Live debug session

The recommended way to run the browser debugger around a real exploration is through anemone.debug.build_live_debug_environment().

Basic workflow

from random import Random

from anemone.debug import (
    build_live_debug_environment,
    serve_live_debug_session,
)

debug_environment = build_live_debug_environment(
    tree_exploration=tree_exploration,
    session_directory="debug-session",
)

result = debug_environment.controlled_exploration.explore(
    random_generator=Random(0)
)
debug_environment.finalize()

serve_live_debug_session("debug-session", port=8000)

What gets written

The live recorder writes a session directory containing:

• index.html for the browser viewer

• session.json with the current timeline payload

• control_state.json and commands.json for live interaction state

• snapshots/ with exported graph snapshots and metadata sidecars

Lifecycle

The session recorder writes incrementally during the search run. Calling debug_environment.finalize() marks session.json as complete. If the process exits early, the browser viewer can still open the session, but it will remain marked incomplete.

Reference example

See examples/debug_live_session_example.py for a small helper-oriented reference workflow.