Release notes#
Release notes summarize intended changes. For reproducible analysis, also record the exact Git commit, model files, quality-control database, and dependency environment.
Unreleased#
Installation changes#
Reworked the mainland-China installer to prefer micromamba, support TUNA, BFSU, USTC, upstream, and institutional channels, and keep mirror settings out of global user configuration.
Added dry-run, cached/offline, update, local model archive, custom model URL, and explicit manager/path options.
Made the official
PKU-EMBL/BASALT_WEIGHTHugging Face repository the first automatic model source, with Figshare and Baidu Netdisk fallbacks.Removed the obsolete
esteiniganddefaultschannel requirements and separated legacy CheckM from the maintained Python 3.12 environment.Removed the duplicate uv/pip scientific-stack transaction from
basalt_environment.yml; the base runtime is now solved once through Conda, while VAMB and LorBin remain explicitly optional adapters.Added shallow-clone installation guidance and a validated, persistent
~/.bashrcconfiguration forBASALT_WEIGHT.Added a bounded Hugging Face reachability check, a release-pinned model revision, and matching mainland-installer controls so model installation remains auditable on blocked or mirrored routes.
Documented VAMB as an isolated optional environment because its pinned PyTorch stack conflicts with the validated BASALT model runtime.
Documented
PKU-EMBL/LorBin-BASALT-Extrabinneras the maintained-e lintegration target, including pinned-source installation, dependency validation, warning-only failure semantics, and fork-commit reporting.
BASALT 1.2.1#
Tagged 15 May 2026.
Changes#
Updated the VAMB adapter for the VAMB 5 command family.
Added VAMB 5 to the Python 3.12 environment definition.
Added warning-based failure handling for optional binners so other candidate generators can continue.
Compatibility boundaries#
Optional-adapter success must be verified from logs and output FASTA files.
VAMB, PyTorch, CUDA, and driver versions remain part of the analysis provenance.
The Conda edition does not implement
BASALT --version; record the Git tag and commit directly.
BASALT 1.2.0#
Released 16 December 2025.
Changes#
Updated the documented environment to Python 3.12.
Added a LorBin adapter under
-e l.Selected CheckM2 as the default quality-control backend.
Added accelerator-compatible dependency paths for SemiBin 2 and model inference.
Made the model directory configurable through
BASALT_WEIGHT.Renamed the historical
--autoparaoption to--sensitive.
Compatibility boundaries#
The Conda edition does not implement
BASALT --versionorBASALT --check-deps.Optional binner adapters require independent installation and validation.
Resuming checkpoints created with earlier code or database versions is not recommended.
BASALT 1.1.0#
Released 12 June 2024.
Revised installation workflows.
Added China-mainland mirror support.
Added a Singularity distribution route.
Included maintenance and stability changes.
BASALT 1.0.0#
Released 18 August 2023.
Introduced multi-assembly candidate generation and bin selection.
Introduced model-based contig screening and read-supported retrieval.
Added bin-level reassembly and OLC stages.
Added external-bin data feeding.
Added CheckM-based quality assessment and textual checkpoints.
Publication#
Qiu, Z., Yuan, L., Lian, C.-A. et al. BASALT refines binning from metagenomic data and increases resolution of genome-resolved metagenomic analysis. Nature Communications 15, 2179 (2024). https://doi.org/10.1038/s41467-024-46539-7