Running Code

Basic Execution

State persists between run() calls within the same sandbox:

cell.run("x = 42")
result = cell.run("print(x * 2)")
print(result.stdout)     # "84"
print(result.exit_code)  # 0
print(result.latency_ms) # 0.24

Streaming Output

For long-running code, use callbacks to stream output in real-time:

Python
TypeScript

result = cell.run(
    "for i in range(10): print(f'Processing {i}')",
    on_stdout=lambda line: print(f"[live] {line}"),
    on_stderr=lambda line: print(f"[err]  {line}"),
)

const result = await cell.run(code, {
  onStdout: (line) => console.log(`[live] ${line}`),
  onStderr: (line) => console.error(`[err]  ${line}`),
});

Background Commands

Run commands asynchronously and poll for results:

handle = cell.command("pip install pandas && python train.py", background=True)

# Poll status
while handle.is_running:
    time.sleep(1)

print(handle.stdout)
print(handle.exit_code)

Execution Receipts

Every execution returns a cryptographic receipt:

result = cell.run("print(2 + 2)")
receipt = result.receipt

print(receipt.execution_id)  # Unique ID
print(receipt.code_hash)     # SHA-256 of input code
print(receipt.result_hash)   # SHA-256 of output
print(receipt.timestamp)     # Unix ms