Distributed quick start
Back to Home
Distributed quick start {#distributed-quick-start}¶
Spin up a local coordinator and worker to sanity-check the distributed stack before deeper testing. Use this workflow when onboarding new contributors or verifying environment changes.
Prerequisites {#prerequisites}¶
- Python environment with project dependencies installed (
pip install -e .[dev]
). - Separate terminals (or tmux panes) for server and worker processes.
- Optional:
.env
file for custom configuration; otherwise defaults apply.
Launch the server {#launch-server}¶
python -m bytebiota server --host 127.0.0.1 --port 8080 --no-reload
- Logs show worker capacity, web UI status, and tuning mode.
- Visit
http://127.0.0.1:8080
for the dashboard once workers connect.
Start a local worker {#start-worker}¶
In a second terminal:
python -m bytebiota worker --preset auto --server-url http://127.0.0.1:8080
- The worker auto-detects CPU/memory limits, registers with the server, and begins pulling assignments.
- Use
--verbose
for detailed execution traces during development.
Verify the run {#verify-run}¶
- Watch the server console for assignment and checkpoint messages.
- Load the web dashboard to confirm population growth, worker heartbeats, and seed-bank activity.
- Optional: run
python -m bytebiota export-diversity --log-file {data_dir}/checkpoints/experiment.jsonl --output diversity.csv
after a few minutes to inspect diversity metrics.
Stop and clean up {#cleanup}¶
- Press
Ctrl+C
in the worker terminal first, then stop the server. Both processes flush logs and checkpoints automatically. - Remove transient data if needed:
bash rm -rf data/checkpoints data/logs data/cache