preview-offline-scope

作者: microsoft

當使用者想在推送給使用者之前,先估算離線設定檔的下載大小與同步頻率成本時使用。僅供讀取。封裝…

npx skills add https://github.com/microsoft/power-platform-skills --skill preview-offline-scope

Shared instructions: shared-instructions.md — read first.

Preview Offline Scope

Read-only diagnostic. Tells you "if you pushed this profile to users right now, here's what their devices would download and how often it would re-sync." No mutations.

Useful before:

  • /setup-offline-profile for a final sanity check
  • /assign-offline-profile (so users don't get surprised by data caps)
  • /edit-offline-profile to gauge impact of a column-list change

Workflow

  1. Verify project + locate profile → 2. Run verify (drift check) → 3. Per-table row counts → 4. Cache-size estimate → 5. Report

Step 1 — Verify project + locate profile

Same as /edit-offline-profile Step 1. Read profileId from offline-profile.json or $ARGUMENTS --profile-id. Do not read profile metadata from power.config.json; it is owned by npx power-apps init.

test -f power.config.json
node "${PLUGIN_ROOT}/scripts/resolve-environment.js" "$(node -e \"console.log(require('./power.config.json').environmentId)\")"

Step 2 — Run verify

Telemetry checkpoint: verify_offline_profile_snapshot

node "${PLUGIN_ROOT}/scripts/verify-offline-profile.js" <envUrl> \
  --project-root "$(pwd)"

If status: drift, surface the drift list verbatim. The estimate that follows is still meaningful but flag that the live profile diverges from offline-profile.json — recommend re-running /setup-offline-profile or /edit-offline-profile to reconcile.

Step 3 — Per-table row counts (with scope-applied filter)

Telemetry checkpoint: count_offline_profile_rows

For each table in the profile, run a count query that applies the same filter the runtime would use:

recorddistributioncriteriarecordsownedby* flagEffective filter
1 (All records)n/a?$count=true&$top=0 (whole table)
2 + merecordsownedbyme=true?$count=true&$top=0&$filter=_ownerid_value eq <current-user-id>
2 + teamrecordsownedbymyteam=trueApprox: _ownerid_value in (current user's team-owned IDs); for estimate, use team count from teamroles_association
2 + burecordsownedbymybusinessunit=true?$count=true&$top=0&$filter=_owningbusinessunit_value eq <current-bu-id>
0 (Related only)n/aCan't estimate independently — depends on parents. Report ~depends on parent counts.

For current-user/current-BU filters, resolve identity only inside this skill by calling Dataverse WhoAmI through scripts/dataverse-request.js against the resolved <envUrl>. If that call fails, report the affected scope as unknown — current user/BU unavailable instead of blocking the whole preview. Do not expect scripts/resolve-environment.js or auth.config.json.environment to provide UserId / BusinessUnitId.

For each table:

node "${PLUGIN_ROOT}/scripts/dataverse-request.js" <envUrl> GET \
  "<entitysetname>?\$count=true&\$top=0&<scope-filter>"

Cap at 5000 (Dataverse non-aggregate count cap). When the result is exactly 5000, prefix with ≥ in the report.

Step 4 — Cache-size estimate

Telemetry checkpoint: estimate_offline_cache_size

Rough byte-per-row heuristics (configurable; replace with measured values once we have a real sync benchmark):

Column typeBytes
Uniqueidentifier40
String (avg 80 chars)100
Integer / Decimal12
DateTime28
Boolean / State / Status4
Picklist (option value only)8
Memo (avg 500 chars)600
Lookup (FK only)40
Image (URL + thumbnail metadata only — see note below for full-image bytes)200
File (URL + name metadata)200

For each table, compute: estimatedBytesPerRow = sum(bytesForEach column in selectedcolumns). Then tableTotalBytes = rows × estimatedBytesPerRow.

Per-table sync overhead: each syncintervalinminutes interval triggers a delta query → assume ~20% of total rows touched on a typical day. dailyTransferBytes = (totalBytes × 0.20 × intervalsPerDay).

Step 5 — Report

═════════════════════════════════════════════════════════════
  Offline Scope Preview — <profileName>
═════════════════════════════════════════════════════════════

Drift status: ok | drift (see verify output)

Per-table breakdown:

| Table              | Scope          | Rows     | Cols | Est. bytes | Sync (min) |
|--------------------|----------------|----------|------|------------|------------|
| chnl_region        | All records    | 12       | 7    | 4 KB       | 60         |
| chnl_product       | All records    | 312      | 13   | 280 KB     | 60         |
| chnl_rmprofile     | Org+me         | 1        | 10   | 1 KB       | 10         |
| chnl_store         | Org+me         | 84       | 17   | 112 KB     | 10         |
| chnl_storevisit    | Org+me         | 412      | 13   | 220 KB     | 5          |
| chnl_order         | Org+me         | 156      | 14   | 95 KB      | 10         |
| chnl_orderline     | Related only   | ~depends | 11   | ~150 KB    | 10         |

Total initial download : ~860 KB (5000-rows cap not hit)
Daily transfer estimate: ~12 MB / device / day (assuming 20% delta rate)

Sync intervals per day:
  chnl_storevisit (5 min)  → 288 syncs/day  ⚠ high
  chnl_store      (10 min) → 144 syncs/day
  chnl_order      (10 min) → 144 syncs/day
  chnl_rmprofile  (10 min) → 144 syncs/day
  chnl_region     (60 min) → 24  syncs/day
  chnl_product    (60 min) → 24  syncs/day

Concerns:
  - chnl_storevisit at 5 min could exceed mobile data caps in low-signal areas
    (battery drain too). Consider 10 min unless 5 min is mission-critical.

Image / File bytes NOT counted by these heuristics. The metadata (URL +
thumbnail size) IS counted at 200 bytes/row, but the JPEG/file payload
itself transfers separately at runtime sync time. To estimate image
storage, multiply expected non-empty image rows × average JPEG size
(typically 10–500 KB per thumbnail).

Status code (final line)

  • DONE — estimate produced, no concerns
  • DONE_WITH_CONCERNS: <list> — estimate produced but flagged items (high sync rates, near-cap row counts, drift detected, etc.)
  • NEEDS_CONTEXT: <missing> — no profile to estimate against
  • BLOCKED: <reason> — auth or query failures

Notes

This skill is intentionally estimate-grade, not precise. The byte heuristics are rule-of-thumb; real cache size depends on Dataverse's storage layout + JSON serialization + per-device cache compression. The numbers are useful for comparative decisions (one scope vs another, one column list vs another), not for absolute capacity planning.

For precise figures, the only source of truth is the runtime sync log itself — only available once the app is deployed and an RM has signed in.

來自 microsoft 的更多技能

oss-growth
microsoft
開源增長駭客角色
agent-framework-azure-ai-py
microsoft
使用Microsoft Agent Framework Python SDK(agent-framework-azure-ai)构建Azure AI Foundry代理。适用于使用AzureAIAgentsProvider创建持久化代理、使用托管工具(代码解释器、文件搜索、网络搜索)、集成MCP服务器、管理对话线程或实现流式响应。涵盖函数工具、结构化输出和多工具代理。
development
airunway-aks-setup
microsoft
在AKS上設定AI Runway——從裸叢集到執行模型。涵蓋叢集驗證、控制器安裝、GPU評估、供應商設定及首次部署。時機:「設定AI Runway」、「上線AKS叢集」、「安裝AI Runway」、「airunway設定」、「部署模型至AKS」、「在AKS上進行GPU推論」、「在AKS上設定KAITO」、「在AKS上執行LLM」、「在AKS上使用vLLM」、「在AKS上設定模型服務」、「AI Runway控制器」。
devops
appinsights-instrumentation
microsoft
使用Azure Application Insights檢測Web應用程式的指南。提供遙測模式、SDK設定與組態參考。適用時機:如何檢測應用程式、App Insights SDK、遙測模式、什麼是App Insights、Application Insights指南、檢測範例、APM最佳實踐。
devops
applicationinsights-web-ts
microsoft
使用Application Insights JavaScript SDK(@microsoft/applicationinsights-web)為瀏覽器/Web應用程式進行檢測。適用於真實使用者監控(RUM)——頁面檢視、點擊、AJAX/fetch依賴、例外、自訂事件,以及與後端OpenTelemetry追蹤關聯的瀏覽器端GenAI代理追蹤。涵蓋SDK載入器指令碼與npm設定、框架擴充(React、React Native、Angular)、點擊分析、遙測初始化器,以及從瀏覽器發出的代理/工具/模型span的OTel GenAI語意慣例。
devops
azure-ai-anomalydetector-java
microsoft
使用適用於 Java 的 Azure AI 異常偵測器 SDK 建置異常偵測應用程式。在實作單變量/多變量異常偵測、時間序列分析或 AI 驅動監控時使用。
development
azure-ai-language-conversations-py
microsoft
使用 azure-ai-language-conversations Python SDK 實作對話語言理解(CLU)。當使用 ConversationAnalysisClient 分析對話意圖與實體、建置 NLP 功能,或將語言理解整合至應用程式時使用。
development
azure-ai-ml-py
microsoft
Azure Machine Learning SDK v2 for Python。用於機器學習工作區、作業、模型、資料集、計算資源與管線。 觸發詞:「azure-ai-ml」、「MLClient」、「workspace」、「model registry」、「training jobs」、「datasets」。
development