Quick Start
Get Runicorn running quickly with the current 0.7.0 workflow.
This page intentionally focuses on the shortest path from installation to your first visible run.
Step 1: Install Runicorn
Requirements: Python 3.10 or higher
Verify installation:
Step 2: Choose a storage root
Set a persistent storage location before you start collecting runs:
Runicorn resolves storage in this order:
rn.init(storage=...)RUNICORN_DIRrunicorn config --set-user-root ..../.runicorn
See Installation & Storage for the full layout and platform-specific examples.
Step 3: Run your first experiment
Create demo.py:
import random
import time
import runicorn as rn
run = rn.init(
path="quickstart/demo-experiment",
alias="baseline",
capture_console=True,
)
run.set_primary_metric("accuracy", mode="max")
for step in range(1, 51):
loss = 2.0 * (0.9 ** step) + random.uniform(-0.05, 0.05)
accuracy = min(0.98, 0.5 + step * 0.01 + random.uniform(-0.02, 0.02))
run.log({
"loss": round(loss, 4),
"accuracy": round(accuracy, 4),
"learning_rate": 0.001,
}, step=step)
time.sleep(0.1)
if step % 10 == 0:
print(f"Step {step}/50: loss={loss:.4f}, acc={accuracy:.4f}")
run.summary({
"final_accuracy": 0.95,
"total_steps": 50,
"notes": "Demo experiment from quickstart guide",
})
run.finish()
print("Completed:", run.id)
Run it:
Step 4: Open the viewer
Start the local web viewer:
Open http://127.0.0.1:23300.
You should now see the run in the experiments table.
Step 5: Explore the run
Open the run to inspect:
- charts and best-metric tracking
- live and captured logs
- summary fields
- assets, images, and config metadata
See Web UI Overview for the page-by-page tour.
What next?
Getting help
Need help?
- Search this documentation
- Check FAQ
- Read Troubleshooting
- Report issues on GitHub