Environment¶
The genomics environment is created by scripts/env/install.sh (see Installation).
Use the project activation script instead of calling conda activate genomics manually:
The universal script:
- detects common Conda/Miniforge locations;
- activates the
genomicsenvironment; - cleans problematic CUDA library paths on affected systems;
- loads Bash completion for
genomicswhen the command is installed.
If your Conda installation is always under ~/miniforge3, this simpler script is also available:
Completion¶
The activation scripts run this automatically:
To install completion persistently for your user:
mkdir -p ~/.local/share/bash-completion/completions
genomics completion bash > ~/.local/share/bash-completion/completions/genomics
AlphaGenome server environment¶
The local AlphaGenome model server does not run in genomics: it needs JAX with CUDA, TensorFlow and
alphagenome>=0.7, which conflict with PyTorch. genomics alphagenome server setup creates a second
conda environment, alphagenome (or a venv inside the alphagenome_research checkout when there is no
conda), and the visualizer and genomics alphagenome server start run the server with its interpreter.
You never need to activate it. Override the choice with ALPHAGENOME_SERVER_PYTHON /
ALPHAGENOME_SERVER_DIR.
| Variable | Default | Meaning |
|---|---|---|
ALPHAGENOME_API_KEY |
– (also read from ~/.env) |
hosted API key |
ALPHAGENOME_ADDRESS |
– | use a self-hosted server, e.g. grpc://127.0.0.1:50051 |
ALPHAGENOME_TLS_CA_CERT |
– | CA certificate of a self-signed TLS server |
ALPHAGENOME_SERVER_DIR |
../alphagenome_research, then ~/.local/share/genomics/alphagenome_research |
checkout used for the local server |
ALPHAGENOME_SERVER_PYTHON |
the env alphagenome or another env with CUDA jax |
interpreter of the local server |
ALPHAGENOME_SERVER_HOST / ALPHAGENOME_SERVER_PORT |
127.0.0.1 / 50051 |
where the local server listens |
GENOMICS_DATA_ROOT / GENOMICS_RESULTS_ROOT |
/dados/GENOMICS_DATA / results/ |
datasets / runs and caches |