Add telemetry for Interactive Shell, Query Playground, and cross-feature navigation

[tnaum-ms]

Adds telemetry instrumentation to the new features launching in 0.8.0 to help identify areas for improvement.

What's covered

• Interactive Shell: session lifecycle (start/end), per-command evaluation with operation classification, error rates
• Query Playground: execution analytics with command classification, run mode (Run All vs Run Selected), schema scan usage
• Cross-feature navigation: activation source on every command — distinguishes context menu, inline buttons, CodeLens, shell action lines, and Collection View toolbar
• Completion acceptance: tracks which completion categories users select (fields, operators, BSON constructors, etc.) in both playground and Collection View
• Worker thread health: spawn count, lifetime, termination reason
• Server metadata correlation: connectionCorrelationId links shell/playground events to existing connect.* metadata for dashboard joins

Key design decisions

• shell.eval wraps the evaluate call in a single callWithTelemetryAndErrorHandling — the framework captures duration, result, and errors automatically
• Inline vs context menu activation uses separate command IDs (.inline variants) in package.json
• Collection View completion tracking routes through a tRPC mutation to avoid double telemetry — the middleware is the single emission point
• Command classification uses regex pattern matching (classifyCommand.ts) — no user data captured, only operation categories

New files

File	Purpose
src/utils/classifyCommand.ts	Classifies shell/playground commands into telemetry-safe categories

[Copilot]

Pull request overview

This PR adds telemetry instrumentation across the new 0.8.0 Interactive Shell, Query Playground, Collection View completions, and cross-feature navigation flows, enabling usage analytics and health signals without capturing user query content.

Changes:

• Instrument shell + playground execution with session lifecycle telemetry, command classification, and connection correlation.
• Add completion-acceptance telemetry in both VS Code completions (playground) and Monaco (Collection View) paths.
• Add activation-source telemetry via dedicated .inline command IDs and navigation-specific events.

Reviewed changes

Copilot reviewed 19 out of 19 changed files in this pull request and generated 11 comments.

Show a summary per file

File	Description
src/webviews/query-language-support/registerLanguage.ts	Adds Monaco command wiring to report completion acceptance back to the extension host.
src/webviews/query-language-support/completions/createCompletionItems.ts	Attaches completion-acceptance command to generated Monaco completion items + maps kind→category.
src/webviews/documentdb/collectionView/components/queryEditor/QueryEditor.tsx	Passes completion-accepted callback to language registration and routes via tRPC.
src/webviews/documentdb/collectionView/collectionViewRouter.ts	Adds telemetry properties for toolbar-driven navigation + new completionAccepted mutation.
src/utils/classifyCommand.ts	New utility to classify shell/playground commands into telemetry-safe categories.
src/documentdb/shell/ShellTerminalLinkProvider.ts	Adds activationSource telemetry for shell action-line navigation links.
src/documentdb/shell/DocumentDBShellPty.ts	Adds shell session start/end + per-eval telemetry (category, index, result type, correlation).
src/documentdb/query-language/playground-completions/PlaygroundCompletionItemProvider.ts	Adds completion-acceptance command on VS Code completion items (playground).
src/documentdb/playground/WorkerSessionManager.ts	Adds worker lifecycle telemetry (spawn count, lifetime, termination events).
src/documentdb/ClustersExtension.ts	Registers new inline command IDs and internal completionAccepted telemetry command.
src/commands/playground/scanCollectionSchema.ts	Records schema scan results in telemetry measurements.
src/commands/playground/playgroundOpenInShell.ts	Adds activationSource telemetry for CodeLens-driven navigation to shell.
src/commands/playground/playgroundOpenInCollectionView.ts	Adds activationSource + hasFilter telemetry for CodeLens-driven navigation to Collection View.
src/commands/playground/newPlayground.ts	Adds activationSource + context properties for playground creation and cross-feature starts.
src/commands/playground/executePlaygroundCode.ts	Adds code-block command classification + correlation to existing connection metadata.
src/commands/openInteractiveShell/openInteractiveShell.ts	Adds activationSource + viewId telemetry for shell open flows.
package.json	Adds .inline command contributions and updates inline menu wiring.
.github/skills/telemetry-instrumentation/SKILL.md	Expands telemetry instrumentation guidance (framework usage + common mistakes).
.github/copilot-instructions.md	Links the telemetry skill doc for repo guidance discoverability.

[Copilot]

Pull request overview

Copilot reviewed 26 out of 26 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (1)

src/documentdb/playground/WorkerSessionManager.ts:385

• terminateWorker() computes reason before setting _terminatingIntentionally = true, so reason will almost always be reported as 'forced' (including graceful shutdown). Consider passing an explicit reason into terminateWorker() (e.g., from shutdown() vs killWorker()), or set the reason after deciding which termination path is being used.

    private terminateWorker(): void {
        const wasAlive = !!this._worker;
        const reason = this._terminatingIntentionally ? 'intentional' : 'forced';

        if (this._worker) {
            this._terminatingIntentionally = true;
            void this._worker.terminate();
            this._worker = undefined;
        }

[tnaum-ms]

Telemetry Review — Deep Review Items Resolution

Addressing issues #8–#14 from the deep code review.

Fixed

#8 — scanCollectionSchema swallows errors (High)
Fixed in 0b6dcd8. The catch block now re-throws after showing a minimal user message ("Failed to scan schema for X") with a "Show Details" button that opens the output channel. The telemetry framework correctly captures the failure.

#9 — shell.sessionEnd fires without shell.sessionStart on connection failure (High)
Fixed in 79bba80. Added a _sessionStarted flag set to true only after successful initialization. shell.sessionEnd is now guarded behind this flag, preventing unpaired events on connection failure.

#10 — sessionEvalCount stored as string property instead of numeric measurement (Medium)
Fixed in f1bda7d. Changed all three occurrences from context.telemetry.properties.sessionEvalCount = String(...) to context.telemetry.measurements.sessionEvalCount = ... for direct numeric aggregation in dashboards.

#13 — Playground init/connect phase has no dedicated telemetry event (Medium)
Fixed in 3fbdc7a. Wrapped ensureWorker() in callWithTelemetryAndErrorHandling('playground.connect', ...). The framework now auto-captures duration, result, and error for the connection phase independently. Includes authMethod and needsSpawn. Kept initDurationMs on playground.execute for convenience.

No Action (by design)

#11 — playground.sessionEnd can fire for zero-eval sessions (Low)
Per operator guidance: intentionally kept as-is. Zero-eval session end events are useful for understanding session abandonment patterns (e.g., users opening a playground but never running anything).

#12 — Cross-feature navigation telemetry missing viewId (Low)
Per operator guidance: deferred to a future iteration when more cross-feature events exist and dashboards are being built. Low severity for now.

#14 — Completion acceptance not tracked for TS-provided method completions (Medium)
Per operator guidance: accepted as a known gap. Only Layer 2 completions (operators, fields, BSON, collection names) are tracked. Method completion popularity can be inferred indirectly from commandCategory in shell.eval / playground.execute events.