OmicsBase

OmicsBase-powered omics analysis with transparent, reproducible Quarto reports.

OmicsBase lets you:

• Specify an analysis in natural language and upload your study data
• Sit back while an agent plans the workflow, writes R + Quarto source, and renders an HTML report
• Ask the agent to improve the analysis, tweak recipes, or fix the report
• Work in NoteThreads: literate notebooks where the agent runs R cells for you, backed by a Bioconductor book knowledge base
• Adapt a versioned team ReportPack (an existing R/Quarto directory) rather than generating every report from an empty scaffold

Getting started

Prerequisites

• Docker (with Compose)
• Node.js 20+ and npm

Install / run

cd omicsbase                      # repository root
cp .env.example .env              # add your LLM API key
make dev

• Frontend: http://localhost:3000
• API: http://localhost:8000
• API docs: http://localhost:8000/docs

make dev runs the backend, worker, Postgres, and Redis in Docker, and the Next.js frontend locally. See DOCKER.md for details and other run modes.

make down          # stop containers

Setup API key

Choose one of:

• Put your key in .env (copy from .env.example)
• Export an env variable before starting

Default provider uses ANTHROPIC_API_KEY. You can also set LLM_PROVIDER / LLM_MODEL.

Optional:

Variable	Purpose
API_KEY	Shared-deployment auth (X-API-Key header)
USE_DOCKER_SANDBOX	Run R/Quarto in an isolated container (requires omicsbase-runner image)
REPORT_PACKS_DIR	Optional administrator-managed directory of additional ReportPacks

Create a new analysis

1. Open the app and describe your research question
2. Attach data (CSV, TSV, Excel, BIOM, QIIME2 .qza, RDS, and more)
3. Choose Build (run through) or Plan (review the workflow first)
4. Open the workspace for the report preview, source files, and chat

Improve an existing project

In the workspace, ask for changes in natural language, for example:

• "Enable PERMANOVA and re-run"
• "Use Shannon for alpha diversity"
• "What does the beta diversity page show?"
• "Fix the bar plot"

You can also edit files in the Code panel, save, and re-render.

What it produces

• A structured analysis plan bound to analysis recipes
• An R + Quarto project (plan, source, data copies)
• A rendered HTML report plus machine-readable results
• A downloadable project zip

Post-render review checks the selected ReportPack's declared sources and artifacts, QMD pages, sessionInfo, and portable paths. Available packs are listed at GET /api/report-packs; plans select them by ID rather than raw filesystem path.

NoteThreads

• Ask questions or start analyses in a notebook-style thread; the agent runs R cells with a shared workspace
• Cells produce inline tables, plots, and output files you can inspect and download
• A curated Bioconductor book knowledge base (synced at runtime from public sources) grounds the agent's answers

Architecture

Layer	Stack
Frontend	Next.js, React, Monaco, TanStack Query
Backend	FastAPI, SQLAlchemy, Celery, Redis, PostgreSQL
Analysis	R, Bioconductor, Quarto (Docker)