https://cua.ai/assets/run-CV-2srQ1.js

1!function(){try{var e="undefined"!=typeof window?window:"undefined"!=typeof global?global:"undefined"!=typeof globalThis?globalThis:"undefined"!=typeof self?self:{},n=(new e.Error).stack;n&&(e._posthogChunkIds=e._posthogChunkIds||{},e._posthogChunkIds[n]="01a102ca-23d2-74b3-b081-b517399c563b")}catch(e){}}();import{j as e}from"./jsx-runtime-u17CrQMm.js";import{u as t}from"./index-DPhldNWm.js";import"./chunk-4N6VE7H7-DExpvS7J.js";const c={title:"cb run",description:"Run a task or dataset with an agent or the oracle, run one interactively, and follow runs."};function d(n){const s={a:"a",code:"code",figure:"figure",h2:"h2",h3:"h3",p:"p",pre:"pre",span:"span",strong:"strong",table:"table",tbody:"tbody",td:"td",th:"th",thead:"thead",tr:"tr",...t(),...n.components};return e.jsxs(e.Fragment,{children:[`
2`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Command"}),e.jsx(s.th,{children:"Description"})]})}),e.jsxs(s.tbody,{children:[e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.a,{href:"#cb-run",children:e.jsx(s.code,{children:"cb run"})})}),e.jsx(s.td,{children:"Run a task or dataset, and manage runs."})]}),e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.a,{href:"#cb-run-task",children:e.jsx(s.code,{children:"cb run task"})})}),e.jsx(s.td,{children:"Run one task variant."})]}),e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.a,{href:"#cb-run-dataset",children:e.jsx(s.code,{children:"cb run dataset"})})}),e.jsx(s.td,{children:"Run every task and variant of a dataset in parallel."})]}),e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.a,{href:"#cb-run-list",children:e.jsx(s.code,{children:"cb run list"})})}),e.jsx(s.td,{children:"List all runs with status."})]}),e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.a,{href:"#cb-run-info",children:e.jsx(s.code,{children:"cb run info"})})}),e.jsx(s.td,{children:"Show detailed info about a run."})]}),e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.a,{href:"#cb-run-watch",children:e.jsx(s.code,{children:"cb run watch"})})}),e.jsx(s.td,{children:"Watch a run in real-time with live updates."})]}),e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.a,{href:"#cb-run-stop",children:e.jsx(s.code,{children:"cb run stop"})})}),e.jsx(s.td,{children:"Stop a run and release its sandboxes."})]}),e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.a,{href:"#cb-run-logs",children:e.jsx(s.code,{children:"cb run logs"})})}),e.jsx(s.td,{children:"View combined logs from a run or session."})]}),e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.a,{href:"#cb-interact",children:e.jsx(s.code,{children:"cb interact"})})}),e.jsx(s.td,{children:"Run a task's setup with its desktop visible, then evaluate."})]})]})]}),`
3`,e.jsxs(s.p,{children:["Every command also accepts the ",e.jsx(s.a,{href:"/cua-bench/reference/cli#global-options",children:"global options"}),"."]}),`
4`,e.jsx(s.h2,{id:"cb-run",children:e.jsx(s.code,{children:"cb run"})}),`
5`,e.jsx(s.p,{children:"Run a task or dataset, and manage runs."}),`
6`,e.jsxs(s.p,{children:["Run a task (a directory with main.py) or a dataset, with an agent or the task's oracle. ",e.jsx(s.code,{children:"cb run <path>"})," is shorthand for ",e.jsx(s.code,{children:"cb run task|dataset <path>"}),"."]}),`
7`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"text","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"text","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:'<span data-line=""><span>cb run [COMMAND]</span></span>'}})})})}),`
8`,e.jsx(s.p,{children:e.jsx(s.strong,{children:"Examples"})}),`
9`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"bash","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"bash","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:`<span data-line=""><span style="color:#8B949E"># Run a task with its oracle solution</span></span>
10<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> ./tasks/hello_file_env</span></span>
11<span data-line=""> </span>
12<span data-line=""><span style="color:#8B949E"># Run a dataset with an agent, four variants at once</span></span>
13<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> cua-bench-basic</span><span style="color:#79C0FF"> --agent</span><span style="color:#A5D6FF"> cua-agent</span><span style="color:#79C0FF"> --model</span><span style="color:#A5D6FF"> anthropic/claude-sonnet-4-20250514</span><span style="color:#79C0FF"> -j</span><span style="color:#79C0FF"> 4</span></span>
14<span data-line=""> </span>
15<span data-line=""><span style="color:#8B949E"># List runs</span></span>
16<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> list</span></span>`}})})})}),`

17`,e.jsx(s.h3,{id:"cb-run-task",children:e.jsx(s.code,{children:"cb run task"})}),`
18`,e.jsx(s.p,{children:"Run one task variant."}),`
19`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"text","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"text","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:'<span data-line=""><span>cb run task [OPTIONS] &#x3C;task_path></span></span>'}})})})}),`
20`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Argument"}),e.jsx(s.th,{children:"Type"}),e.jsx(s.th,{children:"Default"}),e.jsx(s.th,{children:"Description"})]})}),e.jsx(s.tbody,{children:e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.code,{children:"<task_path>"})}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{children:"required"}),e.jsx(s.td,{children:"Path to task directory (containing main.py)."})]})})]}),`
21`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Flag"}),e.jsx(s.th,{children:"Short"}),e.jsx(s.th,{children:"Type"}),e.jsx(s.th,{children:"Default"}),e.jsx(s.th,{children:"Description"})]})}),e.jsxs(s.tbody,{children:[e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--variant-id"}),e.jsx(s.code,{children:"--variant-id"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"0"})}),e.jsx(s.td,{children:"Task variant index (default: 0)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--agent"}),e.jsx(s.code,{children:"--agent"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Agent to use for evaluation (e.g., cua-agent)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--agent-import-path"}),e.jsx(s.code,{children:"--agent-import-path"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:'Import path for custom agent (e.g., "path.to.agent:MyCustomAgent").'})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--model"}),e.jsx(s.code,{children:"--model"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Model to use with the agent (e.g., anthropic/claude-sonnet-4-20250514)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--max-steps"}),e.jsx(s.code,{children:"--max-steps"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"100"})}),e.jsx(s.td,{children:"Maximum number of steps for agent execution (default: 100)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--on"}),e.jsx(s.code,{children:"--on"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:[e.jsx(s.code,{children:"local"})," | ",e.jsx(s.code,{children:"cloud"})," | ",e.jsx(s.code,{children:"e2b"})," | ",e.jsx(s.code,{children:"daytona"})," | ",e.jsx(s.code,{children:"modal"})," | ",e.jsx(s.code,{children:"cloudflare"})," | ",e.jsx(s.code,{children:"vercel"})," | ",e.jsx(s.code,{children:"morph"})," | ",e.jsx(s.code,{children:"runloop"})," | ",e.jsx(s.code,{children:"fly"})," | ",e.jsx(s.code,{children:"blaxel"})," | ",e.jsx(s.code,{children:"codesandbox"})," | ",e.jsx(s.code,{children:"northflank"})," | ",e.jsx(s.code,{children:"aws"})," | ",e.jsx(s.code,{children:"gcp"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:["Where sandboxes run: local (this machine), cloud, your own cloud account (aws, gcp, modal; connect it first with ",e.jsx(s.code,{children:"cua cloud connect <provider>"}),"), or a contrib provider (e2b, daytona; needs a cua SDK built with contrib and the provider's key). Default: CUA_DEFAULT_ON, else ",e.jsx(s.code,{children:"cua config set default.on"}),", else local."]})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--kind"}),e.jsx(s.code,{children:"--kind"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:[e.jsx(s.code,{children:"auto"}

21)," | ",e.jsx(s.code,{children:"container"})," | ",e.jsx(s.code,{children:"vm"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:["What kind of sandbox: container or vm. Default auto: Linux as a container unless the task or image says VM; Windows, macOS and Android are VM-only. Default: CUA_DEFAULT_KIND, else ",e.jsx(s.code,{children:"cua config set default.kind"}),"."]})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--runtime"}),e.jsx(s.code,{children:"--runtime"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:[e.jsx(s.code,{children:"auto"})," | ",e.jsx(s.code,{children:"gvisor"})," | ",e.jsx(s.code,{children:"runc"})," | ",e.jsx(s.code,{children:"qemu"})," | ",e.jsx(s.code,{children:"lume"})," | ",e.jsx(s.code,{children:"kubevirt"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:["Which engine: gvisor or runc (local containers), qemu or lume (local VMs), gvisor or kubevirt (cloud containers, VMs). Implies its kind. Default auto: the SDK picks (CUA_DEFAULT_RUNTIME, else ",e.jsx(s.code,{children:"cua config set default.runtime"}),")."]})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--image"}),e.jsx(s.code,{children:"--image"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Image for every task (overrides the task's setup_config.image): an OS alias (linux, windows, macos:tahoe), ghcr.io/trycua/<os>, any registry ref or a digest, or pool:<name> for an existing Fleet pool. Env: CUA_BENCH_IMAGE."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--cpu"}),e.jsx(s.code,{children:"--cpu"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"vCPUs per sandbox."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--memory"}),e.jsx(s.code,{children:"--memory"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Memory per sandbox (e.g. 4096, 4G)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--claim-ttl"}),e.jsx(s.code,{children:"--claim-ttl"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"cloud: how long a claim outlives a crashed run (default 15m; renewed while running)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--output-dir"}),e.jsx(s.code,{children:"--output-dir"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Output directory for session results."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--attempts"}),e.jsx(s.code,{children:"--attempts"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"1"})}),e.jsx(s.td,{children:"Run every variant this many times (fresh sandbox each); summary.json reports pass@k (default: 1)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--retries"}),e.jsx(s.code,{children:"--retries"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"0"})}),e.jsx(s.td,{children:"Retry a variant that failed before evaluation (sandbox start, connection) up to this many times, with backoff (default: 0)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--keep-runs"}),e.jsx(s.code,{children:"--keep-runs"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"After the run, delete all but the N newest runs in the runs directory (default: keep everything; env CUA_BENCH_KEEP_RUNS)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--max-age"}),e.jsx(s.code,{children:"--max-age"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"number"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"After the run, delete runs older than DAYS (env CUA_BENCH_MAX_AGE_DAYS)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--max-results-size"}),e.jsx(s.code,{children:"--max-results-size"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"After the run, delete the oldest runs until the runs directory fits in SIZE, e.g. 20G (env CUA_BENCH_MAX_RESULTS_SIZE)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--oracle"}

21),e.jsx(s.code,{children:"--oracle"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Run the oracle solution (the default when no agent is given)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--noop"}),e.jsx(s.code,{children:"--noop"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Set up and evaluate with no actions (the null baseline for eval parity)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--warm"}),e.jsx(s.code,{children:"--warm"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"cloud: keep one replica ready when the managed pool is first created."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--detach"}),e.jsx(s.code,{children:"--detach"})]}),e.jsx(s.td,{children:e.jsx(s.code,{children:"-d"})}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Run in the background and return (follow with cb run watch <id>)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-task--dry-run"}),e.jsx(s.code,{children:"--dry-run"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Print each variant's image, kind, variant and backend, then exit without starting a sandbox."})]})]})]}),`
22`,e.jsx(s.p,{children:e.jsx(s.strong,{children:"Examples"})}),`
23`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"bash","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"bash","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:`<span data-line=""><span style="color:#8B949E"># Run variant 0 with the oracle solution</span></span>
24<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> task</span><span style="color:#A5D6FF"> ./tasks/hello_file_env</span></span>
25<span data-line=""> </span>
26<span data-line=""><span style="color:#8B949E"># Run variant 2 with an agent in the cloud</span></span>
27<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> task</span><span style="color:#A5D6FF"> ./tasks/hello_file_env</span><span style="color:#79C0FF"> --variant-id</span><span style="color:#79C0FF"> 2</span><span style="color:#79C0FF"> --agent</span><span style="color:#A5D6FF"> cua-agent</span><span style="color:#79C0FF"> --model</span><span style="color:#A5D6FF"> anthropic/claude-sonnet-4-20250514</span><span style="color:#79C0FF"> --on</span><span style="color:#A5D6FF"> cloud</span></span>
28<span data-line=""> </span>
29<span data-line=""><span style="color:#8B949E"># Show what would run without starting a sandbox</span></span>
30<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> ./tasks/hello_file_env</span><span style="color:#79C0FF"> --dry-run</span></span>`}})})})}),`
31`,e.jsx(s.h3,{id:"cb-run-dataset",children:e.jsx(s.code,{children:"cb run dataset"})}),`
32`,e.jsx(s.p,{children:"Run every task and variant of a dataset in parallel."}),`
33`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"text","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"text","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:'<span data-line=""><span>cb run dataset [OPTIONS] &#x3C;dataset_path></span></span>'}})})})}),`
34`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Argument"}),e.jsx(s.th,{children:"Type"}),e.jsx(s.th,{children:"Default"}),e.jsx(s.th,{children:"Description"})]})}),e.jsx(s.tbody,{children:e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.co

34de,{children:"<dataset_path>"})}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{children:"required"}),e.jsx(s.td,{children:"Path to dataset directory, or dataset name from registry."})]})})]}),`
35`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Flag"}),e.jsx(s.th,{children:"Short"}),e.jsx(s.th,{children:"Type"}),e.jsx(s.th,{children:"Default"}),e.jsx(s.th,{children:"Description"})]})}),e.jsxs(s.tbody,{children:[e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--max-parallel"}),e.jsx(s.code,{children:"--max-parallel"})]}),e.jsx(s.td,{children:e.jsx(s.code,{children:"-j"})}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"4"})}),e.jsx(s.td,{children:"Variants running at once; in the cloud also the managed pool's max size (default: 4)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--max-variants"}),e.jsx(s.code,{children:"--max-variants"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Maximum number of variants to run per task (default: all)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--task-filter"}),e.jsx(s.code,{children:"--task-filter"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Filter tasks by name pattern (glob)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--agent"}),e.jsx(s.code,{children:"--agent"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Agent to use for evaluation (e.g., cua-agent)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--agent-import-path"}),e.jsx(s.code,{children:"--agent-import-path"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:'Import path for custom agent (e.g., "path.to.agent:MyCustomAgent").'})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--model"}),e.jsx(s.code,{children:"--model"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Model to use with the agent (e.g., anthropic/claude-sonnet-4-20250514)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--max-steps"}),e.jsx(s.code,{children:"--max-steps"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"100"})}),e.jsx(s.td,{children:"Maximum number of steps for agent execution (default: 100)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--on"}),e.jsx(s.code,{children:"--on"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:[e.jsx(s.code,{children:"local"})," | ",e.jsx(s.code,{children:"cloud"})," | ",e.jsx(s.code,{children:"e2b"})," | ",e.jsx(s.code,{children:"daytona"})," | ",e.jsx(s.code,{children:"modal"})," | ",e.jsx(s.code,{children:"cloudflare"})," | ",e.jsx(s.code,{children:"vercel"})," | ",e.jsx(s.code,{children:"morph"})," | ",e.jsx(s.code,{children:"runloop"})," | ",e.jsx(s.code,{children:"fly"})," | ",e.jsx(s.code,{children:"blaxel"})," | ",e.jsx(s.code,{children:"codesandbox"})," | ",e.jsx(s.code,{children:"northflank"})," | ",e.jsx(s.code,{children:"aws"})," | ",e.jsx(s.code,{children:"gcp"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:["Where sandboxes run: local (this machine), cloud, your own cloud account (aws, gcp, modal; connect it first with ",e.jsx(s.code,{children:"cua cloud connect <provider>"}),"), or a contrib provider (e2b, daytona; needs a cua SDK built with contrib and the provider's key). Default: CUA_DEFAULT_ON, else ",e.jsx(s.code,{children:"cua config set default.on"}),", else local."]})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--kind"}),e.jsx(s.code,{children:"--kind"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:[e.jsx(s.code,{children:"auto"})," | ",e.jsx(s.code,{children:"container"})," | ",e.jsx(s.code,{children:"vm"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:["What kind of sandbox: container or vm. Default auto: Linux as a container unless the task or image says VM; Windows, macOS and Android are VM-only. Default: CUA_DEFAULT_KIND, else ",e.jsx(s.co

35de,{children:"cua config set default.kind"}),"."]})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--runtime"}),e.jsx(s.code,{children:"--runtime"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:[e.jsx(s.code,{children:"auto"})," | ",e.jsx(s.code,{children:"gvisor"})," | ",e.jsx(s.code,{children:"runc"})," | ",e.jsx(s.code,{children:"qemu"})," | ",e.jsx(s.code,{children:"lume"})," | ",e.jsx(s.code,{children:"kubevirt"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:["Which engine: gvisor or runc (local containers), qemu or lume (local VMs), gvisor or kubevirt (cloud containers, VMs). Implies its kind. Default auto: the SDK picks (CUA_DEFAULT_RUNTIME, else ",e.jsx(s.code,{children:"cua config set default.runtime"}),")."]})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--image"}),e.jsx(s.code,{children:"--image"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Image for every task (overrides the task's setup_config.image): an OS alias (linux, windows, macos:tahoe), ghcr.io/trycua/<os>, any registry ref or a digest, or pool:<name> for an existing Fleet pool. Env: CUA_BENCH_IMAGE."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--cpu"}),e.jsx(s.code,{children:"--cpu"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"vCPUs per sandbox."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--memory"}),e.jsx(s.code,{children:"--memory"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Memory per sandbox (e.g. 4096, 4G)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--claim-ttl"}),e.jsx(s.code,{children:"--claim-ttl"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"cloud: how long a claim outlives a crashed run (default 15m; renewed while running)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--output-dir"}),e.jsx(s.code,{children:"--output-dir"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Output directory for session results."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--attempts"}),e.jsx(s.code,{children:"--attempts"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"1"})}),e.jsx(s.td,{children:"Run every variant this many times (fresh sandbox each); summary.json reports pass@k (default: 1)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--retries"}),e.jsx(s.code,{children:"--retries"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"0"})}),e.jsx(s.td,{children:"Retry a variant that failed before evaluation (sandbox start, connection) up to this many times, with backoff (default: 0)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--keep-runs"}),e.jsx(s.code,{children:"--keep-runs"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"After the run, delete all but the N newest runs in the runs directory (default: keep everything; env CUA_BENCH_KEEP_RUNS)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--max-age"}),e.jsx(s.code,{children:"--max-age"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"number"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"After the run, delete runs older than DAYS (env CUA_BENCH_MAX_AGE_DAYS)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--max-results-size"}),e.jsx(s.code,{children:"--max-results-size"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"After the run, delete the oldest runs until the runs directory fits in SIZE, e.g. 20G (env CUA_BENCH_MAX_RESULTS_SIZE)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--oracle"}

35),e.jsx(s.code,{children:"--oracle"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Run the oracle solution (the default when no agent is given)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--noop"}),e.jsx(s.code,{children:"--noop"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Set up and evaluate with no actions (the null baseline for eval parity)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--warm"}),e.jsx(s.code,{children:"--warm"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"cloud: keep one replica ready when the managed pool is first created."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--detach"}),e.jsx(s.code,{children:"--detach"})]}),e.jsx(s.td,{children:e.jsx(s.code,{children:"-d"})}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Run in the background and return (follow with cb run watch <id>)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-dataset--dry-run"}),e.jsx(s.code,{children:"--dry-run"})]}),e.jsx(s.td,{}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Print each variant's image, kind, variant and backend, then exit without starting a sandbox."})]})]})]}),`
36`,e.jsx(s.p,{children:e.jsx(s.strong,{children:"Examples"})}),`
37`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"bash","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"bash","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:`<span data-line=""><span style="color:#8B949E"># Run a registry dataset with its oracle solutions</span></span>
38<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> dataset</span><span style="color:#A5D6FF"> cua-bench-basic</span></span>
39<span data-line=""> </span>
40<span data-line=""><span style="color:#8B949E"># Run one variant of each matching task, eight at once, in the background</span></span>
41<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> ./tasks</span><span style="color:#79C0FF"> --task-filter</span><span style="color:#A5D6FF"> 'click*'</span><span style="color:#79C0FF"> --max-variants</span><span style="color:#79C0FF"> 1</span><span style="color:#79C0FF"> -j</span><span style="color:#79C0FF"> 8</span><span style="color:#79C0FF"> --detach</span></span>
42<span data-line=""> </span>
43<span data-line=""><span style="color:#8B949E"># Run every variant three times and report pass@k</span></span>
44<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> dataset</span><span style="color:#A5D6FF"> cua-bench-basic</span><span style="color:#79C0FF"> --agent</span><span style="color:#A5D6FF"> cua-agent</span><span style="color:#79C0FF"> --attempts</span><span style="color:#79C0FF"> 3</span></span>`}})})})}),`
45`,e.jsx(s.h3,{id:"cb-run-list",children:e.jsx(s.code,{children:"cb run list"})}),`
46`,e.jsx(s.p,{children:"List all runs with status."}),`
47`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"text","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"text","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:'<span data-line=""><span>cb run list [OPTIONS]</span></span>'}})})})}),`
48`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Flag"}),e.jsx(s.th,{children:"Short"}),e.jsx(s.th,{children:"Type"}),e.jsx(s.th,{children:"Default"}),e.jsx(s.th,{children:"Description"})]})}),e.jsx(s.tbody,{children:e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-list--verbose"}

48),e.jsx(s.code,{children:"--verbose"})]}),e.jsx(s.td,{children:e.jsx(s.code,{children:"-v"})}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Show verbose debugging information."})]})})]}),`
49`,e.jsx(s.p,{children:e.jsx(s.strong,{children:"Examples"})}),`
50`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"bash","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"bash","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:`<span data-line=""><span style="color:#8B949E"># List runs</span></span>
51<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> list</span></span>
52<span data-line=""> </span>
53<span data-line=""><span style="color:#8B949E"># Include debugging details</span></span>
54<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> list</span><span style="color:#79C0FF"> --verbose</span></span>`}})})})}),`
55`,e.jsx(s.h3,{id:"cb-run-info",children:e.jsx(s.code,{children:"cb run info"})}),`
56`,e.jsx(s.p,{children:"Show detailed info about a run."}),`
57`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"text","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"text","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:'<span data-line=""><span>cb run info &#x3C;run_id></span></span>'}})})})}),`
58`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Argument"}),e.jsx(s.th,{children:"Type"}),e.jsx(s.th,{children:"Default"}),e.jsx(s.th,{children:"Description"})]})}),e.jsx(s.tbody,{children:e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.code,{children:"<run_id>"})}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{children:"required"}),e.jsx(s.td,{children:"Run ID to show info for."})]})})]}),`
59`,e.jsx(s.p,{children:e.jsx(s.strong,{children:"Examples"})}),`
60`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"bash","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"bash","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:`<span data-line=""><span style="color:#8B949E"># Show one run</span></span>
61<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> info</span><span style="color:#A5D6FF"> 30c12572</span></span>`}})})})}),`
62`,e.jsx(s.h3,{id:"cb-run-watch",children:e.jsx(s.code,{children:"cb run watch"})}),`
63`,e.jsx(s.p,{children:"Watch a run in real-time with live updates."}),`
64`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"text","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"text","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:'<span data-line=""><span>cb run watch &#x3C;run_id></span></span>'}})})})}),`
65`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Argument"}),e.jsx(s.th,{children:"Type"}),e.jsx(s.th,{children:"Default"}),e.jsx(s.th,{children:"Description"})]})}),e.jsx(s.tbody,{children:e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.code,{children:"<run_id>"})}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{children:"required"}),e.jsx(s.td,{children:"Run ID to watch."})]})})]}),`
66`,e.jsx(s.p,{children:e.jsx(s.strong,{children:"Examples"})}),`
67`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"bash","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"bash","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:`<span data-line=""><span style="color:#8B949E"># Follow a detached run</span></span>
68<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> watch</span><span style="color:#A5D6FF"> 30c12572</span></span>`}})})})}),`

69`,e.jsx(s.h3,{id:"cb-run-stop",children:e.jsx(s.code,{children:"cb run stop"})}),`
70`,e.jsx(s.p,{children:"Stop a run and release its sandboxes."}),`
71`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"text","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"text","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:'<span data-line=""><span>cb run stop &#x3C;run_id></span></span>'}})})})}),`
72`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Argument"}),e.jsx(s.th,{children:"Type"}),e.jsx(s.th,{children:"Default"}),e.jsx(s.th,{children:"Description"})]})}),e.jsx(s.tbody,{children:e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.code,{children:"<run_id>"})}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{children:"required"}),e.jsx(s.td,{children:"Run ID to stop."})]})})]}),`
73`,e.jsx(s.p,{children:e.jsx(s.strong,{children:"Examples"})}),`
74`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"bash","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"bash","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:`<span data-line=""><span style="color:#8B949E"># Stop a run</span></span>
75<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> stop</span><span style="color:#A5D6FF"> 30c12572</span></span>`}})})})}),`
76`,e.jsx(s.h3,{id:"cb-run-logs",children:e.jsx(s.code,{children:"cb run logs"})}),`
77`,e.jsx(s.p,{children:"View combined logs from a run or session."}),`
78`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"text","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"text","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:'<span data-line=""><span>cb run logs [OPTIONS] &#x3C;identifier></span></span>'}})})})}),`
79`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Argument"}),e.jsx(s.th,{children:"Type"}),e.jsx(s.th,{children:"Default"}),e.jsx(s.th,{children:"Description"})]})}),e.jsx(s.tbody,{children:e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.code,{children:"<identifier>"})}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{children:"required"}),e.jsx(s.td,{children:"Run ID or Session ID to view logs for."})]})})]}),`
80`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Flag"}),e.jsx(s.th,{children:"Type"}),e.jsx(s.th,{children:"Default"}),e.jsx(s.th,{children:"Description"})]})}),e.jsx(s.tbody,{children:e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-run-logs--tail"}),e.jsx(s.code,{children:"--tail"})]}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Show only the last N lines."})]})})]}),`
81`,e.jsx(s.p,{children:e.jsx(s.strong,{children:"Examples"})}),`
82`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"bash","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"bash","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:`<span data-line=""><span style="color:#8B949E"># Print a run's logs</span></span>
83<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> logs</span><span style="color:#A5D6FF"> 30c12572</span></span>
84<span data-line=""> </span>
85<span data-line=""><span style="color:#8B949E"># Print the last 50 lines</span></span>
86<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> run</span><span style="color:#A5D6FF"> logs</span><span style="color:#A5D6FF"> 30c12572</span><span style="color:#79C0FF"> --tail</span><span style="color:#79C0FF"> 50</span></span>`}})})})}),`

87`,e.jsx(s.h2,{id:"cb-interact",children:e.jsx(s.code,{children:"cb interact"})}),`
88`,e.jsx(s.p,{children:"Run a task's setup with its desktop visible, then evaluate."}),`
89`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"text","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"text","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:'<span data-line=""><span>cb interact [OPTIONS] &#x3C;env_path></span></span>'}})})})}),`
90`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Argument"}),e.jsx(s.th,{children:"Type"}),e.jsx(s.th,{children:"Default"}),e.jsx(s.th,{children:"Description"})]})}),e.jsx(s.tbody,{children:e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.code,{children:"<env_path>"})}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{children:"required"}),e.jsx(s.td,{children:"Path to the task directory, or a task name with --dataset or --dataset-path."})]})})]}),`
91`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Flag"}),e.jsx(s.th,{children:"Type"}),e.jsx(s.th,{children:"Default"}),e.jsx(s.th,{children:"Description"})]})}),e.jsxs(s.tbody,{children:[e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--variant-id"}),e.jsx(s.code,{children:"--variant-id"})]}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"0"})}),e.jsx(s.td,{children:"Task variant index (default: 0)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--dataset"}),e.jsx(s.code,{children:"--dataset"})]}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Registry dataset to resolve the task from (CUA_REGISTRY_HOME)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--dataset-path"}),e.jsx(s.code,{children:"--dataset-path"})]}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Path to dataset directory containing multiple tasks."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--max-steps"}),e.jsx(s.code,{children:"--max-steps"})]}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Maximum number of env.step() calls before stopping."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--screenshot"}),e.jsx(s.code,{children:"--screenshot"})]}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Save a screenshot to this file."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--trace-out"}),e.jsx(s.code,{children:"--trace-out"})]}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Record a trace and save it as a dataset to this path on exit."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--on"}),e.jsx(s.code,{children:"--on"})]}),e.jsxs(s.td,{children:[e.jsx(s.code,{children:"local"})," | ",e.jsx(s.code,{children:"cloud"})," | ",e.jsx(s.code,{children:"e2b"})," | ",e.jsx(s.code,{children:"daytona"})," | ",e.jsx(s.code,{children:"modal"})," | ",e.jsx(s.code,{children:"cloudflare"})," | ",e.jsx(s.code,{children:"vercel"})," | ",e.jsx(s.code,{children:"morph"})," | ",e.jsx(s.code,{children:"runloop"})," | ",e.jsx(s.code,{children:"fly"})," | ",e.jsx(s.code,{children:"blaxel"})," | ",e.jsx(s.code,{children:"codesandbox"})," | ",e.jsx(s.code,{children:"northflank"})," | ",e.jsx(s.code,{children:"aws"})," | ",e.jsx(s.code,{children:"gcp"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:["Where sandboxes run: local (this machine), cloud, your own cloud account (aws, gcp, modal; connect it first with ",e.jsx(s.code,{children:"cua cloud connect <provider>"}),"), or a contrib provider (e2b, daytona; needs a cua SDK built with contrib and the provider's key). Default: CUA_DEFAULT_ON, else ",e.jsx(s.code,{children:"cua config set default.on"}),", else local."]})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--kind"}

91),e.jsx(s.code,{children:"--kind"})]}),e.jsxs(s.td,{children:[e.jsx(s.code,{children:"auto"})," | ",e.jsx(s.code,{children:"container"})," | ",e.jsx(s.code,{children:"vm"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:["What kind of sandbox: container or vm. Default auto: Linux as a container unless the task or image says VM; Windows, macOS and Android are VM-only. Default: CUA_DEFAULT_KIND, else ",e.jsx(s.code,{children:"cua config set default.kind"}),"."]})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--runtime"}),e.jsx(s.code,{children:"--runtime"})]}),e.jsxs(s.td,{children:[e.jsx(s.code,{children:"auto"})," | ",e.jsx(s.code,{children:"gvisor"})," | ",e.jsx(s.code,{children:"runc"})," | ",e.jsx(s.code,{children:"qemu"})," | ",e.jsx(s.code,{children:"lume"})," | ",e.jsx(s.code,{children:"kubevirt"})]}),e.jsx(s.td,{}),e.jsxs(s.td,{children:["Which engine: gvisor or runc (local containers), qemu or lume (local VMs), gvisor or kubevirt (cloud containers, VMs). Implies its kind. Default auto: the SDK picks (CUA_DEFAULT_RUNTIME, else ",e.jsx(s.code,{children:"cua config set default.runtime"}),")."]})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--image"}),e.jsx(s.code,{children:"--image"})]}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Image for every task (overrides the task's setup_config.image): an OS alias (linux, windows, macos:tahoe), ghcr.io/trycua/<os>, any registry ref or a digest, or pool:<name> for an existing Fleet pool. Env: CUA_BENCH_IMAGE."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--cpu"}),e.jsx(s.code,{children:"--cpu"})]}),e.jsx(s.td,{children:"integer"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"vCPUs per sandbox."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--memory"}),e.jsx(s.code,{children:"--memory"})]}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"Memory per sandbox (e.g. 4096, 4G)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--claim-ttl"}),e.jsx(s.code,{children:"--claim-ttl"})]}),e.jsx(s.td,{children:"string"}),e.jsx(s.td,{}),e.jsx(s.td,{children:"cloud: how long a claim outlives a crashed run (default 15m; renewed while running)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--oracle"}),e.jsx(s.code,{children:"--oracle"})]}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Run the solution after setup."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--view"}),e.jsx(s.code,{children:"--view"})]}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Open the trace viewer when done."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--no-wait"}),e.jsx(s.code,{children:"--no-wait"})]}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Skip the interactive prompt (useful for SSH/CI testing)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--no-browser"}),e.jsx(s.code,{children:"--no-browser"})]}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"Print the Display: URL without opening it (also CUA_BENCH_NO_BROWSER=1)."})]}),e.jsxs(s.tr,{children:[e.jsxs(s.td,{children:[e.jsx("span",{id:"cb-interact--warm"}),e.jsx(s.code,{children:"--warm"})]}),e.jsx(s.td,{children:"boolean"}),e.jsx(s.td,{children:e.jsx(s.code,{children:"false"})}),e.jsx(s.td,{children:"cloud: keep one replica ready when the managed pool is first created."})]})]})]}),`
92`,e.jsx(s.p,{children:e.jsx(s.strong,{children:"Examples"})}),`
93`,e.jsx(s.figure,{"data-rehype-pretty-code-figure":"",children:e.jsx(s.pre,{tabIndex:"0","data-language":"bash","data-theme":"github-dark-default",children:e.jsx(s.code,{"data-language":"bash","data-theme":"github-dark-default",style:{display:"grid"},children:e.jsx(s.span,{dangerouslySetInnerHTML:{__html:`<span data-line=""><span style="color:#8B949E"># Open a task and its display, wait for Enter, then evaluate</span></span>
94<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> interact</span><span style="color:#A5D6FF"> ./tasks/hello_file_env</span></span>
95<span data-line=""> </span>
96<span data-line=""><span style="color:#8B949E"># Run variant 1 with its oracle and save a screenshot</span></span>
97<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> interact</span><span style="color:#A5D6FF"> ./tasks/hello_file_env</span><span style="color:#79C0FF"> --variant-id</span><span style="color:#79C0FF"> 1</span><span style="color:#79C0FF"> --oracle</span><span style="color:#79C0FF"> --screenshot</span><span style="color:#A5D6FF"> shot.png</span></span>
98<span data-line=""> </span>
99<span data-line=""><span style="color:#8B949E"># Resolve a task from a registry dataset</span></span>
100<span data-line=""><span style="color:#FFA657">cb</span><span style="color:#A5D6FF"> interact</span><span style="color:#A5D6FF"> click-button</span><span style="color:#79C0FF"> --dataset</span><span style="color:#A5D6FF"> cua-bench-basic</span></span>`}})})})}),`
101`,e.jsx(s.h2,{id:"exit-codes",children:"Exit codes"}),`
102`,e.jsxs(s.table,{children:[e.jsx(s.thead,{children:e.jsxs(s.tr,{children:[e.jsx(s.th,{children:"Code"}),e.jsx(s.th,{children:"Meaning"})]})}),e.jsxs(s.tbody,{children:[e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.co

102de,{children:"0"})}),e.jsx(s.td,{children:"Success: every variant completed."})]}),e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.code,{children:"1"})}),e.jsx(s.td,{children:"Failure: a variant failed or was cancelled, or the command could not run."})]}),e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.code,{children:"2"})}),e.jsx(s.td,{children:"Usage error: an unknown command, flag or value."})]}),e.jsxs(s.tr,{children:[e.jsx(s.td,{children:e.jsx(s.code,{children:"130"})}),e.jsx(s.td,{children:"Interrupted (Ctrl-C); running variants are cancelled."})]})]})]})]})}function i(n={}){const{wrapper:s}={...t(),...n.components};return s?e.jsx(s,{...n,children:e.jsx(d,{...n})}):d(n)}export{i as default,c as frontmatter};
103//# sourceMappingURL=run-CV-2srQ1.js.map
104
105//# chunkId=01a102ca-23d2-74b3-b081-b517399c563b