@xiaoso/dsh-tool-plus 0.1.6 → 0.1.7-beta.0

package/lib/{error-C-f_Wnw4.mjs → error-CORgGg5a.mjs}

@@ -1,14 +1,17 @@
 import { installBunShim as __installBunShim } from "./bun-shim.mjs";
 __installBunShim();
-import { J as getLogsDir, R as isBunTestRuntime } from "./src-BcD0_rQR.mjs";
+import { J as getLogsDir, L as isBunTestRuntime } from "./src-BwW51vA5.mjs";
 import * as path from "node:path";
 import { scheduler } from "node:timers/promises";
-//#region node_modules/.pnpm/@oh-my-pi+pi-utils@17.3.5/node_modules/@oh-my-pi/pi-utils/src/fetch-retry.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-utils@18.1.17/node_modules/@oh-my-pi/pi-utils/src/fetch-retry.ts
 const QUOTA_RESET_PATTERN = /reset after (?:(\d+)h)?(?:(\d+)m)?(\d+(?:\.\d+)?)s/i;
 const PLEASE_RETRY_PATTERN = /Please retry in ([0-9.]+)(ms|s)/i;
 const RETRY_DELAY_FIELD_PATTERN = /"retryDelay":\s*"([0-9.]+)(ms|s)"/i;
 const TRY_AGAIN_PATTERN = /try again in\s+~?\s*([0-9.]+)\s*(ms|sec|s|minutes?|mins?|m|hours?|hrs?|h)\b/i;
 const WILL_RESET_IN_PATTERN = /(?:will\s+)?reset in\s+~?\s*([0-9.]+)\s*(ms|sec|s|minutes?|mins?|m|hours?|hrs?|h)\b/i;
+const WILL_RESET_AT_PATTERN = /(?:will\s+)?reset at\s+([0-9]{4}-[0-9]{2}-[0-9]{2}[ T][0-9]{2}:[0-9]{2}:[0-9]{2}(?:\.[0-9]+)?(?:Z|[+-][0-9]{2}:?[0-9]{2})?)/i;
+const CN_RESET_AT_PATTERN = /将在\s*([0-9]{4}-[0-9]{2}-[0-9]{2}\s+[0-9]{2}:[0-9]{2}:[0-9]{2})\s*重置/;
+const RETRY_AFTER_MS_BODY_PATTERN = /\bretry-after-ms\s*[:=]\s*([0-9]+)\b/i;
 /**
 * Server-suggested retry delay extraction. Merges the patterns historically used
 * by the OpenAI Codex and Google Gemini retry helpers.
@@ -25,8 +28,12 @@ const WILL_RESET_IN_PATTERN = /(?:will\s+)?reset in\s+~?\s*([0-9.]+)\s*(ms|sec|s
 *  - `Please retry in 250ms` / `Please retry in 12s`
 *  - `"retryDelay": "34.074824224s"` (JSON error detail field)
 *  - `try again in 250ms` / `try again in 12s` / `try again in 5 min` / `try again in ~158 min`
+*  - `retry-after-ms=98497000` / `retry-after-ms: 7200000` / `retry-after-ms = 7200000`
+*  - `Your limit will reset at 2026-09-01 09:44:51` / `将在 2026-09-01 09:44:51 重置`
 *
-* Returns `undefined` if no signal is found.
+* Returns `undefined` if no signal is found, or `0` when the provider
+* explicitly asks for an immediate retry (`retry-after…=0`, or an absolute
+* reset timestamp that has already elapsed).
 */
 function extractRetryHint(source, body) {
 	const headers = source instanceof Headers ? source : source?.headers ?? void 0;
@@ -68,6 +75,16 @@ function extractRetryHint(source, body) {
 		}
 	}
 	if (!body) return void 0;
+	let longestMs;
+	let retryNow = false;
+	const consider = (ms) => {
+		if (ms !== void 0 && ms > 0 && (longestMs === void 0 || ms > longestMs)) longestMs = ms;
+	};
+	const considerClamped = (ms) => {
+		if (ms === void 0) return;
+		if (ms > 0) consider(ms);
+		else retryNow = true;
+	};
 	const quotaMatch = QUOTA_RESET_PATTERN.exec(body);
 	if (quotaMatch) {
 		const hours = quotaMatch[1] ? Number.parseInt(quotaMatch[1], 10) : 0;
@@ -75,11 +92,32 @@ function extractRetryHint(source, body) {
 		const seconds = Number.parseFloat(quotaMatch[3]);
 		if (!Number.isNaN(seconds)) {
 			const totalMs = ((hours * 60 + minutes) * 60 + seconds) * 1e3;
-			if (totalMs > 0) return totalMs;
+			consider(totalMs > 0 ? totalMs : void 0);
+		}
+	}
+	for (const pattern of [WILL_RESET_AT_PATTERN, CN_RESET_AT_PATTERN]) {
+		const match = pattern.exec(body);
+		if (match?.[1]) {
+			const normalized = match[1].replace(" ", "T");
+			const hasOffset = /(?:Z|[+-][0-9]{2}:?[0-9]{2})$/i.test(normalized);
+			const parsed = Date.parse(hasOffset ? normalized : `${normalized}Z`);
+			if (!Number.isNaN(parsed) && parsed > Date.now()) consider(parsed - Date.now());
 		}
 	}
+	const accountResetMatch = WILL_RESET_IN_PATTERN.exec(body);
+	if (accountResetMatch?.[1]) {
+		const value = Number.parseFloat(accountResetMatch[1]);
+		if (Number.isFinite(value) && value > 0) {
+			const unitMs = unitToMs(accountResetMatch[2]);
+			if (unitMs !== void 0) consider(value * unitMs);
+		}
+	}
+	const retryAfterMsMatch = RETRY_AFTER_MS_BODY_PATTERN.exec(body);
+	if (retryAfterMsMatch?.[1]) {
+		const ms = Number(retryAfterMsMatch[1]);
+		if (Number.isFinite(ms)) considerClamped(ms);
+	}
 	for (const pattern of [
-		WILL_RESET_IN_PATTERN,
 		PLEASE_RETRY_PATTERN,
 		RETRY_DELAY_FIELD_PATTERN,
 		TRY_AGAIN_PATTERN
@@ -89,10 +127,31 @@ function extractRetryHint(source, body) {
 			const value = Number.parseFloat(match[1]);
 			if (Number.isFinite(value) && value > 0) {
 				const unitMs = unitToMs(match[2]);
-				if (unitMs !== void 0) return value * unitMs;
+				if (unitMs !== void 0) consider(value * unitMs);
 			}
 		}
 	}
+	const retryAfterMatch = /retry-after\s*[:=]\s*([^\s,;]+)/i.exec(body);
+	if (retryAfterMatch) {
+		const value = retryAfterMatch[1];
+		const seconds = Number(value);
+		if (Number.isFinite(seconds)) considerClamped(seconds * 1e3);
+		else {
+			const dateMs = Date.parse(value);
+			if (!Number.isNaN(dateMs)) considerClamped(dateMs - Date.now());
+		}
+	}
+	const resetMsMatch = /x-ratelimit-reset-ms\s*[:=]\s*(\d+)/i.exec(body);
+	if (resetMsMatch) {
+		const resetMs = Number(resetMsMatch[1]);
+		if (!Number.isNaN(resetMs)) considerClamped(resetMs > 0xe8d4a51000 ? resetMs - Date.now() : resetMs);
+	}
+	const resetMatch = /x-ratelimit-reset\s*[:=]\s*(\d+)/i.exec(body);
+	if (resetMatch) {
+		const resetSeconds = Number(resetMatch[1]);
+		if (!Number.isNaN(resetSeconds)) considerClamped(resetSeconds > 1e9 ? resetSeconds * 1e3 - Date.now() : resetSeconds * 1e3);
+	}
+	return longestMs ?? (retryNow ? 0 : void 0);
 }
 function unitToMs(unit) {
 	switch (unit.toLowerCase()) {
@@ -242,7 +301,7 @@ function isRetryableStatus(status) {
 * proxies surface these for any HTTP/2 stream reset.
 */
 function isUnexpectedSocketCloseMessage(message) {
-	return /\b(?:the\s+)?socket connection (?:was )?closed unexpectedly\b/i.test(message);
+	return /\b(?:the\s+)?socket connection (?:was )?closed unexpectedly\b/i.test(message) || /^(?:error:\s*)?socket is closed\.?$/i.test(message.trim());
 }
 const TRANSIENT_MESSAGE_PATTERN = /overloaded|rate.?limit|too many requests|service.?unavailable|server error|internal error|connection.?error|unable to connect|fetch failed|network error|stream stall|other side closed|HTTP2(?:StreamReset|RefusedStream|EnhanceYourCalm)/i;
 const VALIDATION_MESSAGE_PATTERN = /invalid|validation|bad request|unsupported|schema|missing required|not found|unauthorized|forbidden/i;
@@ -265,7 +324,7 @@ function isRetryableError(error) {
 	return isUnexpectedSocketCloseMessage(message) || TRANSIENT_MESSAGE_PATTERN.test(message);
 }
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/error/aws.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/error/aws.ts
 /** A failure resolving AWS credentials for the Bedrock provider. */
 var AwsCredentialsError = class extends Error {
 	kind;
@@ -283,7 +342,7 @@ var EventStreamFrameError = class extends Error {
 	}
 };
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/error/classes.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/error/classes.ts
 /** Prefix on errors raised when an Anthropic SSE stream envelope is malformed. */
 const STREAM_ENVELOPE_ERROR_PREFIX = "Anthropic stream envelope error:";
 /** Non-2xx HTTP response from a provider. */
@@ -378,8 +437,7 @@ var AnthropicApiError = class AnthropicApiError extends ProviderHttpError {
 		if (signal?.aborted) onAbort();
 		else signal?.addEventListener("abort", onAbort, { once: true });
 		const deadline = performance.now() + anthropicErrorBodyReadTimeoutMs;
-		let timeout;
-		timeout = setTimeout(() => {
+		const timeout = setTimeout(() => {
 			timedOut = true;
 			cancelReader();
 		}, anthropicErrorBodyReadTimeoutMs);
@@ -474,9 +532,10 @@ var OllamaApiError = class extends ProviderHttpError {
 	name = "OllamaApiError";
 };
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/error/rate-limit.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/error/rate-limit.ts
 const ACCOUNT_RATE_LIMIT_PATTERN = /\baccount(?:'s)?\b[^\n]{0,80}\brate.?limit\b|\brate.?limit\b[^\n]{0,80}\baccount\b/i;
 const INSUFFICIENT_BALANCE_PATTERN = /insufficient.?balance/i;
+const CREDITS_EXHAUSTED_PATTERN = /\b(?:exceed\w*|insufficient|not enough)\b[^\n]{0,40}\bcredits?\b|\bcredits?\b[^\n]{0,40}\b(?:exhausted|depleted)\b/i;
 const SPEND_LIMIT_PATTERN = /spend.?limit/i;
 const SUBSCRIPTION_CAP_PATTERN = /\b(?:subscription|plan|membership)\b[^\n]{0,80}\b(?:rate.?limits?|quota|cap)\b|\b(?:rate.?limits?|quota|cap)\b[^\n]{0,80}\b(?:subscription|plan|membership)\b/i;
 const TRANSIENT_INTERVAL_RATE_LIMIT_PATTERN = /\bper\s+(?:second|minute)\b/i;
@@ -484,6 +543,7 @@ function matchesSubscriptionCapText(errorMessage) {
 	return SUBSCRIPTION_CAP_PATTERN.test(errorMessage) && !TRANSIENT_INTERVAL_RATE_LIMIT_PATTERN.test(errorMessage);
 }
 const OPENROUTER_DAILY_FREE_LIMIT_PATTERN = /\bfree[-_ ]models[-_ ]per[-_ ]day\b/i;
+const CLINE_PASS_QUOTA_PATTERN = /clinepass limit|free limit reached on model/i;
 const RESOURCE_EXHAUSTED_PATTERN = /resource.?exhausted/gi;
 const CONCURRENT_LIMIT_PATTERN = /\btoo many\s+concurren\w*\s+(?:requests?|invocations?)\b|\bconcurren\w*\b[^\n]{0,60}\b(?:limit|quota|exceed\w*|reach\w*)\b|\b(?:limit|quota|exceed\w*|reach\w*)\b[^\n]{0,60}\bconcurren\w*\b|\bconcurren[a-z]*[-_](?:[a-z]+[_-])*(?:limit|quota|exceed\w*|reach\w*)/i;
 const ACCOUNT_SCOPED_403_PATTERN = /\b(?:overall|account|organization|team|workspace)\b[^\n]{0,40}\b(?:message |request )?rate.?limit\b|\byour\b[^\n]{0,30}\b(?:limit )?will reset\b/i;
@@ -497,6 +557,7 @@ function isDashScopeTokenLimitText(errorMessage) {
 	return DASHSCOPE_TOKEN_LIMIT_DOC_PATTERN.test(errorMessage) && DASHSCOPE_TOKEN_LIMIT_MESSAGE_PATTERN.test(errorMessage);
 }
 const GOOGLE_RPC_ERROR_INFO_TYPE = "type.googleapis.com/google.rpc.ErrorInfo";
+const ANTIGRAVITY_MODEL_QUOTA_PATTERN = /\bexhausted your capacity on this model\b/i;
 const LONG_RATE_LIMIT_DELAY_MS = 3e5;
 function asRecord(value) {
 	return typeof value === "object" && value !== null && !Array.isArray(value) ? value : void 0;
@@ -527,6 +588,7 @@ function parseGoogleRpcRateLimitReason(errorMessage) {
 			case "QUOTA_EXHAUSTED": return "QUOTA_EXHAUSTED";
 			case "INSUFFICIENT_G1_CREDITS_BALANCE": return "INSUFFICIENT_G1_CREDITS_BALANCE";
 			case "RATE_LIMIT_EXCEEDED": {
+				if (typeof error.message === "string" && ANTIGRAVITY_MODEL_QUOTA_PATTERN.test(error.message)) return "QUOTA_EXHAUSTED";
 				const retryDelayMs = extractRetryHint(void 0, errorMessage);
 				return retryDelayMs !== void 0 && retryDelayMs >= LONG_RATE_LIMIT_DELAY_MS ? "QUOTA_EXHAUSTED" : "RATE_LIMIT_EXCEEDED";
 			}
@@ -561,26 +623,35 @@ function parseRateLimitReason(errorMessage) {
 	if (SPEND_LIMIT_PATTERN.test(errorMessage)) return "QUOTA_EXHAUSTED";
 	if (matchesSubscriptionCapText(errorMessage)) return "QUOTA_EXHAUSTED";
 	if (OPENROUTER_DAILY_FREE_LIMIT_PATTERN.test(errorMessage)) return "QUOTA_EXHAUSTED";
+	if (CLINE_PASS_QUOTA_PATTERN.test(errorMessage)) return "QUOTA_EXHAUSTED";
 	if (lower.includes("per minute") || lower.includes("rate limit") || lower.includes("too many requests") || lower.includes("presque")) return "RATE_LIMIT_EXCEEDED";
-	if (lower.includes("exhausted") || lower.includes("quota") || lower.includes("usage limit") || lower.includes("run out of credits") || lower.includes("out of credits") || lower.includes("spending-limit") || lower.includes("spending limit") || INSUFFICIENT_BALANCE_PATTERN.test(errorMessage)) return "QUOTA_EXHAUSTED";
+	if (lower.includes("exhausted") || lower.includes("quota") || lower.includes("usage limit") || lower.includes("run out of credits") || lower.includes("out of credits") || lower.includes("spending-limit") || lower.includes("spending limit") || INSUFFICIENT_BALANCE_PATTERN.test(errorMessage) || CREDITS_EXHAUSTED_PATTERN.test(errorMessage)) return "QUOTA_EXHAUSTED";
 	if (lower.includes("500") || lower.includes("internal error") || lower.includes("internal server error")) return "SERVER_ERROR";
 	if (hasResourceExhaustedStatus) return "MODEL_CAPACITY_EXHAUSTED";
 	return "UNKNOWN";
 }
 /** Detect usage/quota limit errors in error messages (persistent, requires credential switch). */
-const USAGE_LIMIT_PATTERN = /usage.?limit|usage_limit_reached|usage_not_included|limit_reached|quota.?(?:exceeded|reached|insufficient)|额度不足|额度耗尽|resource.?exhausted|exhausted your capacity|quota will reset|insufficient.?(?:balance|quota)|balance.?exhausted|run out of credits|out of credits|spending[- _]?limit|personal-team-blocked/i;
+const USAGE_LIMIT_PATTERN = /usage.?limit|usage_limit_reached|usage_not_included|limit_reached|quota.?(?:exceeded|reached|insufficient)|额度不足|额度耗尽|resource.?exhausted|exhausted your capacity|quota will reset|insufficient.?(?:balance|quota)|balance.?exhausted|run out of credits|out of credits|spending[- _]?limit|personal-team-blocked|clinepass limit|free limit reached on model/i;
 /**
 * HTTP status codes that, absent richer body classification, represent an
 * account-local usage cap rather than a bad credential or a transient blip.
-* HTTP 402 Payment Required is categorically an account-billing cap (xAI
+* HTTP 402 Payment Required represents an account-billing cap (xAI
 * Grok Build "usage balance exhausted", DeepSeek "Insufficient Balance",
-* OpenRouter credit exhaustion) — never a transient blip or bad credential.
-* Always combine with {@link isUsageLimitOutcome} when a message is available
-* — a 429 carrying transient rate-limit wording is NOT a usage cap.
+* OpenRouter credit exhaustion) when opaque, payment/deactivation/balance-worded,
+* or QUOTA_EXHAUSTED/CONCURRENT_LIMIT, while informative non-quota 402s (e.g.
+* endpoint subscription requirements) remain non-usage-limits. Always combine
+* with {@link isUsageLimitOutcome} when a message is available.
 */
 function isUsageLimitStatus(status) {
 	return status === 429 || status === 402;
 }
+const STATUS_402_QUOTA_PATTERN = /\b(?:payment(?:\s+is)?[-_.\s]*required|deactivated_workspace|insufficient.?balance)\b/i;
+function is402BillingCapBody(message) {
+	if (message === void 0 || isOpaqueStatusBody(message)) return true;
+	if (STATUS_402_QUOTA_PATTERN.test(message)) return true;
+	const reason = parseRateLimitReason(message);
+	return isQuotaExhaustedReason(reason) || reason === "CONCURRENT_LIMIT";
+}
 /**
 * A usage-limit status body is opaque when it carries no signal beyond the
 * status itself — empty, whitespace-only, the status digits with HTTP/JSON
@@ -589,7 +660,7 @@ function isUsageLimitStatus(status) {
 * classifier.
 */
 function isOpaqueStatusBody(message) {
-	const cleaned = message.replace(/\b(?:429|402)\b/g, "").replace(/\b(?:http|https|status|error|code|response|message)\b/gi, "");
+	const cleaned = message.replace(/\b(?:429|402)\b/g, "").replace(/\b(?:http|https|status|error|code|response|message)\b/gi, "").replace(/\(?\bno body\b\)?/gi, "");
 	return !/[a-z\d]{3,}/i.test(cleaned) && !CN_QUOTA_EXHAUSTED_PATTERN.test(cleaned) && !CN_TRANSIENT_CAP_PATTERN.test(cleaned) && !CN_THROTTLE_PATTERN.test(cleaned);
 }
 /**
@@ -602,7 +673,7 @@ function matchesUsageLimitText(errorMessage) {
 	const structuredReason = parseGoogleRpcRateLimitReason(errorMessage);
 	if (structuredReason !== void 0) return isQuotaExhaustedReason(structuredReason);
 	if (isDashScopeTokenLimitText(errorMessage)) return false;
-	return USAGE_LIMIT_PATTERN.test(errorMessage) || CN_QUOTA_EXHAUSTED_PATTERN.test(errorMessage) && !CN_TRANSIENT_CAP_PATTERN.test(errorMessage) || SPEND_LIMIT_PATTERN.test(errorMessage) || ACCOUNT_RATE_LIMIT_PATTERN.test(errorMessage) || matchesSubscriptionCapText(errorMessage) || OPENROUTER_DAILY_FREE_LIMIT_PATTERN.test(errorMessage);
+	return USAGE_LIMIT_PATTERN.test(errorMessage) || CREDITS_EXHAUSTED_PATTERN.test(errorMessage) || CN_QUOTA_EXHAUSTED_PATTERN.test(errorMessage) && !CN_TRANSIENT_CAP_PATTERN.test(errorMessage) || SPEND_LIMIT_PATTERN.test(errorMessage) || ACCOUNT_RATE_LIMIT_PATTERN.test(errorMessage) || matchesSubscriptionCapText(errorMessage) || OPENROUTER_DAILY_FREE_LIMIT_PATTERN.test(errorMessage);
 }
 /**
 * Account-scoped cap phrasing delivered on a 403 (or a statusless Connect
@@ -615,7 +686,7 @@ function isAccountScopedCapText(message) {
 	return ACCOUNT_SCOPED_403_PATTERN.test(message);
 }
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/error/flags.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/error/flags.ts
 const Flag = {
 	Class: 4096,
 	ThinkingLoop: 65536,
@@ -639,11 +710,13 @@ const Flag = {
 	/** Anthropic model/account does not support fast mode / the `speed` parameter. */
 	FastModeUnsupported: 536870912,
 	/** OAuth refresh failed definitively — the stored grant is dead, re-login required. */
-	OAuthExpiry: 1073741824
+	OAuthExpiry: 1073741824,
+	/** HTTP 413 byte/media rejection — token compaction cannot shrink bytes or media budgets (#9235). */
+	PayloadRejected: 2147483648
 };
-const KIND_MASK = Flag.ThinkingLoop | Flag.Transient | Flag.Timeout | Flag.UsageLimit | Flag.StaleResponsesItem | Flag.MalformedFunctionCall | Flag.ProviderFinishError | Flag.EmptyResponse | Flag.ContentBlocked | Flag.AccountPolicy | Flag.ContextOverflow | Flag.AuthFailed | Flag.SilentAbort | Flag.UserInterrupt | Flag.Abort | Flag.Grammar | Flag.FastModeUnsupported | Flag.OAuthExpiry;
+const KIND_MASK = Flag.ThinkingLoop | Flag.Transient | Flag.Timeout | Flag.UsageLimit | Flag.StaleResponsesItem | Flag.MalformedFunctionCall | Flag.ProviderFinishError | Flag.EmptyResponse | Flag.ContentBlocked | Flag.AccountPolicy | Flag.ContextOverflow | Flag.AuthFailed | Flag.PayloadRejected | Flag.SilentAbort | Flag.UserInterrupt | Flag.Abort | Flag.Grammar | Flag.FastModeUnsupported | Flag.OAuthExpiry;
 const RETRIABLE_KINDS = Flag.Transient | Flag.UsageLimit | Flag.ThinkingLoop | Flag.StaleResponsesItem | Flag.ProviderFinishError | Flag.EmptyResponse;
-const OVERFLOW_PATTERNS = [
+const CONTEXT_OVERFLOW_EVIDENCE_PATTERNS = [
 	/prompt is too long/i,
 	/input is too long for requested model/i,
 	/exceeds the context window/i,
@@ -651,7 +724,6 @@ const OVERFLOW_PATTERNS = [
 	/maximum prompt length is \d+/i,
 	/reduce the length of the messages/i,
 	/maximum context length is \d+ tokens/i,
-	/exceeds the limit of \d+/i,
 	/exceeds the available context size/i,
 	/requested tokens?.*exceed.*context (window|length|size)/i,
 	/context (window|length|size).*(exceeded|overflow|too small)/i,
@@ -663,26 +735,94 @@ const OVERFLOW_PATTERNS = [
 	/context[_ ]length[_ ]exceeded/i,
 	/too many tokens/i,
 	/token limit exceeded/i,
-	/request_too_large/i,
-	/request exceeds the maximum size/i,
-	/payload too large/i,
-	/entity too large/i,
-	/\b413\b.*\b(request|payload|entity)\b.*\btoo large\b/i,
+	/request_too_large[^\n]*\btokens?\b/i,
+	/\btokens?\b[^\n]*request_too_large/i,
 	/model_context_window_exceeded/i,
-	/prompt filled the context window/i
+	/prompt filled the context window/i,
+	/exceeds the limit of \d+ tokens?\b/i,
+	/chat history exceeds the \d+-message limit/i
 ];
+const GENERIC_LIMIT_OVERFLOW_PATTERN = /exceeds the limit of \d+/i;
+const OVERFLOW_PATTERNS = [...CONTEXT_OVERFLOW_EVIDENCE_PATTERNS, GENERIC_LIMIT_OVERFLOW_PATTERN];
+/** Token-context evidence only — excludes GENERIC_LIMIT_OVERFLOW_PATTERN which media budgets also match (#9235). */
+function hasTokenContextOverflowEvidence(text) {
+	return CONTEXT_OVERFLOW_EVIDENCE_PATTERNS.some((p) => p.test(text));
+}
+/** Checks every cause-chain link; a wrapper whose nested 413 names the token budget stays overflow-only (#9235). */
+function hasCauseTokenContextOverflowEvidence(error) {
+	const seen = /* @__PURE__ */ new Set();
+	let link = error;
+	while (link !== void 0 && link !== null) {
+		if (typeof link !== "object") {
+			if (typeof link === "string" && hasTokenContextOverflowEvidence(link)) return true;
+			break;
+		}
+		if (seen.has(link)) break;
+		seen.add(link);
+		if ("message" in link) {
+			const message = link.message;
+			if (typeof message === "string" && hasTokenContextOverflowEvidence(message)) return true;
+		}
+		if ("cause" in link) {
+			link = link.cause;
+			continue;
+		}
+		break;
+	}
+	return false;
+}
 const OVERFLOW_NO_BODY_PATTERN = /\b4(00|13)\s*(status code)?\s*\(no body\)/i;
+const PAYLOAD_REJECTION_PATTERNS = [
+	/\b413\s*(?:status code\s*)?\(no body\)/i,
+	/\b413\b[^.\n]{0,120}\b(?:request|payload|entity|body)\b[^.\n]{0,60}\b(?:exceed|too large|limit)/i,
+	/request_too_large/i,
+	/(?:payload|entity) too large/i,
+	/request exceeds the maximum (?:size|number of bytes)/i
+];
+function matchesPayloadRejectionText(text) {
+	if (!PAYLOAD_REJECTION_PATTERNS.some((p) => p.test(text))) return false;
+	if (hasTokenContextOverflowEvidence(text)) return false;
+	return true;
+}
 const TIMEOUT_PATTERN = /\b(?:operation\s+)?timed?\s*out\b|\btimeout\b|\bstream stall\b/i;
 const TRANSIENT_ENVELOPE_PATTERN = /anthropic stream envelope error:/i;
-const TRANSIENT_ENVELOPE_BEFORE_START_PATTERN = /before message_start/i;
+const TRANSIENT_ENVELOPE_TRUNCATION_PATTERN = /before message_(?:start|stop)/i;
 const STREAM_READ_ERROR_PATTERN = /stream[_ -]?read[_ -]?error/i;
-const TRANSIENT_TRANSPORT_PATTERN = /\b(?:no[_ -]?capacity|(?:high|peak)[ _-]?demand|(?:at|over|insufficient)[ _-]?capacity|capacity[ _-]?(?:exceeded|exhausted)|peak[ _-]?load)\b|overloaded|provider.?returned.?error|rate.?limit|too many requests|429|500|502|503|504|service.?unavailable|server.?error|internal.?error|retry your request|network.?error|connection.?error|connection.?refused|unable.?to.?connect\.\s*is the computer able to access the url\?|other side closed|fetch failed|upstream.?connect|upstream.?request.?failed|reset before headers|socket hang up|timed? out|timeout|terminated|retry delay|stream stall|no error details in response|HTTP2(?:StreamReset|RefusedStream|EnhanceYourCalm)|nghttp2_(?:internal_error|refused_stream)|stream closed with error code nghttp2_(?:internal_error|refused_stream)|malformed.?function.?call/i;
+/** Python h2/httpx diagnostics forwarded through provider or proxy error events. */
+const PYTHON_HTTP2_STREAM_RESET_PATTERN = /<StreamReset stream_id:\d+, error_code:(?:2|7), remote_reset:True>/;
+/** Python h11/httpx EOF while reading an HTTP/1.1 chunked response body. */
+const PYTHON_HTTP_INCOMPLETE_CHUNK_PATTERN = /peer closed connection without sending complete message body \(incomplete chunked read\)/;
+const TRANSIENT_TRANSPORT_PATTERN = /\b(?:no[_ -]?capacity|(?:high|peak)[ _-]?demand|(?:at|over|insufficient)[ _-]?capacity|capacity[ _-]?(?:exceeded|exhausted)|peak[ _-]?load)\b|overloaded|provider.?returned.?error|rate.?limit|too many requests|auth-gateway\s+5\d{2}(?=[:\s]|$)|\b(?:429|500|502|503|504)\b|service.?unavailable|server.?error|internal.?error|retry your request|network.?error|connection.?error|connection.?refused|unable.?to.?connect\.\s*is the computer able to access the url\?|other side closed|fetch failed|upstream.?connect|upstream.?request.?failed|reset before headers|socket hang up|timed? out|timeout|terminated|retry delay|stream stall|no error details in response|HTTP2(?:StreamReset|RefusedStream|EnhanceYourCalm)|nghttp2_(?:internal_error|refused_stream)|stream closed with error code nghttp2_(?:internal_error|refused_stream)|malformed.?function.?call/i;
 const AUTH_FAILURE_PATTERN = /\b(?:401|403|unauthorized|forbidden|authentication|auth[_ ]?unavailable|no auth available|(?:invalid|no)[_ ]?api[_ ]?key)\b/i;
 const MALFORMED_FUNCTION_CALL_PATTERN = /\bmalformed.?function.?call\b/i;
 const PROVIDER_FINISH_ERROR_PATTERN = /\bProvider (?:returned error finish_reason|finish_reason:\s*error)\b/i;
 const EMPTY_RESPONSE_PATTERN = /\bthought-only response without final output\b/i;
 const CONTENT_FILTER_PATTERN = /\b(?:incomplete:\s*)?content_filter\b/i;
 const ACCOUNT_POLICY_PATTERN = /\bcyber_policy\b|trusted access for cyber/i;
+const CODEX_CHATGPT_ACCOUNT_MODEL_POLICY_PATTERN = /\bThe ['"]([^'"\r\n]+)['"] model is not supported when using Codex with a ChatGPT account\./i;
+const CODEX_CHATGPT_ACCOUNT_MODEL_MAX_LENGTH = 256;
+const CURSOR_PLAN_POLICY_MARKER_PATTERN = /\bERROR_RATE_LIMITED_CHANGEABLE\b/i;
+const CURSOR_PLAN_POLICY_PATTERN = /\bNamed models unavailable\b|\bModel unavailable on\b|\bFree plans can only use\b/i;
+function isCursorPlanPolicyText(text) {
+	return CURSOR_PLAN_POLICY_MARKER_PATTERN.test(text) && CURSOR_PLAN_POLICY_PATTERN.test(text);
+}
+function normalizeCodexChatGPTAccountPolicyModel(modelId) {
+	if (typeof modelId !== "string") return void 0;
+	const separator = modelId.lastIndexOf("/");
+	const bareModelId = (separator === -1 ? modelId : modelId.slice(separator + 1)).trim().toLowerCase();
+	if (!bareModelId || bareModelId.length > CODEX_CHATGPT_ACCOUNT_MODEL_MAX_LENGTH || bareModelId.includes("\0")) return;
+	return bareModelId;
+}
+function codexChatGPTAccountPolicyModelFromText(text) {
+	const modelId = CODEX_CHATGPT_ACCOUNT_MODEL_POLICY_PATTERN.exec(text)?.[1]?.trim();
+	return normalizeCodexChatGPTAccountPolicyModel(modelId) === void 0 ? void 0 : modelId;
+}
+function isCodexChatGPTAccountPolicyText(text, provider, modelId) {
+	if (provider !== "openai-codex") return false;
+	const deniedIdentity = normalizeCodexChatGPTAccountPolicyModel(codexChatGPTAccountPolicyModelFromText(text));
+	const requestedIdentity = normalizeCodexChatGPTAccountPolicyModel(modelId);
+	return deniedIdentity !== void 0 && deniedIdentity === requestedIdentity;
+}
 const STALE_RESPONSE_ITEM_PATTERNS = [/\bItem with id ['"][^'"]+['"] not found\.?/i, /previous[ _]?response/i];
 const STALE_RESPONSE_ITEM_DETAIL_PATTERN = /not[ _]?found|invalid|expired|stale|zero[ _-]?data[ _-]?retention/i;
 /**
@@ -692,8 +832,7 @@ const STALE_RESPONSE_ITEM_DETAIL_PATTERN = /not[ _]?found|invalid|expired|stale|
 * output, so callers strip {@link Flag.Transient} when this matches.
 */
 const LLAMA_CPP_TOOL_CALL_PARSE_PATTERN = /failed to parse tool call arguments as json|\[json\.exception\.parse_error\.101\]/i;
-const COPILOT_TRANSIENT_MODEL_CODES = { model_not_supported: true };
-const COPILOT_TRANSIENT_MODEL_PATTERN = /model_not_supported/i;
+const GENERATION_NAN_PATTERN = /floating[ _-]?point nan\b.*\bdetected in generation/is;
 const GRAMMAR_TOO_LARGE_PATTERN = /compiled grammar/i;
 const GRAMMAR_TOO_LARGE_DETAIL_PATTERN = /too large/i;
 const SCHEMA_TOO_COMPLEX_PATTERN = /schema/i;
@@ -722,7 +861,7 @@ function matchesFastModeUnsupported(message, errorStatus) {
 	if (errorStatus === 400 && INVALID_REQUEST_PATTERN.test(message) && FAST_MODE_SPEED_PARAM_PATTERN.test(message) && FAST_MODE_NOT_SUPPORTED_PATTERN.test(message)) return true;
 	return errorStatus === 429 && FAST_MODE_RATE_LIMIT_PATTERN.test(message) && FAST_MODE_ENTITLEMENT_PATTERN.test(message);
 }
-Flag.ThinkingLoop, Flag.Transient, Flag.Timeout, Flag.UsageLimit, Flag.StaleResponsesItem, Flag.MalformedFunctionCall, Flag.ProviderFinishError, Flag.EmptyResponse, Flag.ContentBlocked, Flag.AccountPolicy, Flag.ContextOverflow, Flag.AuthFailed, Flag.SilentAbort, Flag.UserInterrupt, Flag.Abort;
+Flag.ThinkingLoop, Flag.Transient, Flag.Timeout, Flag.UsageLimit, Flag.StaleResponsesItem, Flag.MalformedFunctionCall, Flag.ProviderFinishError, Flag.EmptyResponse, Flag.ContentBlocked, Flag.AccountPolicy, Flag.ContextOverflow, Flag.PayloadRejected, Flag.AuthFailed, Flag.SilentAbort, Flag.UserInterrupt, Flag.Abort;
 const STATUS_MESSAGE_PATTERNS = [
 	/\bstatus(?:_code)?[:=]\s*(\d{3})\b/i,
 	/\bstatus\s+(\d{3})\b/i,
@@ -740,6 +879,7 @@ function is(id, flag) {
 }
 function retriable(id, opts) {
 	if (is(id, Flag.ContentBlocked)) return false;
+	if (is(id, Flag.PayloadRejected)) return false;
 	if (opts?.replayUnsafe) return false;
 	if (is(id, Flag.MalformedFunctionCall)) return true;
 	return ((id ?? 0) & RETRIABLE_KINDS) !== 0;
@@ -783,7 +923,7 @@ function isStreamReadErrorText(text) {
 	return STREAM_READ_ERROR_PATTERN.test(text);
 }
 function isTransientErrorText(text) {
-	return isUnexpectedSocketCloseMessage(text) || isStreamReadErrorText(text) || TRANSIENT_ENVELOPE_PATTERN.test(text) && TRANSIENT_ENVELOPE_BEFORE_START_PATTERN.test(text) || TRANSIENT_TRANSPORT_PATTERN.test(text);
+	return isUnexpectedSocketCloseMessage(text) || isStreamReadErrorText(text) || PYTHON_HTTP2_STREAM_RESET_PATTERN.test(text) || PYTHON_HTTP_INCOMPLETE_CHUNK_PATTERN.test(text) || TRANSIENT_ENVELOPE_PATTERN.test(text) && TRANSIENT_ENVELOPE_TRUNCATION_PATTERN.test(text) || TRANSIENT_TRANSPORT_PATTERN.test(text);
 }
 function isTimeoutText(text) {
 	return TIMEOUT_PATTERN.test(text);
@@ -806,31 +946,35 @@ function isContentBlockedText(text) {
 function matchesOverflowText(text) {
 	return OVERFLOW_PATTERNS.some((p) => p.test(text)) || OVERFLOW_NO_BODY_PATTERN.test(text);
 }
-function classifyText(errorMessage, errorStatus, api) {
+function classifyText(errorMessage, errorStatus, priorTokenOverflowEvidence = false, api, provider, modelId) {
 	let kinds = 0;
 	if (errorMessage) {
 		if (matchesOverflowText(errorMessage)) kinds |= Flag.ContextOverflow;
+		if (matchesPayloadRejectionText(errorMessage)) kinds |= Flag.PayloadRejected;
 		if (isMalformedFunctionCallText(errorMessage)) kinds |= Flag.MalformedFunctionCall;
 		if (isProviderFinishErrorText(errorMessage)) kinds |= Flag.ProviderFinishError;
 		if (EMPTY_RESPONSE_PATTERN.test(errorMessage)) kinds |= Flag.EmptyResponse | Flag.Transient;
 		if (isContentBlockedText(errorMessage)) kinds |= Flag.ContentBlocked;
-		if (ACCOUNT_POLICY_PATTERN.test(errorMessage)) kinds |= Flag.AccountPolicy | Flag.ContentBlocked;
+		if (ACCOUNT_POLICY_PATTERN.test(errorMessage) || isCodexChatGPTAccountPolicyText(errorMessage, provider, modelId) || provider === "cursor" && isCursorPlanPolicyText(errorMessage)) kinds |= Flag.AccountPolicy | Flag.ContentBlocked;
 		if (isAuthFailureText(errorMessage)) kinds |= Flag.AuthFailed;
 		const statusClean = errorStatus ? errorStatus : status({ message: errorMessage }) ?? void 0;
 		const cleanMessage = errorMessage;
 		const isOpaque = isOpaqueStatusBody(cleanMessage);
 		const isLimitStatus = isUsageLimitStatus(statusClean);
 		const reason = parseRateLimitReason(cleanMessage);
-		const isBillingCapStatus = statusClean === 402;
-		if (!(reason === "CONCURRENT_LIMIT" && !isBillingCapStatus) && (matchesUsageLimitText(cleanMessage) || (statusClean === 403 || statusClean === void 0) && isAccountScopedCapText(cleanMessage) || isLimitStatus && (isOpaque || reason === "QUOTA_EXHAUSTED" || isBillingCapStatus && reason === "CONCURRENT_LIMIT"))) kinds |= Flag.UsageLimit;
+		const is402BillingCap = statusClean === 402 && is402BillingCapBody(cleanMessage);
+		if (!(reason === "CONCURRENT_LIMIT" && statusClean !== 402) && (is402BillingCap || matchesUsageLimitText(cleanMessage) || (statusClean === 403 || statusClean === void 0) && isAccountScopedCapText(cleanMessage) || isLimitStatus && (isOpaque || reason === "QUOTA_EXHAUSTED"))) kinds |= Flag.UsageLimit;
 		if (isTimeoutText(errorMessage)) kinds |= Flag.Transient | Flag.Timeout;
 		else if (isTransientErrorText(errorMessage)) kinds |= Flag.Transient;
 		if (reason === "CONCURRENT_LIMIT") kinds |= Flag.Transient;
 		if ((api === "openai-responses" || api === "openai-codex-responses") && isStaleResponsesText(errorMessage)) kinds |= Flag.StaleResponsesItem;
-		if (statusClean === 400 && COPILOT_TRANSIENT_MODEL_PATTERN.test(cleanMessage)) kinds |= Flag.Transient;
+		if (statusClean === 400 && GENERATION_NAN_PATTERN.test(cleanMessage)) kinds |= Flag.Transient;
 		if (matchesStrictToolsRejection(cleanMessage, statusClean)) kinds |= Flag.Grammar;
 		if (matchesFastModeUnsupported(cleanMessage, statusClean)) kinds |= Flag.FastModeUnsupported;
 	}
+	const statusEvidence = errorStatus ?? (errorMessage ? status({ message: errorMessage }) : void 0);
+	if (statusEvidence === 413 && !priorTokenOverflowEvidence && !(errorMessage && hasTokenContextOverflowEvidence(errorMessage))) kinds |= Flag.PayloadRejected;
+	if (statusEvidence === 402 && (errorMessage === void 0 || isOpaqueStatusBody(errorMessage))) kinds |= Flag.UsageLimit;
 	if (kinds !== 0) return create(kinds);
 	const fallbackStatus = errorStatus ?? (errorMessage ? status({ message: errorMessage }) : void 0);
 	if (fallbackStatus === 401 || fallbackStatus === 403) return create(Flag.AuthFailed);
@@ -839,6 +983,7 @@ function classifyText(errorMessage, errorStatus, api) {
 function classify(error, api) {
 	let kinds = 0;
 	const seen = /* @__PURE__ */ new Set();
+	const causeTokenEvidence = hasCauseTokenContextOverflowEvidence(error);
 	let link = error;
 	while (link !== void 0 && link !== null) {
 		if (typeof link === "object") {
@@ -855,7 +1000,7 @@ function classify(error, api) {
 		else if (link instanceof ProviderHttpError) {
 			let linkKinds = 0;
 			const { status: codeStatus, code } = link;
-			if (code === "usage_limit_reached" || code === "insufficient_quota" && !isDashScopeTokenLimitText(link.message)) linkKinds |= Flag.UsageLimit;
+			if (code === "usage_limit_reached" || code === "insufficient_quota" && !isDashScopeTokenLimitText(link.message) || codeStatus === 402 && (code === "payment_required" || code === "deactivated_workspace" || is402BillingCapBody(link.message))) linkKinds |= Flag.UsageLimit;
 			if (code === "overloaded_error" || code === "rate_limit_error") linkKinds |= Flag.Transient;
 			if ((codeStatus === 401 || codeStatus === 403) && !(codeStatus === 403 && parseRateLimitReason(link.message) === "CONCURRENT_LIMIT")) linkKinds |= Flag.AuthFailed;
 			else if (codeStatus === 429) {
@@ -867,7 +1012,7 @@ function classify(error, api) {
 		if (link instanceof Error) linkMessage = link.message;
 		else if (typeof link === "string") linkMessage = link;
 		else if (typeof link === "object" && "message" in link && typeof link.message === "string") linkMessage = link.message;
-		const textId = classifyText(linkMessage, status(link), api);
+		const textId = classifyText(linkMessage, status(link), causeTokenEvidence, api);
 		kinds |= textId & KIND_MASK;
 		link = typeof link === "object" && "cause" in link ? link.cause : void 0;
 	}
@@ -897,41 +1042,25 @@ function isGrammarError(error) {
 function isFastModeUnsupported(error) {
 	return is(classify(error), Flag.FastModeUnsupported);
 }
+const CLINE_PASS_SURFACE_GATE_PATTERN = /only available via cline product surfaces/i;
 /**
-* Depth-bounded search for a provider error `code`. SDK error objects keep the
-* parsed response body on `.error`, and Copilot's body is itself
-* `{ error: { code } }`, so the code sits up to two envelopes below the thrown
-* error depending on which SDK produced it.
+* Cline's gateway gates some roster entries (certain free-tier models) to its
+* own product surfaces with a 403. The API key is valid — the restriction is
+* per-model client policy — so it must neither rotate sibling credentials (they
+* fail identically) nor surface as an auth failure.
 */
-function providerErrorCode(error) {
-	let node = error;
-	for (let depth = 0; depth < 3; depth++) {
-		if ("code" in node && typeof node.code === "string") return node.code;
-		if (!("error" in node)) return void 0;
-		const nested = node.error;
-		if (!nested || typeof nested !== "object") return void 0;
-		node = nested;
-	}
-}
-/**
-* GitHub Copilot 400 `model_not_supported` response for a model advertised by
-* `/models` — transient fleet skew, not a malformed request. Reads the
-* structural `code` through the SDK/body envelopes, then falls back to the
-* stringified body both SDK families put in `message`.
-*/
-function isCopilotTransientModelError(error) {
-	if (!error || typeof error !== "object" || status(error) !== 400) return false;
-	const code = providerErrorCode(error);
-	if (code !== void 0 && Object.hasOwn(COPILOT_TRANSIENT_MODEL_CODES, code)) return true;
-	const message = "message" in error ? error.message : void 0;
-	return typeof message === "string" && COPILOT_TRANSIENT_MODEL_PATTERN.test(message);
+function isClinePassSurfaceGateMessage(errorMessage) {
+	return errorMessage !== void 0 && CLINE_PASS_SURFACE_GATE_PATTERN.test(errorMessage);
 }
 function classifyMessage(message) {
 	const existingId = message.errorId;
 	const currentStatus = message.errorStatus ?? statusFromId(existingId);
-	const textId = classifyText(message.errorMessage, currentStatus, message.api);
+	const existingOverflowOnly = existingId !== void 0 && is(existingId, Flag.ContextOverflow) && !is(existingId, Flag.PayloadRejected);
+	const classificationMessage = message.errorClassificationMessage ?? message.errorMessage;
+	const textId = classifyText(classificationMessage, currentStatus, existingOverflowOnly, message.api, message.provider, message.model);
 	let kinds = ((existingId ?? 0) | textId) & KIND_MASK;
-	if (message.errorMessage && LLAMA_CPP_TOOL_CALL_PARSE_PATTERN.test(message.errorMessage)) kinds &= ~Flag.Transient;
+	if (currentStatus === 413 && classificationMessage && hasTokenContextOverflowEvidence(classificationMessage) && !(textId & Flag.PayloadRejected)) kinds &= ~Flag.PayloadRejected;
+	if (classificationMessage && LLAMA_CPP_TOOL_CALL_PARSE_PATTERN.test(classificationMessage)) kinds &= ~Flag.Transient;
 	const id = kinds !== 0 ? create(kinds) : statusFromId(textId) ?? statusFromId(existingId) ?? currentStatus ?? 0;
 	message.errorId = id;
 	return id;
@@ -967,7 +1096,7 @@ function isRetryableStreamEnvelopeError(error) {
 	return error instanceof Error && STREAM_EVENT_ORDER_PATTERN.test(error.message);
 }
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/error/abort.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/error/abort.ts
 /**
 * A request was cancelled — by the caller's `AbortSignal` or a provider-local
 * watchdog. Carries the {@link Flag.Abort} classification structurally so retry
@@ -985,7 +1114,7 @@ var AbortError = class extends Error {
 	}
 };
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/error/auth.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/error/auth.ts
 /**
 * No API key / credential was available to dispatch a request.
 *
@@ -1029,7 +1158,28 @@ var LoginCancelledError = class extends Error {
 	}
 };
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/utils/retry-after.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/error/oauth.ts
+/**
+* A failure inside an interactive OAuth / device-code login flow. The `kind`
+* pinpoints the stage. Timeout/polling are classified transient; everything
+* else is a hard auth failure. The auth-retry classifier separately treats
+* `token-refresh` as an explicit request to refresh and replay once.
+*/
+var OAuthError = class extends Error {
+	kind;
+	provider;
+	status;
+	constructor(message, options = {}) {
+		super(message, options.cause === void 0 ? void 0 : { cause: options.cause });
+		this.name = "OAuthError";
+		this.kind = options.kind ?? "http";
+		this.provider = options.provider;
+		this.status = options.status;
+		attach(this, this.kind === "timeout" || this.kind === "polling" ? create(Flag.Transient) : create(Flag.AuthFailed));
+	}
+};
+//#endregion
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/utils/retry-after.ts
 const RETRY_AFTER_HINT = "retry-after-ms=";
 function formatErrorMessageWithRetryAfter(error, headers) {
 	const message = error instanceof Error ? error.message : JSON.stringify(error);
@@ -1105,45 +1255,7 @@ function parseResetHeader(value, unit) {
 	return Math.ceil(targetMs - nowMs);
 }
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/utils/retry.ts
-const COPILOT_MODEL_RETRY_MAX_ATTEMPTS = 8;
-const COPILOT_GENERIC_RETRY_MAX_ATTEMPTS = 3;
-const COPILOT_MODEL_RETRY_BASE_DELAY_MS = 400;
-/** Longest server-requested backoff we are willing to sit out before giving up. */
-const COPILOT_RETRY_AFTER_MAX_WAIT_MS = 3e4;
-/**
-* Wrap an initial Copilot request so transient `model_not_supported` 400s are
-* retried a small number of times. No-op for non-Copilot providers.
-*
-* The callback **MUST** create a fresh in-flight request each invocation — a
-* once-consumed AsyncIterable cannot be re-iterated.
-*/
-async function callWithCopilotModelRetry(fn, options) {
-	if (options.provider !== "github-copilot") return fn();
-	let lastError;
-	const retryBaseDelayMs = options.retryBaseDelayMs ?? COPILOT_MODEL_RETRY_BASE_DELAY_MS;
-	for (let attempt = 0; attempt < COPILOT_MODEL_RETRY_MAX_ATTEMPTS; attempt++) try {
-		return await fn();
-	} catch (error) {
-		lastError = error;
-		if (options.signal?.aborted) throw error;
-		const transientModelError = isCopilotTransientModelError(error);
-		if (!transientModelError && !isRetryableError(error)) throw error;
-		if (attempt >= (transientModelError ? COPILOT_MODEL_RETRY_MAX_ATTEMPTS : COPILOT_GENERIC_RETRY_MAX_ATTEMPTS) - 1) break;
-		let delayMs = transientModelError ? retryBaseDelayMs : retryBaseDelayMs * (attempt + 1);
-		if (!transientModelError) {
-			if (status(error) !== void 0) {
-				const retryAfterMs = getRetryAfterMsFromHeaders(getHeadersFromError(error));
-				if (retryAfterMs === void 0 || retryAfterMs > COPILOT_RETRY_AFTER_MAX_WAIT_MS) throw error;
-				delayMs = Math.max(delayMs, retryAfterMs);
-			}
-		}
-		await scheduler.wait(delayMs, { signal: options.signal });
-	}
-	throw lastError;
-}
-//#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/utils/http-inspector.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/utils/http-inspector.ts
 const SENSITIVE_HEADERS = [
 	"authorization",
 	"x-api-key",
@@ -1174,8 +1286,8 @@ function buildHttp400DumpPayload(dump, error, message) {
 *  are excluded: 429/5xx are retried, so persisting them here would write one
 *  dump per attempt. */
 function shouldDumpRejectedRequest(error) {
-	const status$2 = status(error);
-	return status$2 === 400 || status$2 === 413;
+	const status$1 = status(error);
+	return status$1 === 400 || status$1 === 413;
 }
 async function appendRawHttpRequestDumpFor400(message, error, dump) {
 	if (!dump || isBunTestRuntime() || !shouldDumpRejectedRequest(error)) return message;
@@ -1202,20 +1314,47 @@ async function finalizeErrorMessage(error, rawRequestDump, capturedErrorResponse
 * Rewrite error message for GitHub Copilot request failures.
 * Must run AFTER finalizeErrorMessage since it replaces the message entirely.
 *
-* 400 `model_not_supported` = Copilot fleet skew. A model that `/models`
-*        advertises can flap between 200 and 400 because only part of
-*        Copilot's fleet has it in the integrator allowlist. After the
-*        in-request retry exhausts, surface guidance rather than the raw error.
 * 401 = token invalid/expired → credential removal is safe, prompt re-login.
 * 403 = token valid but access denied (plan, model policy, org restriction) →
 *       do NOT reuse the auth-failed string (which triggers credential removal).
+* 400 is left verbatim: GitHub's body names the cause (`model_not_supported`,
+*       `model_not_available_for_integrator` with its `Available models` list)
+*       and a rewrite only hides it.
 */
 function rewriteCopilotError(errorMessage, error, provider) {
 	if (provider !== "github-copilot") return errorMessage;
-	const status$1 = status(error);
-	if (status$1 === 401) return `GitHub Copilot authentication failed (HTTP 401). Your token may have been revoked. Please re-login with /login github-copilot`;
-	if (status$1 === 403) return `GitHub Copilot access denied (HTTP 403). Your account may not have access to this model or feature. Check your Copilot plan or model policy settings.`;
-	if (isCopilotTransientModelError(error)) return `GitHub Copilot rejected this model (HTTP 400) after retries: only part of its fleet currently serves this model id, even though /models advertises it. Try again in a few seconds or switch to a model Copilot serves fleet-wide (claude-opus-4.7, claude-sonnet-4.5, gpt-4.1).`;
+	const status$2 = status(error);
+	if (status$2 === 401) return `GitHub Copilot authentication failed (HTTP 401). Your token may have been revoked. Please re-login with /login github-copilot`;
+	if (status$2 === 403) return `GitHub Copilot access denied (HTTP 403). Your token is valid but the account may not have access to this model or feature. Check your Copilot plan or model policy settings. Business organizations can also restrict which clients may call the API: omp sends Copilot-Integration-Id copilot-chat by default (COPILOT_INTEGRATION_ID overrides it) and retries a denied default-identity request once as the Copilot CLI (copilot-developer-cli). If both identities are denied, ask your org admin to allow one of them; if you pinned an identity, try the other.`;
+	return errorMessage;
+}
+const CLINE_PASS_NOT_SUBSCRIBED_PATTERN = /not subscribed to required model plan|no access to clinepass subscription models/i;
+const CLINE_PASS_ORG_ACCOUNT_PATTERN = /organization accounts cannot use individual model inference subscriptions/i;
+const CLINE_PASS_MODEL_NOT_FOUND_PATTERN = /model not found/i;
+/**
+* Rewrite error messages for ClinePass request failures. Gated to the
+* cline-pass provider: the "model not found" marker is too generic to match
+* for other hosts.
+*
+* not-subscribed (400) = the key is valid but the account has no ClinePass
+*        subscription; free-tier models remain usable on the same key.
+* org restriction (400) = organization accounts cannot use individual
+*        inference subscriptions; a personal-account key is required.
+* model-not-found (400) = roster rotation removed the model since selection;
+*        the fix is reselection, not retry. (Quota windows — "clinepass
+*        limit", "free limit reached on model" — are classified upstream in
+*        error/rate-limit and need no rewrite.)
+* surface-gate (403) = the model is restricted to Cline's official clients.
+*        Requests carry the mirrored CLI identity headers, so reaching this
+*        means Cline's gate policy changed; the classifier exempts it from
+*        credential rotation (sibling keys fail identically).
+*/
+function rewriteClinePassError(errorMessage, provider) {
+	if (provider !== "cline-pass") return errorMessage;
+	if (CLINE_PASS_NOT_SUBSCRIBED_PATTERN.test(errorMessage)) return "This model requires a ClinePass subscription. Free-tier models (marked \"(free)\" in the picker) work with any Cline account.";
+	if (CLINE_PASS_ORG_ACCOUNT_PATTERN.test(errorMessage)) return "ClinePass is unavailable for organization accounts: individual inference subscriptions are personal-plan only. Log in with a personal Cline API key.";
+	if (isClinePassSurfaceGateMessage(errorMessage)) return "Cline restricts this model to its official product surfaces and the mirrored CLI client identity was not accepted. Pick another model with /model and report the regression — the header mirror may need updating.";
+	if (CLINE_PASS_MODEL_NOT_FOUND_PATTERN.test(errorMessage)) return "Cline removed this model from the roster since it was selected. Pick another with /model — the roster refreshes automatically while your API key is configured.";
 	return errorMessage;
 }
 function sanitizeDump(dump) {
@@ -1278,7 +1417,7 @@ function isObject(value) {
 	return typeof value === "object" && value !== null && !Array.isArray(value);
 }
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/error/format.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/error/format.ts
 function rewriteOllamaToolCallJsonError(message) {
 	if (!LLAMA_CPP_TOOL_CALL_PARSE_PATTERN.test(message)) return message;
 	return `Local Ollama model emitted malformed tool-call JSON and llama.cpp rejected it (HTTP 500). This is usually a deterministic model-output failure after context degradation, not a transient server outage; reload the model or reduce context, then retry.\n${message}`;
@@ -1295,11 +1434,12 @@ function rewriteOllamaToolCallJsonError(message) {
 async function formatMessage(error, opts = {}) {
 	let message = opts.rawRequestDump ? await finalizeErrorMessage(error, opts.rawRequestDump, opts.capturedErrorResponse) : formatErrorMessageWithRetryAfter(error);
 	if (opts.provider === "github-copilot") message = rewriteCopilotError(message, error, opts.provider);
+	if (opts.provider === "cline-pass") message = rewriteClinePassError(message, opts.provider);
 	if (opts.provider === "ollama") message = rewriteOllamaToolCallJsonError(message);
 	return message;
 }
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/error/finalize.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/error/finalize.ts
 /**
 * Build the complete error bundle for a provider catch block, replacing the
 * `stopReason` / `errorStatus` / `errorId` / `errorMessage` boilerplate.
@@ -1321,6 +1461,8 @@ async function finalize(error, opts = {}) {
 	return {
 		id: classifyMessage({
 			api: opts.api,
+			provider: opts.provider,
+			model: opts.model,
 			errorId: classify(error, opts.api),
 			errorMessage: message,
 			errorStatus: currentStatus
@@ -1331,27 +1473,7 @@ async function finalize(error, opts = {}) {
 	};
 }
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/error/oauth.ts
-/**
-* A failure inside an interactive OAuth / device-code login flow. The `kind`
-* pinpoints the stage. Timeout/polling are classified transient; everything
-* else is a hard auth failure so the credential layer does not silently retry.
-*/
-var OAuthError = class extends Error {
-	kind;
-	provider;
-	status;
-	constructor(message, options = {}) {
-		super(message, options.cause === void 0 ? void 0 : { cause: options.cause });
-		this.name = "OAuthError";
-		this.kind = options.kind ?? "http";
-		this.provider = options.provider;
-		this.status = options.status;
-		attach(this, this.kind === "timeout" || this.kind === "polling" ? create(Flag.Transient) : create(Flag.AuthFailed));
-	}
-};
-//#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/error/provider.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/error/provider.ts
 /**
 * A non-HTTP provider failure: a truncated stream, an error stop reason, an
 * empty body, a malformed envelope, or a runtime fault. For non-2xx HTTP
@@ -1375,7 +1497,7 @@ var DevinApiError = class extends ProviderHttpError {
 	name = "DevinApiError";
 };
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/error/retryable.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/error/retryable.ts
 /**
 * Whether a numeric HTTP status is in the canonical transient/retryable set:
 * 408 (Request Timeout), 429 (Too Many Requests), and any 5xx.
@@ -1398,12 +1520,11 @@ function isTransientTransportMessage(message) {
 * here — they are owned by the credential-rotation layer (auth-gateway /
 * `streamSimple` a/b/c policy), not this seconds-scale provider backoff.
 *
-* Provider-specific transient cases are injected via {@link ProviderRetryableHooks}
-* so this stays free of provider imports.
+* Every 4xx other than 408/429 is terminal: a request the provider rejected
+* as malformed, unauthorized, or unentitled fails identically on replay.
 */
-function isProviderRetryableError(error, hooks = {}) {
+function isProviderRetryableError(error) {
 	if (!(error instanceof Error)) return false;
-	if (hooks.isProviderTransient?.(error)) return true;
 	if (isUsageLimit(error)) return false;
 	const httpStatus = status(error);
 	if (httpStatus !== void 0 && httpStatus >= 400 && httpStatus < 500 && httpStatus !== 408 && httpStatus !== 429) return false;
@@ -1412,7 +1533,7 @@ function isProviderRetryableError(error, hooks = {}) {
 	return isRetryableError(error);
 }
 //#endregion
-//#region node_modules/.pnpm/@oh-my-pi+pi-ai@17.3.5/node_modules/@oh-my-pi/pi-ai/src/error/validation.ts
+//#region node_modules/.pnpm/@oh-my-pi+pi-ai@18.1.17/node_modules/@oh-my-pi/pi-ai/src/error/validation.ts
 /**
 * Caller-supplied input failed validation before/while building a provider
 * request: bad request body, malformed tool arguments, unsupported content
@@ -1446,4 +1567,4 @@ var StreamTimeoutError = class extends Error {
 	}
 };
 //#endregion
-export { isTransientStreamParseError as A, OpenAIHttpError as B, classifyMessage as C, isFastModeUnsupported as D, isCopilotTransientModelError as E, AnthropicStreamEnvelopeError as F, fetchWithRetry as G, AwsCredentialsError as H, BedrockApiError as I, GeminiCliApiError as L, AnthropicApiError as M, AnthropicConnectionError as N, isGrammarError as O, AnthropicConnectionTimeoutError as P, GoogleApiError as R, classify as S, is as T, EventStreamFrameError as U, ProviderHttpError as V, extractHttpStatusFromError as W, OnPromptRequiredError as _, isTransientStatus as a, LLAMA_CPP_TOOL_CALL_PARSE_PATTERN as b, OAuthError as c, callWithCopilotModelRetry as d, getHeadersFromError as f, MissingApiKeyError as g, LoginCancelledError as h, isProviderRetryableError as i, retriable as j, isStreamEnvelopeError as k, finalize as l, ApiKeyRequiredError as m, StreamTimeoutError as n, DevinApiError as o, getRetryAfterMsFromHeaders as p, ValidationError as r, ProviderResponseError as s, ConfigurationError as t, finalizeErrorMessage as u, AbortError as v, create as w, attach as x, Flag as y, OllamaApiError as z };
+export { isTransientStreamParseError as A, OpenAIHttpError as B, classify as C, isFastModeUnsupported as D, is as E, AnthropicStreamEnvelopeError as F, fetchWithRetry as G, AwsCredentialsError as H, BedrockApiError as I, GeminiCliApiError as L, AnthropicApiError as M, AnthropicConnectionError as N, isGrammarError as O, AnthropicConnectionTimeoutError as P, GoogleApiError as R, attach as S, create as T, EventStreamFrameError as U, ProviderHttpError as V, extractHttpStatusFromError as W, AbortError as _, isTransientStatus as a, PYTHON_HTTP2_STREAM_RESET_PATTERN as b, finalize as c, getRetryAfterMsFromHeaders as d, OAuthError as f, OnPromptRequiredError as g, MissingApiKeyError as h, isProviderRetryableError as i, retriable as j, isStreamEnvelopeError as k, finalizeErrorMessage as l, LoginCancelledError as m, StreamTimeoutError as n, DevinApiError as o, ApiKeyRequiredError as p, ValidationError as r, ProviderResponseError as s, ConfigurationError as t, getHeadersFromError as u, Flag as v, classifyMessage as w, PYTHON_HTTP_INCOMPLETE_CHUNK_PATTERN as x, LLAMA_CPP_TOOL_CALL_PARSE_PATTERN as y, OllamaApiError as z };