Templates
Error triage
Decide which team owns a new error tracker issue, how much it hurts users, and whether the latest release caused it.
Decide which team owns a new error tracker issue, how much it hurts users, and whether the latest release caused it. Pattern: fan_out. Model: jev-1.13.0. Outage rule: review.
When to use it
- New issues from your error tracker sit unassigned because nobody knows whose they are.
- An engineer can tell the owner and the impact from the title, stack trace and release notes in a few seconds.
When not to use it
- You need a root cause analysis or a fix. That takes longer than 10 seconds and is text generation.
- Ownership follows file paths exactly, such as a CODEOWNERS file. Look it up in code.
- The impact is a number you already have, such as affected users per hour. Threshold it in code.
Questions
| Id | Type | Asks | Answers | Gating |
|---|---|---|---|---|
owning_team | choice | Which team owns the code that raised issue, judged from its title, culprit and stack trace? | web_frontend, checkout_and_billing, search, platform, none_of_these | yes |
user_impact | score | How badly does issue hurt the people using the product when it happens? | 4 levels | yes |
new_regression | noul | Is issue likely caused by one of the changes shipped in the latest release, listed in release.changes? | true or false | no |
Reading the result
- Replace the
owning_teamoptions with your own teams, and keepnone_of_these. - Assign only when
overallActionisauto. Routepagegoes to on-call,regressionto the release owner,backlogto the owning team's queue. release.changesshould list the changes in the latest release, one line each. The regression question compares the issue against that list, not against dates.
Spec
Save it as bandwise/sets/error-triage.json and edit it for your data.
{
"schemaVersion": 1,
"model": "jev-1.13.0",
"input": {
"schema": {
"type": "object",
"required": [
"issue",
"release"
],
"properties": {
"issue": {
"type": "object",
"required": [
"title",
"culprit",
"stack"
],
"properties": {
"title": {
"type": "string"
},
"culprit": {
"type": "string"
},
"stack": {
"type": "string",
"maxLength": 8000
},
"url_path": {
"type": "string"
}
}
},
"release": {
"type": "object",
"required": [
"version",
"changes"
],
"properties": {
"version": {
"type": "string"
},
"changes": {
"type": "array",
"items": {
"type": "string"
}
}
}
}
}
}
},
"stages": [
{
"id": "triage",
"questions": {
"owning_team": {
"type": "choice",
"instructions": "Which team owns the code that raised `issue`, judged from its title, culprit and stack trace?",
"criteria": {
"web_frontend": "Browser code: pages, components, client-side state.",
"checkout_and_billing": "Carts, payments, taxes, invoices and subscriptions.",
"search": "Search indexing, ranking and query handling.",
"platform": "Auth, databases, queues, deploys and shared infrastructure.",
"none_of_these": null
},
"meta": {
"label": "Owning team"
}
},
"user_impact": {
"type": "score",
"instructions": "How badly does `issue` hurt the people using the product when it happens?",
"criteria": [
"None. Users do not notice.",
"Minor. A cosmetic glitch or a retry that works.",
"Major. A feature fails for the user, with a workaround.",
"Blocking. Users cannot finish a core task such as signing in or paying."
],
"meta": {
"label": "User impact"
}
},
"new_regression": {
"type": "noul",
"instructions": "Is `issue` likely caused by one of the changes shipped in the latest release, listed in `release.changes`?",
"criteria": {
"true": "The stack trace or culprit points at code one of the listed changes touched.",
"false": "The failing code is unrelated to every listed change."
},
"meta": {
"label": "New regression"
}
}
}
}
],
"policies": {
"owning_team": {
"type": "choice",
"gating": true,
"thresholds": {
"high": 0.65,
"medium": 0.4
},
"actions": {
"high": {
"kind": "auto"
},
"medium": {
"kind": "review"
},
"low": {
"kind": "fallback",
"config": {
"kind": "value",
"value": "none_of_these"
}
}
}
},
"user_impact": {
"type": "score",
"gating": true,
"thresholds": {
"high": 0.6,
"medium": 0.35
},
"actions": {
"high": {
"kind": "auto"
},
"medium": {
"kind": "review"
},
"low": {
"kind": "review"
}
}
},
"new_regression": {
"type": "noul",
"gating": false,
"noul": {
"trueAt": 0.75,
"falseAt": 0.25,
"reviewMargin": 0.1
},
"actions": {
"high": {
"kind": "auto"
},
"medium": {
"kind": "auto"
},
"low": {
"kind": "auto"
}
}
}
},
"routes": [
{
"when": {
"q": "user_impact",
"eq": 3
},
"output": "page"
},
{
"when": {
"q": "new_regression",
"eq": true
},
"output": "regression"
}
],
"defaultRoute": "backlog",
"savings": {
"comparatorModel": "claude-haiku-4-5",
"estOutputTokensPerQuestion": 60,
"kind": "decision"
},
"onUnavailable": "review"
}Example states
The expected outcome is what a person would decide. It is not a recorded model answer.
Tax lookup failing at checkout
Expected: checkout_and_billing, blocking, a regression from the tax change: page.
{
"issue": {
"title": "TypeError: Cannot read properties of undefined (reading 'rate')",
"culprit": "checkout/tax.ts in computeTax",
"stack": "at computeTax (checkout/tax.ts:41)\nat buildOrder (checkout/order.ts:88)\nat POST /api/checkout",
"url_path": "/checkout"
},
"release": {
"version": "2026.9.3",
"changes": [
"checkout: switch tax calculation to the new rates service",
"search: add typo tolerance",
"deps: bump the image resizing library"
]
}
}Avatar image fails to load
Expected: web_frontend, minor, not related to the release: backlog.
{
"issue": {
"title": "Failed to load resource: avatar.png 404",
"culprit": "components/Avatar.tsx",
"stack": "at Avatar (components/Avatar.tsx:12)",
"url_path": "/settings/profile"
},
"release": {
"version": "2026.9.3",
"changes": [
"checkout: switch tax calculation to the new rates service",
"search: add typo tolerance",
"deps: bump the image resizing library"
]
}
}Borderline cases
One case near the line for each question. Use them to test your wording before you trust the thresholds.
owning_team
A database timeout raised inside search code: platform owns the database, search owns the query.
{
"issue": {
"title": "QueryTimeout: statement timeout after 30000ms",
"culprit": "search/indexer.ts in reindexBatch",
"stack": "at pool.query (db/pool.ts:20)\nat reindexBatch (search/indexer.ts:64)"
},
"release": {
"version": "2026.9.3",
"changes": [
"checkout: switch tax calculation to the new rates service",
"search: add typo tolerance",
"deps: bump the image resizing library"
]
}
}user_impact
Search results come back without typo tolerance: the feature degrades but still works.
{
"issue": {
"title": "Warning: fuzzy matcher unavailable, falling back to exact match",
"culprit": "search/query.ts",
"stack": "at runQuery (search/query.ts:102)",
"url_path": "/search"
},
"release": {
"version": "2026.9.3",
"changes": [
"checkout: switch tax calculation to the new rates service",
"search: add typo tolerance",
"deps: bump the image resizing library"
]
}
}new_regression
The failing code sits next to a changed library, but the error is in code the release did not touch.
{
"issue": {
"title": "Error: unsupported image format 'heic'",
"culprit": "media/upload.ts",
"stack": "at detectFormat (media/upload.ts:33)\nat resize (node_modules/image-lib/index.js:210)",
"url_path": "/upload"
},
"release": {
"version": "2026.9.3",
"changes": [
"checkout: switch tax calculation to the new rates service",
"search: add typo tolerance",
"deps: bump the image resizing library"
]
}
}Try it
Save an example state as state.json, then run the spec locally. Local mode makes no network call and needs no key; answers are synthetic unless a recorded fixture matches.
pnpm bandwise run --local bandwise/sets/error-triage.json state.jsonBandwise is an independent product built on TypeSafe's System One models. It is not TypeSafe's documentation. For the System One models themselves, see docs.typesafe.ai.