Newt app development errors

When a Newt app task stops, the app task message shows the error code · stage · problem · cause · next action together.
If the cause couldn't be confirmed it says "cause unknown" — it never guesses that permissions or the network are to blame.

This page lists, per code, what happened and what to do. Even if you can't open this page, you can carry on with the message's next action alone. The whole flow is in Build apps by chatting in Newt.

Three things to know first

Check the status of the Request Inbox task again

Find it by situation

Apps, permissions, and the feature

app_not_found The app wasn't found

Meaning — no app you can see matches the name or link you gave. The name differs, it belongs to another company, you lack access, or it was deleted. To protect other companies' information, it doesn't say which.

What to do — check the exact name with "show me the apps I can change", and check that the right company is selected in Newt. If you need access, ask the app owner.

app_ambiguous It can't tell which app

Meaning — several apps have the same or similar names. The Agent never writes to an arbitrary app; it shows you the candidates.

What to do — pick one of the candidates. Mentioning something distinctive, like the app's address, makes it certain.

app_create_forbidden You can't create apps

Meaning — you don't have permission to create new apps in this company. Nothing was created.

What to do — ask your company admin for permission to create apps. Alternatively, get write access to an existing app and change that one.

app_unsupported This app can't be changed in Newt

Meaning — Newt app development works only on apps that use AxHub's own Git repository. It doesn't work on apps connected to GitHub, AxHub plugin-type apps, apps whose repository is being migrated, or archived or suspended apps. The message says which reason applies.

What to do — change GitHub-connected apps the existing way (Quickstart). Resume a suspended app first (Manage deployments); if the repository is being migrated, ask again after it finishes.

permission_revoked You lost a permission during the task

Meaning — permissions are re-checked right before each stage, and one was missing. You're no longer a member of the company, or your write or deploy permission on the app was removed. Nothing ran from that stage on; saves and deploys that finished earlier are shown in the message.

What to do — get the permission back from the app owner, then say "continue the Request Inbox task".

capability_disabled The app development feature is off

Meaning — Newt app development isn't available in this company. Nothing was created or changed.

What to do — ask your company admin. Meanwhile, you can build apps the existing way in Quickstart.

agent_coding_not_allowed This Agent can't build or change apps

Meaning — before building an app or changing code, it checked whether this conversation and Agent can do development, and stopped because they can't. The message's problem and cause tell you which case it is. This check only runs when building an app or changing code. Listing apps, previewing a saved version, deploying, and changing settings aren't blocked by this code.

What to do — find the row in the tables below that matches the message's problem · cause, and do what that row says. In most cases, ask your default Newt (the default Agent you start with) the same thing.

When it stopped before a task started — this shows in the Agent's reply. It stopped before a task was created, so there's no new app and no saved work.

When it stopped after the task started, before coding — this shows in the app task message. Right before coding, it checked again, couldn't continue, and didn't start coding. The app and saved work stay as they are. If it was building a new app, the app and its template are already saved.

Tasks and goals

operation_in_progress This app already has a task running

Meaning — only one task runs per app at a time. A task paused waiting for your answer or an approval also counts as running. The Agent tells you which task is running.

What to do — wait for it to finish, or check its status with "how far along is the Request Inbox task?". If it's waiting for your answer, answer it. If you don't need it, say "stop the Request Inbox task" and then ask again.

goal_not_authorized That's outside the goal you gave

Meaning — the current stage isn't part of the goal you gave. For example, you only said "change it" and the Agent tried to deploy to production. That stage wasn't run. A task that ended because you asked it to stop, such as "cancel the deploy", shows cancelled instead of this code.

What to do — if you want it, state the goal clearly. For example: "Deploy Request Inbox to production".

goal_unverified Part of the work is on hold because the goal is unclear

Meaning — it isn't sure whether your last message changes the goal, so it put production deploys · publishing to the internet · setting changes on hold. Writing code, validation, saving, and the private preview continue.

What to do — say the scope you want in one sentence. For example: "deploy it to production too" or "only go as far as the preview".

cancelled The task was stopped

Meaning — the task ended because you asked it to stop ("cancel it", "don't deploy it"). If it stopped a production deploy, a visibility change or a setting change, the message's first line says which, for example "Stopped the production deploy as you asked". In-progress coding and any deploy that could be stopped were stopped, and a review this task requested that was still pending was withdrawn ("withdrew the review request"). A review approved in the meantime isn't withdrawn; the message says it was approved. Code already saved and deploys already finished aren't rolled back. The message says what remains.

What to do — to do it again, ask again. It continues from the last saved version.

temporarily_unavailable The task paused after a temporary problem

Meaning — this code appears when a task in progress paused, and when it stopped before a task started. What you say to try again differs between the two, so tell them apart by the message.

When a task in progress paused — a service the task needs didn't answer, so the same step was retried automatically. For example, the app repository (Git) account or permissions weren't ready yet, the coding workspace or deploy system briefly didn't respond, or another deploy kept using the preview environment. After 90 seconds a "retrying the … step automatically" message tells you the step and the cause once; if it still isn't resolved after 10 minutes, the task pauses at that step. Anything already saved or deployed stays as is, and the task keeps this app's in-progress slot.

What to do — in a little while, say "continue" in the same conversation. It retries from the paused step, and if it gets stuck again it tells you the same way (a notice after 90 seconds, a pause after 10 minutes). In another conversation, name the app: "continue developing Request Inbox". If it pauses again and again, give your admin the task number · stage · cause.

When it stopped before a task started — it couldn't finish a check needed before building or changing an app: it couldn't confirm settings, couldn't save a record, ran into another change, or hit a server error. The Agent's reply gives a one-line cause and an inquiry number. It stopped before a task was created, so there's no new app and no saved work.

What to do — there's no task to continue, so instead of "continue", ask the same thing again, as is, in a little while. For example, if you said "build a Request Inbox app", send that sentence again. If the same message keeps coming back, share the inquiry number · cause · time checked as described in When you ask for help.

repair_limit_reached It stopped while fixing things on its own

Meaning — validation failed and the Agent tried to fix it on its own, but it didn't pass within three tries, or the same error kept repeating, so it stopped. Code that didn't pass wasn't saved.

What to do — read the last output in the message, then ask again more specifically or with a smaller scope.

stale_revision The task's state changed in the meantime

Meaning — the Agent asked for the next step based on an older task state, but the task had already moved on. Nothing ran.

What to do — say "check the Request Inbox task status again" and the Agent continues from the latest state.

request_conflict The same request came in with different content

Meaning — a request with the same request number as one already accepted came in with different content. It was refused to prevent running twice. The request accepted first carries on.

What to do — check the status of the task accepted first. If you want different content, say the new content as a new request.

invalid_request The request couldn't be understood

Meaning — something the request needs is missing or malformed. For example, no app was settled on, or a secret was put where a setting value goes. Nothing ran.

What to do — say the app name and the result you want again in one sentence. Register secrets on the environment variables screen, not in the chat.

Keeping and saving code

source_conflict The app's code changed in the meantime

Meaning — the app's saved version changed after the task started, because another conversation or person saved first. The current result was built on the older version, so it wasn't saved over the newer one.

What to do — ask "redo it on the latest version of Request Inbox".

source_unsupported_file There's a file that can't be saved

Meaning — the result contains files that must not be saved: .env · .env.* files (.env.example is fine) or symlinks. The message names the files. Nothing was saved.

What to do — ask it to delete those files. Register secrets as environment variables, not in files.

source_too_large The code is too large

Meaning — it's over what one save can hold: up to 4,000 files · 24 MiB in total · 16 MiB per file, and under 32 MiB compressed. Nothing was saved.

What to do — ask it to take large images, data files, or build output out of the repository.

capture_failed The task's result couldn't be collected

Meaning — the changed code couldn't be collected from the workspace. The app's code is still the last saved version.

What to do — check the status of the same task, then ask for the same thing again.

cleanup_unknown Workspace cleanup couldn't be confirmed

Meaning — it couldn't confirm the coding workspace was cleaned up properly. AxHub keeps checking. Whether the result was kept is shown in the message's save status.

What to do — check the same task's status again in a little while. If it stays this way, tell your admin along with the task number.

git_save_failed It couldn't save to Git

Meaning — the result that passed validation was kept in AxHub, but couldn't be saved to the Git repository. The repository didn't respond for a moment or was read-only, or the Git repository's storage is nearly full. If storage is the cause, the message says so.

What to do — in a little while, ask "save Request Inbox again". It saves the kept result as-is. If the message says storage is full, ask your company admin to add space, then ask it to save again.

Asking "show me the deploy logs" in a conversation shows the stored deploy status and stages, plus the last lines a failed build left. Live build logs aren't available in conversations; check them on the app's deployment page in the AxHub web console.

git_conflict The branch it was saving to had changed

Meaning — at the moment of saving, the newt/dev branch wasn't at the expected version, because something else saved to that branch first. Nothing was overwritten, and the result is kept in AxHub.

What to do — ask "apply it again on the latest version of Request Inbox". If you don't know who changed the branch, check with the app owner.

Validation and builds

dependency_resolution_failed Packages couldn't be installed

Meaning — AxHub's install environment tried to write or update the lockfile to match package.json and install the packages, and that failed. Nothing was validated or saved. Typical causes:

• A package name or version that isn't on the npm registry
• A package given as a Git address, file path, URL, or workspace
• A package that needs to run scripts when installed
• The registry couldn't be reached, it didn't finish within 5 minutes, or the installed dependency tree is too large
• The current lockfile can't be updated to match package.json

What to do — check that the package names and versions you asked for are right, and ask it to switch to a version on the npm registry or build without that package. If it looks like a registry problem, ask it to continue the same task in a little while.

dependency_set_unavailable The package install environment isn't set up

Meaning — this AxHub has no package install environment configured. Until it does, only the default template's exact package set works, so an app whose packages were added or changed couldn't get its dependencies prepared. It isn't a problem with your app's code. Nothing was validated or saved.

What to do — tell your company admin along with the task number. To avoid waiting, ask "build it with the packages already there, no new packages". Once it's set up, say "continue the Request Inbox task".

validation_failed Validation didn't pass

Meaning — npm run build (or npm test) failed in the workspace. The message shows the command · exit code · last output. Code that didn't pass wasn't saved.

What to do — read the last output and ask "fix this error", or ask again with a smaller scope.

build_failed The deploy build failed

Meaning — workspace validation passed, but the build at deploy time (Dockerfile) failed. The code is saved. Development validation and the deploy build run in different environments, so this can happen.

What to do — find the cause with "show me the Request Inbox deploy logs" and ask it to fix it. Common causes are in A deploy shows as failed.

runner_unavailable A workspace can't be prepared right now

Meaning — there's no room to start a coding workspace, or it's under maintenance. The code didn't change.

What to do — ask for the same thing again in a little while.

When the coding environment is turned off — if the Agent's reply says "The coding environment can't be used" and "Coding is turned off in this environment", coding is turned off in the AxHub environment you're using, so this Agent didn't start developing the app. It stopped before a task was created, so there's no new app and no saved work.

What to do — ask again in a little while. If the same message keeps coming back, ask your default Newt the same thing.

Previews and deploys

staging_unavailable This app can't use a preview

Meaning — staging (the preview) can't be turned on for this app, because of company policy, the plan or the kind of app. It never quietly deploys to production instead of a preview. If this task already made the first private deploy a preview needs, the message says so.

When your company deploys straight to production without a preview environment (auto-approved production changes, a public workspace, or a plan without staging), Newt works like this:

• A request that includes deploying ("build it and deploy it to production") says it deploys without a preview, then deploys the saved version to production. If this task's first private deploy already has that saved version, it doesn't deploy again and only checks that the service responds.
• A request for a preview only finishes with the private production app link when this task's first private deploy already contains the change. Otherwise it stops with this code and leaves the production app as it is.

What to do — ask your company admin whether staging can be used. If you want a production deploy, say so clearly: "Deploy Request Inbox to production".

staging_busy Another deploy is using staging

Meaning — a web console or CLI deploy, or another Newt task, is using this app's staging. So they don't overwrite each other, the preview deploy waits and retries automatically. If staging stays in use for more than 10 minutes, the task pauses with temporarily_unavailable.

What to do — if you wait, it continues once that deploy finishes. If it paused, say "continue" after that deploy finishes. The other way around, if someone deploys to staging from the web console or CLI during a Newt preview, that deploy takes priority, and the Newt preview link then says a newer version exists.

first_deployment_failed The first private production deploy failed

Meaning — to prepare the new app's preview, it deployed privately to production for the first time, and that failed. The code is saved, and staging isn't on yet.

What to do — check the cause with "show me the Request Inbox deploy logs", ask it to fix it, then ask "preview it again".

preview_failed The preview deploy failed

Meaning — the staging deploy failed, or the service-response check after the deploy failed. The saved code and production are unchanged.

What to do — read the deploy logs and ask it to fix the problem. If the app takes a long time to first respond, see Health check failure codes.

deploy_failed The production deploy failed

Meaning — the deploy releasing to production failed. Which version production is on now is shown in the message's production status.

What to do — read the deploy logs and ask it to fix the problem. If you need to restore the service first, see roll back to the previous version.

review_pending Waiting for admin approval

Meaning — the production release or the visibility widening is waiting for an admin review. This isn't a failure. The Agent never approves on anyone's behalf or shows it as approved.

What to do — just wait. When there's a result, an app task message tells you and the task continues. If you don't want to wait, say "cancel the deploy that's waiting for review" in the conversation. The review request is withdrawn and the task ends as cancelled; you don't need to withdraw it in the web console. The review process is in Visibility & access.

App settings

config_failed The setting couldn't be changed

Meaning — the setting change you asked for (create a table, add a column, add or change an environment variable, set or clear a staging-only value) was refused, or the settings changed in the meantime and it conflicted. Each setting is handled separately, so the message lists what was applied and what wasn't. Settings that weren't applied are unchanged.

What to do — check which settings weren't applied, review the current settings, then ask for just those again. If it was refused for lack of permission, ask the app owner. Environment variable rules are in Set environment variables.

Usage limits and unconfirmed results

quota_exceeded A usage limit was exceeded

Meaning — a usage limit such as the run queue or storage was exceeded, so it can't continue. Nothing new was run.

What to do — wait for other tasks to finish, or check the limit with your company admin.

credit_insufficient Not enough credit

Meaning — there isn't enough credit to start coding or a deploy. Nothing new was run.

What to do — check credit with your company admin. Once topped up, say "continue the Request Inbox task".

effect_unknown The result couldn't be confirmed

Meaning — a save or deploy was requested but no result came back, so it doesn't yet know whether it happened. AxHub looks the result up again by the same request number to confirm. It never runs the same thing again before that's confirmed.

What to do — don't ask again from scratch. Ask "check the status of that task again". If it still can't be confirmed, tell your admin along with the task number.

internal An unexpected error

Meaning — an unexpected error happened inside AxHub. The cause couldn't be confirmed.

What to do — check the same task's status again. If it keeps happening, give your admin the task number · stage · time checked.

The browser keeps asking me to sign in

Symptom — opening a preview or production link keeps showing the AxHub sign-in screen, or after signing in it says you can't access it.

Cause — a private app opens only for people signed in with a company account who have access to it. The browser is signed in with a different account, or the account that opened the link doesn't have access to the app.

Fix — sign in to AxHub in the browser with the same account as Newt, then click the link again. If it's a different account, sign out and sign in again. To show it to other people, widen its visibility (Visibility & access).