Getting started
Explore the offline workflow before connecting application data or making real model calls.
1. Start from the repository root
Use Python 3.11 or newer. On Windows, replace python3.11 with py -3.11. The first workflow uses only the committed synthetic service snapshot.
2. Create a fresh offline workspace
python3.11 testing/create_test_db.py --config testing/configs/fake_1_medium.tomlThe helper creates one working database with 30 shared tasks, four runs, and 120 pending attempts. It refuses an existing destination; do not repeat creation for an existing workspace.
3. Open the local operator dashboard
python3.11 monitor.py --config testing/configs/fake_1_medium.tomlOpen http://127.0.0.1:8766 on the machine running the framework. Select one of the four stored runs and start it explicitly. Each interactive fake worker waits a random 5–30 seconds before returning its deterministic transformation.
4. Inspect one selected run
python3.11 -m codex_agenting --config testing/configs/fake_1_medium.toml status --run-id 1All four sample configurations reference the same database. Stored run parameters determine execution; switching config files is unnecessary.
Before real execution
Choose an approved model and reasoning level, preserve the intended authentication path, and set an explicit compatible HTTPS endpoint in the actual application config. Validate that same configuration locally before a separately authorized, small test-data pilot.
See configuration for the distinction between defaults, stored run parameters, and session controls.