Skip to main content

Upgrade to Stim 2.0

Stop workspaces and switch versions​

  1. With 1.40 still installed, run stim status to find your running workspaces. From each workspace, run stim stop. Repeat for each Stim home you use. Stopping first also prevents duplicate web logs from a Chrome instance that a 1.x supervisor started.
  2. Upgrade the CLI and any project pins or CI inputs that still select 1.x. Keep the same STIM_HOME so 2.0 can read your settings, workspace records, and device ownership. Review legacy settings before restarting.
  3. Update Stim Desktop and any remote workers you use, then run stim doctor in each app directory and restart the workspaces you need. Run stim doctor only in checkouts whose project code you trust. Verify the connection and the next build on each worker before relying on offload.
Command examples

Commands use stim. Use npx stim if you run without a global installation. For step 1, use npx stim@1.40.0 stop if your workspace was running 1.40 and you need to invoke it explicitly.

The resident daemon and engine​

The per-home stimd runs workspace services such as Metro, managed tunnels, and device logs. daemon.mode controls automatic maintenance. Even in off mode, a command that needs one of those workspace services can start stimd.

A 1.40–1.42 maintenance daemon can keep its claim after all workspaces stop. The 2.0 daemon takes over automatically when it can verify that the old maintenance daemon owns the claim. If the claim remains held, follow the refusal from stim start or stim serve. Check the reported PID before using the suggested kill <pid> or, on Windows, taskkill /PID <pid> /F. Retry once the holder has stopped. Metro supervisors, log collectors, and devices already running are separate processes and can outlive the old daemon. stim start can adopt a verified dev server. If it reports a port held by another process, stop the old workspace first.

An npm installation can run directly. If you want a managed machine engine, use stim engine install once from the upgraded CLI. stim engine status shows the active build and channel. On an existing managed engine, stim engine update follows its current channel and stim engine rollback selects its previous build. See engine and service updates.

2.0 workers use peer compatibility version 3 and build compatibility version 6. If stim remote update <machine> can reach and update the worker, use it. If the old worker cannot negotiate an update, upgrade it locally and reinstall its service with stim service install, retaining its service options. On Windows, reinstall the service even after an engine update so the scheduled task runs the new build. If an old server still holds the serving port, stop that server or choose a different port with stim serve --port <port> and point the clients at it.

If stim service install refuses because a Stim daemon serves the service's port outside the service's job and install cannot prove it belongs to the service, Stim Desktop, stim serve, or a Stim build older than this one started that daemon. Stop it once its work drains with the kill <pid> it prints, then run the same stim service install command again. Stopping it also stops what Stim Desktop or that command is using; its dev servers keep running and the next daemon takes them over.

If stim service uninstall reports that the service's daemon is still listening after an engine handoff, the service definition and any route it owns are kept. Let that daemon finish its work, stop the reported daemon, and retry stim service uninstall --label <label>. Ubuntu uninstall check

The live 1.40-to-2.0 upgrade checks kept the existing Metro process and verified the same bundle on macOS, Linux, and Windows. macOS and Linux exercised legacy daemon takeover. Windows started its first resident daemon while the 1.40 Metro was still running. Upgrade results

Update the phone app alongside Desktop if you use it. Upgrades with paired clients and mixed-version app and engine combinations are still being verified. App and engine compatibility checks

Let an agent migrate​

After the upgrade, stim guide agent prints a STATUS line while the home or project still holds something an earlier Stim left: settings or environment variables that 2.0 does not read, a daemon from 1.x, a stim 1.x or bundled @stim-cli/* dependency in package.json, or a doctor check last run with 1.x. The line, the refusals for leftover settings and the old daemon, and the matching stim doctor findings link the agent to For agents. The line clears once nothing is left.

Follow https://stim.appandflow.com/docs/migrate-to-2#for-agents to migrate this Stim home and project to the installed Stim. Ask me before you change my shell, CI or service environment, edit ~/.stim/config.json, keep or change a default, reinstall a service, update another machine, or stop a process.

First native build after the upgrade​

  • Native cache. The 2.0 fingerprint uses new keys. Expect one cold build per distinct native fingerprint whose only cache entry was made by 1.40. Matching 2.0 builds can reuse the new entry across worktrees. If a worktree warmed by 1.40 reports a lockfile newer than its npm install stamp, run npm install in the directory named by the warning so future builds can be cached. Native artifact cache
  • Expo-generated native directories. Preserve manual edits before the first native run. For an untracked ios/ or android/ with a readable prebuild record from an older Stim, 2.0 runs Expo prebuild once for that platform in each generated worktree. A local run prebuilds in place. An offload worker can regenerate its own copy with --clean. If STIM_PREBUILD_FAILED says the record is unreadable, preserve your edits, move only the generated directory named by the error aside, and rerun. Leave tracked and hand-maintained native directories in place. React Native and Expo inputs
  • fingerprint.config.js options. 1.40 passed the file's options to @expo/fingerprint, so preset: 'strict' hashed the contents of native dependencies. 2.0 applies only extraSources (not when cache.fingerprintInputs is set) and keys a registry package by its identity. It does not refuse or warn at build time, and stim doctor lists each ignored option. If your file sets other options, a native dependency edited in place, such as a hand edit in node_modules, reuses a cached build. Move needed inputs to extraSources or cache.fingerprintInputs, or commit the change as a patch. For a temporary native edit, run stim ios --no-build-cache --remote-build local (or stim android).
  • Build credentials. An offloaded build removes Stim's own access and setup credentials from the worker's build environment. A build whose embedded config or dependency reads a private Stim credential, or that enumerates the project environment while one is present, is uncached. Remove the credential from the build environment to get caching back. Project build tools can still print the environment values they receive.
  • Owned iOS simulators. ios.simslim now defaults to on, including for previously created Stim simulators at their next boot. Set ios.simslim to off for a project that needs stock simulator services. Owned devices

Change cleanup scripts before running them​

stim gc now deletes. Update scripts that used it to preview, including scripts that parse JSON.

stim gc --dry-run
stim gc --dry-run --json
stim gc --dry-run --pressure high
  • --delete is gone. Use plain stim gc to apply the measured level after reviewing a dry run.
  • --cache, --older-than, and --worktrees are gone. Review the maintenance settings for age and worktree policy. Use --pressure normal, high, or critical to choose a level for one pass.
  • normal includes age and dormancy cleanup, such as eligible worktree dependencies, parked devices, and unused iOS simulator runtimes. It can also clear dependencies from a dormant source checkout. high also applies pressure cleanup to idle workspaces and trims caches. critical can empty shared Stim caches and clean other tools' caches on the default home. An explicit --pressure critical runs the last stage even if space is recovered earlier. Full cleanup rules
  • stim worktree remove and the automatic sweep keep a changed project.pbxproj and any ios/Podfile.lock change they cannot prove came from pod install. The proof is Pods/Manifest.lock holding the same text, or, when gc deleted Pods, the lock hash gc recorded. worktree remove refuses a changed project.pbxproj, and a tracked file marked assume-unchanged or skip-worktree, naming the file. Commit or discard the change, or pass --force.
  • Review configured maintenance.discoverRoots in the dry run. A root you explicitly configured brings eligible checkouts under the cleanup rules even if the root is in a home dot-folder or the system temp directory. Discovery scope

If you relied on the 1.40 defaults, review these values. An explicitly configured value still takes precedence.

Setting1.402.0
idle.dormantDays31
maintenance.olderThanDays73
disk.minFreeGb2040
disk.criticalFreeGb620
ios.simslimoffon

These commands keep the 1.40 dormancy and age thresholds.

stim settings set idle.dormantDays 3 --scope machine
stim settings set maintenance.olderThanDays 7 --scope machine

Legacy settings​

Use stim settings and stim doctor to find leftover keys. Old keys can be reported as unknown or inert, and previously exported variables can be ignored. Migrate the setting in the layer that defines it. Remove unknown keys from the config file that defines them, and remove old exports from your shell, service, or CI environment.

Unknown names​

The CLI gives no replacement hint for these names.

Old keyWhat to use in 2.0
remote.buildModeremote.build / STIM_REMOTE_BUILD. Map auto to auto, force to remote, and off to local. Remove STIM_REMOTE_BUILD_MODE.
share.webrtcOnly a next build could set it, with STIM_WEBRTC_STUN; 1.40 had neither. Remove both. Shared web video uses the share socket.
devices.reclaimIdleMinutesRemove it. While a run waits to boot, Stim can reclaim the longest-idle device of another workspace idle for at least two minutes.
daemon.enableddaemon.mode. Map true to on and false to off.
caches.injectMetroStoreoptimizations.metroSharedCache.

The pressure ladder uses fixed rules for the choices below. Remove these keys and any matching STIM_ exports, such as STIM_BUDGET_MAX_COMMITTED_MEMORY_GB, STIM_BUDGET_MAX_LIVE_WORKSPACES, and STIM_MAINTENANCE_PRESSURE_CHECK_MINUTES.

  • budget.maxCommittedMemoryGb, budget.maxLiveWorkspaces
  • maintenance.pressureCheckMinutes, maintenance.sizeCheckMinutes, maintenance.maxLoadPerCore, maintenance.memoryPressureLevel, maintenance.memoryWarningMinutes, maintenance.minAvailableMemoryGb, maintenance.capTargetPercent, maintenance.idleCheckMinutes, maintenance.urgentFreeDiskGb, maintenance.pressureProtectMinutes

Names that can still block a command​

start, ios, android, web, macos, revive, and gc refuse when a removed environment variable in the table is set to a nonempty value. Set the replacement and remove the old export. They also refuse when old config sets maintenance.mode to off or report, or sets either old local idle timeout to 0. On these config refusals, follow the printed stim settings steps to preserve the intended behavior.

Old key or environment variableReplacement
metro.idleStopMinutes, devices.idleShutdownMinutesidle.minutes stops the dev server and owned devices together.
maintenance.deviceIdleDays / STIM_MAINTENANCE_DEVICE_IDLE_DAYSidle.dormantDays.
budget.minFreeDiskGb / STIM_BUDGET_MIN_FREE_DISK_GBdisk.minFreeGb / STIM_DISK_MIN_FREE_GB.
budget.hardFloorDiskGb / STIM_BUDGET_HARD_FLOOR_DISK_GBdisk.hardFloorGb / STIM_DISK_HARD_FLOOR_GB.
maintenance.criticalFreeDiskGb / STIM_MAINTENANCE_CRITICAL_FREE_DISK_GBdisk.criticalFreeGb / STIM_DISK_CRITICAL_FREE_GB.
maintenance.mode / STIM_MAINTENANCEdaemon.mode / STIM_DAEMON. Keep the same on, report, or off value.

maintenance.idleStopHours still applies to hosted device sessions. stim doctor can warn about an old 0 setting because local workspace dormancy now follows idle.dormantDays. Review both values if you used it to disable local cleanup.

Update CLI automation and integrations​

Every command now runs from one operation registry. A subcommand refuses an option it does not define, for example stim pairings grant <id> --json. The 1.40 stim-server pair and stim-server devices commands are now stim pair and stim pairings (list, grant, revoke), and stim-server log is stim pairings log. Remove the option from scripts that passed one.

The server and CI runner ship in stim. Use stim serve in place of stim-server. Use stim pair, stim pairings, stim service, and stim setup for those workflows, and replace stim-ci with stim ci. For the GitHub Action, set version to 2.0.0 or later, or stay on an older action ref with its supported version. If cli-path points into a Stim checkout, build it and use packages/stim-cli/bin/stim.mjs.

Packages that stop publishing​

@stim-cli/core, cache, agent-driver, agent-driver-agent-device, metro, expo-build-cache, ci, transport, transport-tailscale, transport-local and server are bundled into stim and no longer publish. Versions already on npm stay installable but get no updates, and the stim-server and stim-ci binaries are gone. @stim-cli/cache-s3 stays published. If your code imports one of these packages, import from stim, stim/ci or stim/plugin instead.

  • @stim-cli/metro. Remove sharedCacheStores and the package from metro.config.js and your dependencies. Stim sets Metro's shared transform store itself when it starts Metro; optimizations.metroSharedCache controls it.
  • @stim-cli/expo-build-cache. Remove it from your dependencies and from the Expo buildCacheProvider config. Stim's engine no longer reads a cache it wrote. Build with stim ios and stim android.

JavaScript and TypeScript​

1.402.0
StimOptions.buildCachebuildCacheDir for the directory passed to createStim; also use it in stim/ci options. The run/build buildCache boolean stays. stim ci --build-cache <path> sets a directory. stim ios and stim android keep --no-build-cache to skip a cache read.
Run option waitwaitSeconds accepts a positive whole number or false. Omit it to use the default.
diagnostics({ errors })diagnostics({ errorsOnly }) with the same boolean.
gc()previewGarbage() for a preview. Call collectGarbage() to apply cleanup.

Update every call site. A temporary compatibility check (C33) refuses the old spellings with STIM_BAD_ARG and names the replacement. Both garbage methods take { pressure } in place of olderThan, cache, worktrees, and delete. previewGarbage() returns the report directly, with actions and blocked describing the preview and an empty results array. collectGarbage() returns { report, exitCode }. Read report.results for the outcomes of actions it applied. Programmatic API

StimLogsOptions.sources and minLevel now use the LogSource and LogLevel unions. Treat operation errors as rejected promises. Import supported functions from stim or stim/ci. The bundled server's stim/internal/* paths and APIs marked @internal have no public stability guarantee. Replace deep imports like stim/api-self and stim/engine-state with public operations where available. If you register cache entries with stim/cache-manifest, use register(entry) and unregister(dir). The manifest readers and custom manifest path argument were removed.

Plugins​

The stable 2.x plugin contract is cache-provider. A provider declares the kind integer in its package manifest and exports CacheProviderModule.apiVersion. Make builds.resolve and builds.store return promises to satisfy the 2.0 types. Stim still awaits synchronous results at runtime.

transport and agent-driver are experimental and bundled. Use the public types, contract kits, and loadCacheProvider from stim/plugin. Replace cacheProviderContractChecks with runCacheProviderContract or runContract, and remove CacheProvider.metro. The transport and agent-driver host loaders, their internal helpers, and cache-tier utilities are gone from stim/plugin. Plugin contracts and testing

For agents​

Follow this section when stim guide agent prints Not migrated from an earlier Stim, or when a refusal or a stim doctor finding links here. Work through the steps in order. A step marked Ask first changes the user's machine, environment, defaults or another machine: describe the change and wait for the user's answer. When you are done, run stim guide agent again; its STATUS line clears once nothing is left.

1. Settings​

Check: list each key Stim does not read with the scope and file that hold it, without its value, which can be a secret:

stim settings --json 2>/dev/null | node -e "let s = ''; process.stdin.on('data', (d) => (s += d)).on('end', () => { for (const { key, scope, file } of JSON.parse(s).unknown) console.log(key, scope, file); })"

Do not print the whole payload, and send its errors to /dev/null as above, since a validation error can quote a value. Rerun stim settings without the filter only if the user asks. stim doctor rates each leftover.

Migrate:

  • When start, ios, android, web, macos, revive or gc refuses, its printed stim settings set and stim settings unset commands are the plan. The refusal leaves out the set when the replacement is already configured. Run each unset with the --scope of the occurrence it removes, as the bullets below say, and for maintenance.mode check the machine layer first.
  • Otherwise, check the replacement before you touch it: stim settings get <replacement> --json shows its origin, a layer, env for its environment variable, or default. A 1.40 home can hold an explicit replacement, such as daemon.mode off, next to an inert old key, such as maintenance.mode on.
    • If the origin is a layer or env, keep the replacement as it is and remove only the old key, as the table says.
    • If the origin is default, the old key's value would change what Stim does. Ask first whether to carry it over with stim settings set <replacement> <value>, adding --scope <scope> of the old key when the replacement accepts that layer. Then remove the old key.
  • Handle each reported occurrence on its own: decide its replacement, then remove only that one with stim settings unset <key> --scope <scope>, using the scope the inventory reported. Without --scope, unset removes the key from every layer at once, including a committed .stim.json that has its own decision to make.
  • Ask the user which value each layer should keep before you remove anything when the same retired key is reported in more than one file, or when the replacement comes only from another layer or from your shell. When one scoped unset would touch several files, edit the approved file directly instead: a --scope committed unset removes the key from every committed .stim.json, the app's and the repository's.
Old keyReplacementRemove the old key with
maintenance.modedaemon.mode; for the machine occurrence, see the machine-layer check belowstim settings unset maintenance.mode --scope <scope>
metro.idleStopMinutes, devices.idleShutdownMinutesidle.minutesstim settings unset <key> --scope <scope>
maintenance.deviceIdleDaysidle.dormantDaysstim settings unset <key> --scope <scope>
budget.minFreeDiskGbdisk.minFreeGbstim settings unset <key> --scope <scope>
budget.hardFloorDiskGbdisk.hardFloorGbstim settings unset <key> --scope <scope>
maintenance.criticalFreeDiskGbdisk.criticalFreeGbstim settings unset <key> --scope <scope>
remote.buildModeremote.build: auto to auto, force to remote, off to locala file edit
daemon.enableddaemon.mode: true to on, false to offa file edit
caches.injectMetroStoreoptimizations.metroSharedCachea file edit
devices.reclaimIdleMinutes, share.webrtcnonea file edit
budget.maxCommittedMemoryGb, budget.maxLiveWorkspaces, maintenance.pressureCheckMinutes, maintenance.sizeCheckMinutes, maintenance.maxLoadPerCore, maintenance.memoryPressureLevel, maintenance.memoryWarningMinutes, maintenance.minAvailableMemoryGb, maintenance.capTargetPercent, maintenance.idleCheckMinutes, maintenance.urgentFreeDiskGb, maintenance.pressureProtectMinutesnonea file edit
  • A file edit: stim settings unset refuses these names, because 2.0 no longer registers them. Remove the property from the file that stim settings --json reports for it: .stim.json for the committed scope, and ~/.stim/config.json (or $STIM_HOME/config.json) for machine, under projects["<path>"].settings for workspace and repos["<git dir>"].settings for repo. Ask first before editing the config file, and edit it only while no Stim command runs. Remove any other unknown key the same way after asking the user what it was for.
  • Keep maintenance.idleStopHours: hosted device sessions still read it.
  • While the machine config holds maintenance.mode off or report and neither STIM_DAEMON nor a machine daemon.mode is set, the daemon stays at report and acts on nothing. Setting either one, to any value, ends that hold.
  • Before you remove the machine occurrence of maintenance.mode when it is off or report, read the machine layer on its own with stim settings get daemon.mode --scope machine --json. A STIM_DAEMON in your shell does not reach Stim Desktop, stim serve or other shells. When the machine layer has no valid daemon.mode (its value is null or not on, report or off), ask first which machine policy to keep, off, report or on, and run stim settings set daemon.mode <value> --scope machine before stim settings unset maintenance.mode --scope machine. Do this even when a refusal or your shell's STIM_DAEMON suggests only the unset.

2. Environment​

Check: list the names of the STIM_ variables, never their values, which can hold secrets such as passwords and access tickets:

node -e "console.log(Object.keys(process.env).filter((k) => k.startsWith('STIM_')).join('\n'))"

Compare the names with the table. STATUS names the ones set in your own shell. The user's shell profile, CI and service definitions can also export them; read those only with the user's permission, and never print a value.

Migrate (ask first): the user changes each export where it is set. Leave other STIM_ variables alone; for example STIM_MAINTENANCE_SWEEP_HOURS and STIM_MAINTENANCE_IDLE_STOP_HOURS are still read.

Retired exportWhat to do
STIM_MAINTENANCERename it to STIM_DAEMON, same value.
STIM_MAINTENANCE_DEVICE_IDLE_DAYSRun stim settings set idle.dormantDays <value>, then remove the export; idle.dormantDays has no variable.
STIM_BUDGET_MIN_FREE_DISK_GBRename it to STIM_DISK_MIN_FREE_GB.
STIM_BUDGET_HARD_FLOOR_DISK_GBRename it to STIM_DISK_HARD_FLOOR_GB.
STIM_MAINTENANCE_CRITICAL_FREE_DISK_GBRename it to STIM_DISK_CRITICAL_FREE_GB.
STIM_REMOTE_BUILD_MODERename it to STIM_REMOTE_BUILD: auto to auto, force to remote, off to local.
STIM_WEBRTC_STUN, STIM_BUDGET_MAX_COMMITTED_MEMORY_GB, STIM_BUDGET_MAX_LIVE_WORKSPACES, STIM_MAINTENANCE_CAP_TARGET_PERCENT, STIM_MAINTENANCE_IDLE_CHECK_MINUTES, STIM_MAINTENANCE_MAX_LOAD_PER_CORE, STIM_MAINTENANCE_MEMORY_PRESSURE_LEVEL, STIM_MAINTENANCE_MEMORY_WARNING_MINUTES, STIM_MAINTENANCE_MIN_AVAILABLE_MEMORY_GB, STIM_MAINTENANCE_PRESSURE_CHECK_MINUTES, STIM_MAINTENANCE_PRESSURE_PROTECT_MINUTES, STIM_MAINTENANCE_SIZE_CHECK_MINUTES, STIM_MAINTENANCE_URGENT_FREE_DISK_GBRemove it; nothing replaces it.

The first five refuse start, ios, android, web, macos, revive and gc while set. 2.0 ignores the rest without a warning.

3. Changed defaults​

Ask first whether to keep the 1.40 values, and apply the answer before step 4: the first stim start that takes the old daemon over runs cleanup under the values then in force. 2.0 cleans up sooner by default: idle.dormantDays 1 (was 3), maintenance.olderThanDays 3 (was 7), disk.minFreeGb 40 (was 20), disk.criticalFreeGb 20 (was 6), and ios.simslim on (was off). To keep a 1.40 value, run stim settings set <key> <value> --scope machine.

4. Old processes​

Do this after steps 1 to 3: stim start refuses while a refusing setting or export is left, and the takeover runs cleanup under the defaults step 3 settled.

Check: stim status. It lists the live workspaces of this Stim home, and the line Daemon: pid <n> from an older stim holds the daemon claim means a 1.x daemon still runs.

Migrate (ask first): a shared home can hold other people's or other agents' active work. List the live workspaces and the old daemon to the user, and wait for an answer before you stop anything.

  • With the user's go, run stim stop in each workspace they name that a 1.x Stim started before the upgrade.
  • When the daemon line says the next command that starts stimd stops it, the next stim start takes the old daemon over; run it only after the user agrees. Otherwise, with the user's go, run the stop command the line prints (kill <pid>, or taskkill /PID <pid> /F on Windows).

5. Project files and scripts​

Check: from the repository root, git grep -nE '@stim-cli/|stim-server|stim-ci|stim gc|gc\(|"stim": *"', and review each stim dependency range in package.json.

Migrate:

In 1.40In 2.0
stim gc, stim gc --json (they only previewed)stim gc --dry-run, stim gc --dry-run --json. Plain stim gc now deletes.
stim gc --deletestim gc, after reviewing stim gc --dry-run.
stim gc --cache, --older-than, --worktreesNo replacement with the same scope: --pressure normal, high or critical runs a whole cleanup pass. Change the script to stim gc --dry-run, and ask first before any stim gc that deletes.
stim-serverstim serve, stim pair, stim pairings.
stim-cistim ci.
the GitHub Action pinned to a 1.x ref, with version below 2.0.0Move the uses: ref to a 2.x release tag or commit and set version to 2.0.0 or later in the same change. Only the 2.x action runs stim ci; a new version under the v1.40.0 action breaks the workflow.
a stim range that admits no 2.x versionA 2.x range, then install.
@stim-cli/metroRemove it and sharedCacheStores from metro.config.js and the dependencies.
@stim-cli/expo-build-cacheRemove it from the dependencies and from the Expo buildCacheProvider config.
@stim-cli/ci/artifacts (diagnosticArtifactFiles, copyDiagnosticLogs)stim/ci/artifacts, with the same names.
other @stim-cli/* importsImport from stim, stim/ci or stim/plugin.
@stim-cli/cache-s3 1.xKeep the name and upgrade it to a 2.x release: 1.40.0 lacks the plugin manifest 2.0 reads. Then check that stim plugin list shows it.
StimOptions.buildCache, run option wait, diagnostics({ errors }), gc()buildCacheDir, waitSeconds, diagnostics({ errorsOnly }), and previewGarbage() or collectGarbage(). See JavaScript and TypeScript.

6. Services and remote machines​

Check: stim service status on this machine, and stim remote list for the workers it uses.

Migrate (ask first):

  • A service that a 1.x Stim installed: run stim service install with its existing options. stim service status prints the exact command on its stale: line when the LaunchAgent is out of date.
  • A worker: stim remote update <machine>. A stock 1.40 worker cannot take that update. On the worker itself, install the target Stim, reinstall its service with stim service install and the same options, then check the connection with stim remote list and the next offloaded build.

7. First build​

Expect one cold native build per fingerprint. An untracked ios/ or android/ that a 1.x Stim generated is prebuilt again once: ask first whether it holds manual edits to preserve. stim doctor lists the fingerprint.config.js options 2.0 ignores.

8. Verify​

Run stim doctor, then stim guide agent. STATUS no longer lists leftovers.