gsd-pi 2.34.0 → 2.35.0-dev.34ce717

package/packages/pi-coding-agent/src/core/retry-handler.ts

@@ -0,0 +1,359 @@
+/**
+ * RetryHandler - Automatic retry logic with exponential backoff and credential/provider fallback.
+ *
+ * Handles retryable errors (overloaded, rate limit, server errors) by:
+ * 1. Trying alternate credentials for the same provider
+ * 2. Falling back to other providers via FallbackResolver
+ * 3. Exponential backoff with configurable max retries
+ *
+ * Context overflow errors are NOT handled here (see compaction).
+ */
+
+import type { Agent } from "@gsd/pi-agent-core";
+import type { AssistantMessage, Model } from "@gsd/pi-ai";
+import { isContextOverflow } from "@gsd/pi-ai";
+import type { UsageLimitErrorType } from "./auth-storage.js";
+import type { FallbackResolver } from "./fallback-resolver.js";
+import type { ModelRegistry } from "./model-registry.js";
+import type { SettingsManager } from "./settings-manager.js";
+import { sleep } from "../utils/sleep.js";
+import type { AgentSessionEvent } from "./agent-session.js";
+
+/** Dependencies injected from AgentSession into RetryHandler */
+export interface RetryHandlerDeps {
+	readonly agent: Agent;
+	readonly settingsManager: SettingsManager;
+	readonly modelRegistry: ModelRegistry;
+	readonly fallbackResolver: FallbackResolver;
+	getModel: () => Model<any> | undefined;
+	getSessionId: () => string;
+	emit: (event: AgentSessionEvent) => void;
+	/** Called when the retry handler switches to a fallback model */
+	onModelChange: (model: Model<any>) => void;
+}
+
+export class RetryHandler {
+	private _retryAbortController: AbortController | undefined = undefined;
+	private _retryAttempt = 0;
+	private _retryPromise: Promise<void> | undefined = undefined;
+	private _retryResolve: (() => void) | undefined = undefined;
+
+	constructor(private readonly _deps: RetryHandlerDeps) {}
+
+	/** Current retry attempt (0 if not retrying) */
+	get retryAttempt(): number {
+		return this._retryAttempt;
+	}
+
+	/** Whether auto-retry is currently in progress */
+	get isRetrying(): boolean {
+		return this._retryPromise !== undefined;
+	}
+
+	/** Whether auto-retry is enabled */
+	get autoRetryEnabled(): boolean {
+		return this._deps.settingsManager.getRetryEnabled();
+	}
+
+	/** Toggle auto-retry setting */
+	setAutoRetryEnabled(enabled: boolean): void {
+		this._deps.settingsManager.setRetryEnabled(enabled);
+	}
+
+	/**
+	 * Create a retry promise synchronously for agent_end events.
+	 * Must be called synchronously from the agent event handler before
+	 * any async processing, so that waitForRetry() doesn't miss in-flight retries.
+	 */
+	createRetryPromiseForAgentEnd(messages: Array<{ role: string } & Record<string, any>>): void {
+		if (this._retryPromise) return;
+
+		const settings = this._deps.settingsManager.getRetrySettings();
+		if (!settings.enabled) return;
+
+		const lastAssistant = this._findLastAssistantInMessages(messages);
+		if (!lastAssistant || !this.isRetryableError(lastAssistant)) return;
+
+		this._retryPromise = new Promise((resolve) => {
+			this._retryResolve = resolve;
+		});
+	}
+
+	/**
+	 * Handle a successful assistant response by resetting retry state.
+	 * Call this when an assistant message completes without error.
+	 */
+	handleSuccessfulResponse(): void {
+		if (this._retryAttempt > 0) {
+			this._deps.emit({
+				type: "auto_retry_end",
+				success: true,
+				attempt: this._retryAttempt,
+			});
+			this._retryAttempt = 0;
+			this._resolveRetry();
+		}
+	}
+
+	/**
+	 * Check if an error is retryable (overloaded, rate limit, server errors).
+	 * Context overflow errors are NOT retryable (handled by compaction instead).
+	 */
+	isRetryableError(message: AssistantMessage): boolean {
+		if (message.stopReason !== "error" || !message.errorMessage) return false;
+
+		// Context overflow is handled by compaction, not retry
+		const contextWindow = this._deps.getModel()?.contextWindow ?? 0;
+		if (isContextOverflow(message, contextWindow)) return false;
+
+		const err = message.errorMessage;
+		return /overloaded|rate.?limit|too many requests|429|500|502|503|504|service.?unavailable|server.?error|internal.?error|connection.?error|connection.?refused|other side closed|fetch failed|upstream.?connect|reset before headers|terminated|retry delay|network.?(?:is\s+)?unavailable|credentials.*expired|temporarily backed off/i.test(
+			err,
+		);
+	}
+
+	/**
+	 * Handle retryable errors with exponential backoff.
+	 * When multiple credentials are available, marks the failing credential
+	 * as backed off and retries immediately with the next one.
+	 * @returns true if retry was initiated, false if max retries exceeded or disabled
+	 */
+	async handleRetryableError(message: AssistantMessage): Promise<boolean> {
+		const settings = this._deps.settingsManager.getRetrySettings();
+		if (!settings.enabled) {
+			this._resolveRetry();
+			return false;
+		}
+
+		// Retry promise is created synchronously in createRetryPromiseForAgentEnd.
+		// Keep a defensive fallback here in case a future refactor bypasses that path.
+		if (!this._retryPromise) {
+			this._retryPromise = new Promise((resolve) => {
+				this._retryResolve = resolve;
+			});
+		}
+
+		// Try credential fallback before counting against retry budget.
+		if (this._deps.getModel() && message.errorMessage) {
+			const errorType = this._classifyErrorType(message.errorMessage);
+			const isCredentialError = errorType !== "unknown";
+			const hasAlternate =
+				isCredentialError &&
+				this._deps.modelRegistry.authStorage.markUsageLimitReached(
+					this._deps.getModel()!.provider,
+					this._deps.getSessionId(),
+					{ errorType },
+				);
+
+			if (hasAlternate) {
+				this._removeLastAssistantError();
+
+				this._deps.emit({
+					type: "auto_retry_start",
+					attempt: this._retryAttempt + 1,
+					maxAttempts: settings.maxRetries,
+					delayMs: 0,
+					errorMessage: `${message.errorMessage} (switching credential)`,
+				});
+
+				// Retry immediately with the next credential - don't increment _retryAttempt
+				setTimeout(() => {
+					this._deps.agent.continue().catch(() => {});
+				}, 0);
+
+				return true;
+			}
+
+			// All credentials are backed off. Try cross-provider fallback before giving up.
+			if (isCredentialError) {
+				const fallbackResult = await this._deps.fallbackResolver.findFallback(
+					this._deps.getModel()!,
+					errorType,
+				);
+
+				if (fallbackResult) {
+					const previousProvider = this._deps.getModel()!.provider;
+					this._deps.agent.setModel(fallbackResult.model);
+					this._deps.onModelChange(fallbackResult.model);
+					this._removeLastAssistantError();
+
+					this._deps.emit({
+						type: "fallback_provider_switch",
+						from: `${previousProvider}/${this._deps.getModel()?.id}`,
+						to: `${fallbackResult.model.provider}/${fallbackResult.model.id}`,
+						reason: fallbackResult.reason,
+					});
+
+					this._deps.emit({
+						type: "auto_retry_start",
+						attempt: this._retryAttempt + 1,
+						maxAttempts: settings.maxRetries,
+						delayMs: 0,
+						errorMessage: `${message.errorMessage} (${fallbackResult.reason})`,
+					});
+
+					// Retry immediately with fallback provider - don't increment _retryAttempt
+					setTimeout(() => {
+						this._deps.agent.continue().catch(() => {});
+					}, 0);
+
+					return true;
+				}
+
+				// No fallback available either
+				if (errorType === "quota_exhausted") {
+					this._deps.emit({
+						type: "fallback_chain_exhausted",
+						reason: `All providers exhausted for ${this._deps.getModel()!.provider}/${this._deps.getModel()!.id}`,
+					});
+					this._deps.emit({
+						type: "auto_retry_end",
+						success: false,
+						attempt: this._retryAttempt,
+						finalError: message.errorMessage,
+					});
+					this._retryAttempt = 0;
+					this._resolveRetry();
+					return false;
+				}
+			}
+		}
+
+		this._retryAttempt++;
+
+		if (this._retryAttempt > settings.maxRetries) {
+			this._deps.emit({
+				type: "auto_retry_end",
+				success: false,
+				attempt: this._retryAttempt - 1,
+				finalError: message.errorMessage,
+			});
+			this._retryAttempt = 0;
+			this._resolveRetry();
+			return false;
+		}
+
+		// Use server-requested delay when available, capped by maxDelayMs.
+		// Fall back to exponential backoff when no server hint is present.
+		const exponentialDelayMs = settings.baseDelayMs * 2 ** (this._retryAttempt - 1);
+		let delayMs: number;
+		if (message.retryAfterMs !== undefined) {
+			const cap = settings.maxDelayMs > 0 ? settings.maxDelayMs : Infinity;
+			if (message.retryAfterMs > cap) {
+				this._deps.emit({
+					type: "auto_retry_end",
+					success: false,
+					attempt: this._retryAttempt - 1,
+					finalError: `Rate limit reset in ${Math.ceil(message.retryAfterMs / 1000)}s (max: ${Math.ceil(cap / 1000)}s). ${message.errorMessage || ""}`.trim(),
+				});
+				this._retryAttempt = 0;
+				this._resolveRetry();
+				return false;
+			}
+			delayMs = message.retryAfterMs;
+		} else {
+			delayMs = exponentialDelayMs;
+		}
+
+		this._deps.emit({
+			type: "auto_retry_start",
+			attempt: this._retryAttempt,
+			maxAttempts: settings.maxRetries,
+			delayMs,
+			errorMessage: message.errorMessage || "Unknown error",
+		});
+
+		this._removeLastAssistantError();
+
+		// Wait with exponential backoff (abortable)
+		this._retryAbortController = new AbortController();
+		try {
+			await sleep(delayMs, this._retryAbortController.signal);
+		} catch {
+			// Aborted during sleep
+			const attempt = this._retryAttempt;
+			this._retryAttempt = 0;
+			this._retryAbortController = undefined;
+			this._deps.emit({
+				type: "auto_retry_end",
+				success: false,
+				attempt,
+				finalError: "Retry cancelled",
+			});
+			this._resolveRetry();
+			return false;
+		}
+		this._retryAbortController = undefined;
+
+		// Retry via continue() - use setTimeout to break out of event handler chain
+		setTimeout(() => {
+			this._deps.agent.continue().catch(() => {});
+		}, 0);
+
+		return true;
+	}
+
+	/** Cancel in-progress retry */
+	abortRetry(): void {
+		this._retryAbortController?.abort();
+		this._resolveRetry();
+	}
+
+	/**
+	 * Wait for any in-progress retry to complete.
+	 * Returns immediately if no retry is in progress.
+	 */
+	async waitForRetry(): Promise<void> {
+		if (this._retryPromise) {
+			await this._retryPromise;
+		}
+	}
+
+	/** Resolve the pending retry promise */
+	resolveRetry(): void {
+		this._resolveRetry();
+	}
+
+	// =========================================================================
+	// Private helpers
+	// =========================================================================
+
+	private _resolveRetry(): void {
+		if (this._retryResolve) {
+			this._retryResolve();
+			this._retryResolve = undefined;
+			this._retryPromise = undefined;
+		}
+	}
+
+	private _findLastAssistantInMessages(
+		messages: Array<{ role: string } & Record<string, any>>,
+	): AssistantMessage | undefined {
+		for (let i = messages.length - 1; i >= 0; i--) {
+			const message = messages[i];
+			if (message.role === "assistant") {
+				return message as AssistantMessage;
+			}
+		}
+		return undefined;
+	}
+
+	/**
+	 * Classify an error message into a usage-limit error type for credential backoff.
+	 */
+	private _classifyErrorType(errorMessage: string): UsageLimitErrorType {
+		const err = errorMessage.toLowerCase();
+		if (/quota|billing|exceeded.*limit|usage.*limit/i.test(err)) return "quota_exhausted";
+		if (/rate.?limit|too many requests|429/i.test(err)) return "rate_limit";
+		if (/500|502|503|504|server.?error|internal.?error|service.?unavailable/i.test(err)) return "server_error";
+		return "unknown";
+	}
+
+	/** Remove the last assistant error message from agent state */
+	private _removeLastAssistantError(): void {
+		const messages = this._deps.agent.state.messages;
+		if (messages.length > 0 && messages[messages.length - 1].role === "assistant") {
+			this._deps.agent.replaceMessages(messages.slice(0, -1));
+		}
+	}
+}

package/packages/pi-coding-agent/src/core/session-manager.ts

@@ -16,8 +16,8 @@ import {
 import { atomicWriteFileSync } from "./fs-utils.js";
 import { readdir, readFile, stat } from "fs/promises";
 import { join, resolve } from "path";
-import lockfile from "proper-lockfile";
 import { getAgentDir as getDefaultAgentDir, getBlobsDir, getSessionsDir } from "../config.js";
+import { tryAcquireLockSync } from "./lock-utils.js";
 import {
 	type BashExecutionMessage,
 	type CustomMessage,
@@ -577,8 +577,7 @@ function resolveBlobRefsInEntries(entries: FileEntry[], blobStore: BlobStore): v
 	}
 }
 
-/** Exported for testing */
-export function loadEntriesFromFile(filePath: string): FileEntry[] {
+function loadEntriesFromFile(filePath: string): FileEntry[] {
 	if (!existsSync(filePath)) return [];
 
 	const content = readFileSync(filePath, "utf8");
@@ -620,8 +619,7 @@ function isValidSessionFile(filePath: string): boolean {
 	}
 }
 
-/** Exported for testing */
-export function findMostRecentSession(sessionDir: string): string | null {
+function findMostRecentSession(sessionDir: string): string | null {
 	try {
 		const files = readdirSync(sessionDir)
 			.filter((f) => f.endsWith(".jsonl"))
@@ -955,39 +953,12 @@ export class SessionManager {
 		}
 	}
 
-	private acquireSessionLock(path: string): (() => void) | undefined {
-		const maxAttempts = 10;
-		const delayMs = 20;
-		let lastError: unknown;
-
-		for (let attempt = 1; attempt <= maxAttempts; attempt++) {
-			try {
-				return lockfile.lockSync(path, { realpath: false });
-			} catch (error) {
-				const code =
-					typeof error === "object" && error !== null && "code" in error
-						? String((error as { code?: unknown }).code)
-						: undefined;
-				if (code !== "ELOCKED" || attempt === maxAttempts) {
-					// Non-fatal: proceed without lock rather than losing data
-					return undefined;
-				}
-				lastError = error;
-				const start = Date.now();
-				while (Date.now() - start < delayMs) {
-					// Busy-wait to avoid async
-				}
-			}
-		}
-		return undefined;
-	}
-
 	private _rewriteFile(): void {
 		if (!this.persist || !this.sessionFile) return;
 		const content = `${this.fileEntries.map((e) => JSON.stringify(e)).join("\n")}\n`;
 		let release: (() => void) | undefined;
 		try {
-			release = this.acquireSessionLock(this.sessionFile);
+			release = tryAcquireLockSync(this.sessionFile);
 			atomicWriteFileSync(this.sessionFile, content);
 		} finally {
 			release?.();
@@ -1026,7 +997,7 @@ export class SessionManager {
 
 		let release: (() => void) | undefined;
 		try {
-			release = this.acquireSessionLock(this.sessionFile);
+			release = tryAcquireLockSync(this.sessionFile);
 			if (!this.flushed) {
 				for (const e of this.fileEntries) {
 					const prepared = prepareForPersistence(e, this.blobStore) as FileEntry;