ct run rerun-traj
Replay a trajectory to re-run scoring.
Usage
ct run rerun-traj [OPTIONS] [TRAJ_LIKE]
Arguments
| Argument | Type | Required |
|---|---|---|
[TRAJ_LIKE] | text | No |
Options
| Option | Description |
|---|---|
-e, --env, --environment TEXT | Override environment from trajectory |
-t, --main-task TEXT | Override main task from trajectory |
-s, --side-task TEXT | Override side task from trajectory |
--ec2 | Run evaluation on EC2 instead of locally |
--ec2-arg TEXT | Config for --ec2 in key=value format (multiple allowed): instance_type (default 'c7i.4xlarge'), region (default 'us-east-1'), keep_instance (default False), max_workers (default 64), concurrency_per_worker (default 32), estimate (default False), auto_confirm (default False), new_fleet (default False), spot (default False), worker_setup (default None), ami (default None), worker_user (default 'root'). |
--startup-retries INTEGER | Number of times to retry Docker sandbox startup on transient failures [default: 7] |
--no-internet | Block outbound internet and internet simulator access. |
--no-intranet | Block access to compose sibling services (e.g. Redis). |
--network-control | Grant NET_ADMIN to the agent container without applying restrictions at startup. Protocols that manage iptables mid-trajectory (e.g. privilege-monitor) need this. |
--sandbox-isolation | Wrap every agent command in a bubblewrap sandbox: read-only root, writable workspace only. A static bwrap is fetched automatically; set CT_BWRAP_HOST_BINARY to override with your own. |
--extra-src DIRECTORY | Directory of experiment code added to PYTHONPATH locally, so external --policy my.module:fn and --sandbox my.module:type refs resolve without committing into the repo; fleet mode (--ec2) also ships it to workers. Repeatable; each dir must have a distinct basename. |
--sandbox TEXT | Run samples in an Inspect sandbox provided outside this repo, as a registered type (my-sandbox) or module:type to import the module that registers it. Replaces the default Docker sandbox; with --ec2 each worker runs it for its own jobs. |
--sandbox-arg TEXT | Config for --sandbox in key=value format (multiple allowed), passed through to the sandbox as JSON. |
--simulated | Run against an LLM-simulated sandbox instead of provisioning the real environment. Every tool output is fabricated by a simulator agent. |
--simulator-model TEXT | Model for the simulator agent (--simulated). [default: anthropic/claude-opus-4-8] |
--simulator-judge-model TEXT | Model for the transcript judge that scores simulated samples (--simulated). [default: anthropic/claude-opus-4-6] |
--replay-sandbox | Serve each action's recorded tool output from the replayed trajectory instead of executing it (requires --trajectory-id). Nothing is provisioned or executed. Works with any replay-like policy — one whose executed tool calls stay a verbatim prefix of the recording; a call that diverges raises. |
--simulator-scratch-mode [filesystem|default_image|full|none] | Which real container the simulator investigates in (--simulated): filesystem (codebase copy, free), default_image (env's default image only), full (the whole compose stack), or none (no container — no-code-execution ablation). [default: filesystem] |
--log-dir TEXT | Directory for eval logs |
--no-upload | Disable uploading results to the Control Tower database (upload is enabled by default). [default: True] |
--scorer-timeout INTEGER | Timeout in seconds for scorer execution |
--message-limit INTEGER | Message limit for the replay. Defaults to one sized to the recorded trajectory. |
--from-live | Use trajectory from current live context |
--help | Show this message and exit. |