What's new in refgenie 1.0
Refgenie 1.0 is a rewrite. One package now does the work of the three 0.x packages (refgenie, refgenconf, and refgenieserver). It installs with pip install "refgenie>=1.0.0a3". You can move an existing 0.x setup to 1.0 without rebuilding your assets; see Upgrade from legacy.
For users
Section titled “For users”- Your own recipes and asset types. Asset classes and recipes come from data channels, simple collections of YAML files on any web host. Subscribe to others' channels or publish your own. See Use data channels and Set up your own data channel.
- Single-file downloads. Servers can serve each asset file at its own URL, including straight from cloud storage, as well as whole-asset
.tgzarchives. See Serving modes. - Standard genome identifiers. Genomes are identified by GA4GH sequence collection digests and sequences by refget digests, with aliases like
hg38on top. See Manage aliases. - Sequence retrieval.
refgenie getseqreads a sequence or subsequence straight from the local sequence store, with no FASTA file needed. See Genome initialization. - A local dashboard.
refgenie dashopens a web app that shows your genomes, aliases, and assets, plus assets on your subscribed servers. See Use the dashboard. - AI-assistant access. A built-in, read-only MCP server lets Claude or another MCP client answer questions like "what genomes do I have?" See Connect AI assistants.
- GA4GH DRS endpoints. The server exposes standard DRS endpoints, so any DRS-aware workflow tool can fetch refgenie assets. See Run a server.
Under the hood
Section titled “Under the hood”- Database instead of a YAML config. Refgenie keeps its records in SQLite by default, or PostgreSQL for shared and server setups. See Database backends.
- RefgetStore.
refgenie genome initloads a genome's sequences into a content-addressable store, where identical sequences are kept once. This store powersgetseqand thefastaasset. - Built-in server.
refgenie serve(frompip install "refgenie[server]>=1.0.0a3") runs the REST API, DRS endpoints, and data channel aggregation. See Run a server. - Staging and remotes. Assets are staged for serving, as symlinks or
.tgzarchives, and can be pushed to remote storage like S3. See Stage assets for serving and Configure remote storage.