Jupyter Integration

Installation

pip install niki[jupyter]

Quick Start

# Load the extension
%load_ext niki

# Monitor a cell
%%niki_watch
import scanpy as sc
adata = sc.read_h5ad("large_dataset.h5ad")
sc.pp.filter_cells(adata, min_genes=200)

When the cell completes, you’ll receive a notification with:

Execution time
Exit status
Captured output
Any matplotlib figures

Named Tasks

Give cells descriptive names:

%%niki_watch preprocessing_sample1
sc.pp.normalize_total(adata)
sc.pp.log1p(adata)
sc.pp.highly_variable_genes(adata)

Options

Option	Description
`--silent` / `-s`	Suppress local output
`--no-plots`	Don’t capture matplotlib figures
`--priority P0\|P1\|P2`	Set notification priority

Examples

# Silent execution (notification only)
%%niki_watch --silent
model.fit(X_train, y_train, epochs=100)

# High priority notification
%%niki_watch critical_step --priority P0
results = expensive_computation()

# Skip plot capture
%%niki_watch --no-plots
for i in range(100):
    plt.figure()
    plt.plot(data[i])
    plt.savefig(f"plot_{i}.png")
    plt.close()

How It Works

Output Capture: stdout/stderr are captured while still displaying in real-time
Figure Detection: After cell execution, NIKI detects open matplotlib figures
Context Gathering: Cell index and notebook name are included
Notification: Data sent to NIKI daemon via HTTP

Requirements

NIKI daemon must be running: niki daemon start
IPython/Jupyter must be installed

Troubleshooting

Extension not loading

# Check if extension is available
import niki.jupyter
print(niki.jupyter.IPYTHON_AVAILABLE)  # Should be True

No notifications received

Check daemon is running: niki daemon status
Check notification config: niki doctor