eg3, the third-generation WashU Epigenome Browser engine, with an LLM agent that drives it through declared tools for ENCODE search, track loading and simple statistics. The tools issue citation tokens, and the chat shows a token as a clickable chip only when it resolves to a stored record.

Read the docsGetting started
  • 3agent modes
  • 23declared tools (14 / 6 / 8)
  • 57assemblies in the genome picker
  • 5 Mbscope limit on four numeric tools

Find your way

Getting started

Install the workspace, run the desktop app or the local web app, and open your first project.

Genome browser

Search genes, zoom and pan, add tracks from public hubs or URLs, and tune track settings.

Assistant tools

Every declared tool by mode: parameters, results, limits and the citation each one issues.

Citations and evidence

How tool-issued tokens become chips, what the evidence panel shows, and what a chip does not prove.

Developers

The renderer, the local backend, the agent runtime bridge, the eg3 engine and the on-disk layout.

Reference

Keyboard shortcuts, the genome assemblies you can pick, configuration settings and a glossary.

How a claim is grounded

The agent runtime calls Genie tools, which search ENCODE, read bigWig files, add tracks drawn by eg3 and return citation tokens for stored evidence records. A chip opens the tool's stored result, the ENCODE file record, or, for navigation and viewport summaries, only the region.

A chip shows which record a claim points to, not that the sentence is correct.

Citations and evidence →
A Genie chat answer about H3K27ac signal at the MYC locus, with numbered citation chips after each claim
An answer whose claims carry numbered citation chips.