feat(telemetry): local JSONL sink with per-session files

package.json

@@ -189,6 +189,76 @@
 						"Logs everything from *headers* plus request and response bodies (may include sensitive data)"
 					],
 					"default": "basic"
+				},
+				"coder.telemetry.level": {
+					"markdownDescription": "Controls Coder extension telemetry collection. Used to diagnose extension issues.",
+					"type": "string",
+					"enum": [
+						"off",
+						"local"
+					],
+					"markdownEnumDescriptions": [
+						"Disable telemetry collection.",
+						"Record events on this machine only."
+					],
+					"default": "local",
+					"tags": [
+						"telemetry"
+					]
+				},
+				"coder.telemetry.local": {
+					"markdownDescription": "Tunables for the local telemetry sink, which writes events as JSON Lines under the extension's global storage. Used when `#coder.telemetry.level#` is `local`. Missing or invalid fields fall back to defaults.",
+					"type": "object",
+					"additionalProperties": false,
+					"properties": {
+						"flushIntervalMs": {
+							"type": "number",
+							"minimum": 1000,
+							"default": 15000,
+							"markdownDescription": "How often, in milliseconds, buffered events are written to disk."
+						},
+						"flushBatchSize": {
+							"type": "number",
+							"minimum": 1,
+							"default": 100,
+							"markdownDescription": "Number of buffered events that triggers an immediate write, ahead of the scheduled interval."
+						},
+						"bufferLimit": {
+							"type": "number",
+							"minimum": 10,
+							"default": 500,
+							"markdownDescription": "Maximum number of events held in memory. Oldest events are dropped on overflow. Should be at least `flushBatchSize`."
+						},
+						"maxFileBytes": {
+							"type": "number",
+							"minimum": 4096,
+							"default": 5242880,
+							"markdownDescription": "Maximum size, in bytes, of a single telemetry file before the sink rotates to a new one."
+						},
+						"maxAgeDays": {
+							"type": "number",
+							"minimum": 1,
+							"default": 30,
+							"markdownDescription": "Telemetry files older than this many days are deleted when the extension activates."
+						},
+						"maxTotalBytes": {
+							"type": "number",
+							"minimum": 4096,
+							"default": 104857600,
+							"markdownDescription": "Cap, in bytes, on the combined size of telemetry files. Oldest files are deleted on activation until the total is under the cap."
+						}
+					},
+					"default": {
+						"flushIntervalMs": 15000,
+						"flushBatchSize": 100,
+						"bufferLimit": 500,
+						"maxFileBytes": 5242880,
+						"maxAgeDays": 30,
+						"maxTotalBytes": 104857600
+					},
+					"tags": [
+						"telemetry"
+					]
 				}
 			}
 		},

src/core/container.ts

@@ -4,6 +4,7 @@ import { CoderApi } from "../api/coderApi";
 import { LoginCoordinator } from "../login/loginCoordinator";
 import { OAuthCallback } from "../oauth/oauthCallback";
 import { TelemetryService } from "../telemetry/service";
+import { LocalJsonlSink } from "../telemetry/sinks/localJsonlSink";
 import { SpeedtestPanelFactory } from "../webviews/speedtest/speedtestPanelFactory";
 import { DuplicateWorkspaceIpc } from "../workspace/duplicateWorkspaceIpc";
 
@@ -90,7 +91,18 @@ export class ServiceContainer implements vscode.Disposable {
 			context.extensionUri,
 			this.logger,
 		);
-		this.telemetryService = new TelemetryService(context, [], this.logger);
+		const localJsonlSink = LocalJsonlSink.start(
+			{
+				baseDir: this.pathResolver.getTelemetryPath(),
+				sessionId: vscode.env.sessionId,
+			},
+			this.logger,
+		);
+		this.telemetryService = new TelemetryService(
+			context,
+			[localJsonlSink],
+			this.logger,
+		);
 	}
 
 	getPathResolver(): PathResolver {

src/core/pathResolver.ts

@@ -49,6 +49,13 @@ export class PathResolver {
 		return path.join(this.basePath, "net");
 	}
 
+	/**
+	 * Return the directory where telemetry files are written.
+	 */
+	public getTelemetryPath(): string {
+		return path.join(this.basePath, "telemetry");
+	}
+
 	/**
 	 * Return the proxy log directory from the `coder.proxyLogDirectory` setting
 	 * or the `CODER_SSH_LOG_DIR` environment variable, falling back to the `log`

src/remote/sshProcess.ts

@@ -4,6 +4,7 @@ import * as path from "node:path";
 import * as vscode from "vscode";
 
 import { findPort } from "../util";
+import { cleanupFiles } from "../util/fileCleanup";
 
 import { NetworkStatusReporter } from "./networkStatus";
 
@@ -80,72 +81,6 @@ export class SshProcessMonitor implements vscode.Disposable {
 	private lastStaleSearchTime = 0;
 	private readonly reporter: NetworkStatusReporter;
 
-	/**
-	 * Helper to clean up files in a directory.
-	 * Stats files in parallel, applies selection criteria, then deletes in parallel.
-	 */
-	private static async cleanupFiles(
-		dir: string,
-		fileType: string,
-		logger: Logger,
-		options: {
-			filter: (name: string) => boolean;
-			select: (
-				files: Array<{ name: string; mtime: number }>,
-				now: number,
-			) => Array<{ name: string }>;
-		},
-	): Promise<void> {
-		try {
-			const now = Date.now();
-			const files = await fs.readdir(dir);
-
-			// Gather file stats in parallel
-			const withStats = await Promise.all(
-				files.filter(options.filter).map(async (name) => {
-					try {
-						const stats = await fs.stat(path.join(dir, name));
-						return { name, mtime: stats.mtime.getTime() };
-					} catch (error) {
-						if ((error as NodeJS.ErrnoException).code !== "ENOENT") {
-							logger.debug(`Failed to stat ${fileType} ${name}`, error);
-						}
-						return null;
-					}
-				}),
-			);
-
-			const toDelete = options.select(
-				withStats.filter((f) => f !== null),
-				now,
-			);
-
-			// Delete files in parallel
-			const results = await Promise.all(
-				toDelete.map(async (file) => {
-					try {
-						await fs.unlink(path.join(dir, file.name));
-						return file.name;
-					} catch (error) {
-						if ((error as NodeJS.ErrnoException).code !== "ENOENT") {
-							logger.debug(`Failed to delete ${fileType} ${file.name}`, error);
-						}
-						return null;
-					}
-				}),
-			);
-
-			const deletedFiles = results.filter((name) => name !== null);
-			if (deletedFiles.length > 0) {
-				logger.debug(
-					`Cleaned up ${deletedFiles.length} ${fileType}(s): ${deletedFiles.join(", ")}`,
-				);
-			}
-		} catch {
-			// Directory may not exist yet, ignore
-		}
-	}
-
 	/**
 	 * Cleans up network info files older than the specified age.
 	 */
@@ -154,15 +89,11 @@ export class SshProcessMonitor implements vscode.Disposable {
 		maxAgeMs: number,
 		logger: Logger,
 	): Promise<void> {
-		await SshProcessMonitor.cleanupFiles(
-			networkInfoPath,
-			"network info file",
-			logger,
-			{
-				filter: (name) => name.endsWith(".json"),
-				select: (files, now) => files.filter((f) => now - f.mtime > maxAgeMs),
-			},
-		);
+		await cleanupFiles(networkInfoPath, logger, {
+			label: "network info file",
+			filter: (name) => name.endsWith(".json"),
+			select: (files, now) => files.filter((f) => now - f.mtime > maxAgeMs),
+		});
 	}
 
 	/**
@@ -175,7 +106,8 @@ export class SshProcessMonitor implements vscode.Disposable {
 		maxAgeMs: number,
 		logger: Logger,
 	): Promise<void> {
-		await SshProcessMonitor.cleanupFiles(logDir, "log file", logger, {
+		await cleanupFiles(logDir, logger, {
+			label: "log file",
 			filter: (name) => name.startsWith("coder-ssh") && name.endsWith(".log"),
 			select: (files, now) =>
 				files

src/settings/telemetry.ts

@@ -0,0 +1,68 @@
+import type { WorkspaceConfiguration } from "vscode";
+
+import type { TelemetryLevel } from "../telemetry/event";
+
+export const TELEMETRY_LEVEL_SETTING = "coder.telemetry.level";
+export const LOCAL_SINK_SETTING = "coder.telemetry.local";
+
+/** Telemetry level. Falls back to `local` for unknown or invalid values. */
+export function readTelemetryLevel(
+	cfg: Pick<WorkspaceConfiguration, "get">,
+): TelemetryLevel {
+	const value = cfg.get<string>(TELEMETRY_LEVEL_SETTING);
+	return value === "off" || value === "local" ? value : "local";
+}
+
+export interface LocalSinkConfig {
+	readonly flushIntervalMs: number;
+	readonly flushBatchSize: number;
+	readonly bufferLimit: number;
+	readonly maxFileBytes: number;
+	readonly maxAgeDays: number;
+	readonly maxTotalBytes: number;
+}
+
+export const LOCAL_SINK_DEFAULTS: LocalSinkConfig = {
+	flushIntervalMs: 15_000,
+	flushBatchSize: 100,
+	bufferLimit: 500,
+	maxFileBytes: 5 * 1024 * 1024,
+	maxAgeDays: 30,
+	maxTotalBytes: 100 * 1024 * 1024,
+};
+
+// Defense in depth: VS Code does not enforce JSON schema at runtime, so users
+// can drop in any value via settings.json. Mirrors the minimums in package.json.
+const LOCAL_SINK_MINIMUMS: LocalSinkConfig = {
+	flushIntervalMs: 1000,
+	flushBatchSize: 1,
+	bufferLimit: 10,
+	maxFileBytes: 4096,
+	maxAgeDays: 1,
+	maxTotalBytes: 4096,
+};
+
+/** Per-field: missing, non-numeric, or below-minimum values fall back to defaults. */
+export function readLocalSinkConfig(
+	cfg: Pick<WorkspaceConfiguration, "get">,
+): LocalSinkConfig {
+	const raw = cfg.get(LOCAL_SINK_SETTING);
+	const obj =
+		raw && typeof raw === "object" && !Array.isArray(raw)
+			? (raw as Record<string, unknown>)
+			: {};
+	const read = (key: keyof LocalSinkConfig): number => {
+		const value = obj[key];
+		return typeof value === "number" && value >= LOCAL_SINK_MINIMUMS[key]
+			? value
+			: LOCAL_SINK_DEFAULTS[key];
+	};
+	return {
+		flushIntervalMs: read("flushIntervalMs"),
+		flushBatchSize: read("flushBatchSize"),
+		bufferLimit: read("bufferLimit"),
+		maxFileBytes: read("maxFileBytes"),
+		maxAgeDays: read("maxAgeDays"),
+		maxTotalBytes: read("maxTotalBytes"),
+	};
+}

src/telemetry/service.ts

@@ -2,6 +2,10 @@ import * as vscode from "vscode";
 
 import { watchConfigurationChanges } from "../configWatcher";
 import { type Logger } from "../logging/logger";
+import {
+	TELEMETRY_LEVEL_SETTING,
+	readTelemetryLevel,
+} from "../settings/telemetry";
 
 import {
 	buildSession,
@@ -15,13 +19,14 @@ import {
 } from "./event";
 import { NOOP_SPAN, type Span } from "./span";
 
-const TELEMETRY_LEVEL_SETTING = "coder.telemetry.level";
-
 const LEVEL_ORDER: Readonly<Record<TelemetryLevel, number>> = {
 	off: 0,
 	local: 1,
 };
 
+const readLevel = (): TelemetryLevel =>
+	readTelemetryLevel(vscode.workspace.getConfiguration());
+
 /** Trace context shared by all events in one trace. */
 interface SpanOptions {
 	traceId: string;
@@ -56,11 +61,13 @@ export class TelemetryService implements vscode.Disposable {
 		this.#configWatcher = watchConfigurationChanges(
 			[{ setting: TELEMETRY_LEVEL_SETTING, getValue: readLevel }],
 			(changes) => {
-				const raw = changes.get(TELEMETRY_LEVEL_SETTING);
-				if (!isTelemetryLevel(raw)) {
+				const next = changes.get(TELEMETRY_LEVEL_SETTING) as
+					| TelemetryLevel
+					| undefined;
+				if (!next) {
 					return;
 				}
-				this.#applyLevelChange(raw).catch((err) => {
+				this.#applyLevelChange(next).catch((err) => {
 					this.logger.warn("Telemetry level change failed", err);
 				});
 			},
@@ -277,14 +284,3 @@ export class TelemetryService implements vscode.Disposable {
 		}
 	}
 }
-
-function readLevel(): TelemetryLevel {
-	const value = vscode.workspace
-		.getConfiguration()
-		.get<string>(TELEMETRY_LEVEL_SETTING);
-	return isTelemetryLevel(value) ? value : "local";
-}
-
-function isTelemetryLevel(value: unknown): value is TelemetryLevel {
-	return value === "off" || value === "local";
-}