github-router 0.3.302 → 0.3.304

package/dist/{server-setup-Zn5wkosb.js → server-setup-CSkdFAPP.js}

@@ -1,12 +1,12 @@
-import { $ as UNKNOWN_EFFORT_ANCHOR, $t as countTokens, An as BUDGET_SMALL_FAST_SLUG, B as resolveAdvisorEffort, Bt as MAX_LUNA_HIGH_ALIAS_ID, Cn as provisionTreeSitterAssets, D as toolbeltEnabled, F as ADVISOR_TOOL_INSTRUCTIONS, Fn as UPSTREAM_FETCH_TIMEOUT_MS, Ft as LUNA_HAIKU_ALIAS_ID, G as repairRejectedThinkingHistory, Gn as oneMContextDisabled, Gt as maxAliasEffort, H as formatThinkingRepairDecline, Hn as upstreamMaxConnections, Ht as canonicalizeAliasModel, I as FAST_ADVISOR_TOOL_INSTRUCTIONS, In as UPSTREAM_INACTIVITY_TIMEOUT_MS, J as isControllerClosedError, Jn as withInstallLock, K as buildAnthropicErrorEvent, Kn as withOneMSuffix, Kt as maxAliasModel, L as buildAdvisorStream, Ln as generateRandomPort, Lt as LUNA_REAL_MODEL_ID, N as searchWeb, On as toolbeltPathOverride, P as ADVISOR_INTERNAL_TOOL_NAME, Pt as LUNA_DRIVER_ALIAS_ID, Q as EFFORT_ORDER, Qt as shimDefaultsToXhigh, R as injectAdvisorTool, Rn as isBudgetClaudeLead, U as rememberThinkingHistoryRepair, Un as classifyMessagesRoute, V as resolveAdvisorModel, Vn as upstreamAllowH2, Vt as MAX_LUNA_MAX_ALIAS_ID, W as repairKnownThinkingHistory, Wn as catalogAdvertises1M, Wt as isMaxModelAlias, X as readIteratorWithTimeout, Y as logStreamError, Yt as resolveModelAlias, Z as relayAnthropicStream, cn as warnOnTokenPriceDrift, dn as createChatCompletions, en as createMessages, et as bucketEffort, fn as MAX_RESPONSE_BODY_BYTES, hn as normalizeOpenAIUsage, i as assertMcpToolSurfaceConsistent, in as findLaunchBySecret, ln as resolveMcpToolTimeoutMs, lt as agentToolsEnabled, mn as parseJsonOrDiagnose, nn as getTokenCount, nt as handleMcpDelete, pn as readResponseBodyCapped, q as buildOpenAIErrorEvent, qn as withOneMSuffixForLead, rn as getTokenizerFromModel, rt as handleMcpPost, sn as assembleResponsesPayload, tn as getTextTokenCount, tt as clampEffort, un as createResponses, z as isAdvisorRequested, zt as LUNA_SONNET_ALIAS_ID } from "./peer-mcp-personas-DwEiYeJS.js";
+import { $ as UNKNOWN_EFFORT_ANCHOR, An as MAX_LUNA_HIGH_ALIAS_ID, At as countTokens, B as resolveAdvisorEffort, Bn as resolveModelAlias, Bt as resolveMcpToolTimeoutMs, Cn as withOneMSuffix, D as toolbeltEnabled, Dn as LUNA_REAL_MODEL_ID, En as LUNA_HAIKU_ALIAS_ID, F as ADVISOR_TOOL_INSTRUCTIONS, Fn as isRetiredFastModelAlias, Ft as findLaunchBySecret, G as repairRejectedThinkingHistory, Gt as parseJsonOrDiagnose, H as formatThinkingRepairDecline, Ht as createChatCompletions, I as FAST_ADVISOR_TOOL_INSTRUCTIONS, In as maxAliasEffort, J as isControllerClosedError, K as buildAnthropicErrorEvent, Kt as normalizeOpenAIUsage, L as buildAdvisorStream, Ln as maxAliasModel, Mn as canonicalizeAliasModel, Mt as getTextTokenCount, N as searchWeb, Nt as getTokenCount, P as ADVISOR_INTERNAL_TOOL_NAME, Pn as isMaxModelAlias, Pt as getTokenizerFromModel, Q as EFFORT_ORDER, R as injectAdvisorTool, Rt as assembleResponsesPayload, Sn as oneMContextDisabled, Tn as LUNA_DRIVER_ALIAS_ID, U as rememberThinkingHistoryRepair, Un as withInstallLock, Ut as MAX_RESPONSE_BODY_BYTES, V as resolveAdvisorModel, Vt as createResponses, W as repairKnownThinkingHistory, Wt as readResponseBodyCapped, X as readIteratorWithTimeout, Y as logStreamError, Z as relayAnthropicStream, an as toolbeltPathOverride, bn as classifyMessagesRoute, en as provisionTreeSitterAssets, et as bucketEffort, fn as UPSTREAM_FETCH_TIMEOUT_MS, hn as isBudgetClaudeLead, i as assertMcpToolSurfaceConsistent, it as agentToolsEnabled, jn as MAX_LUNA_MAX_ALIAS_ID, jt as createMessages, kn as LUNA_SONNET_ALIAS_ID, kt as shimDefaultsToXhigh, mn as generateRandomPort, nt as handleMcpDelete, pn as UPSTREAM_INACTIVITY_TIMEOUT_MS, q as buildOpenAIErrorEvent, rt as handleMcpPost, sn as BUDGET_SMALL_FAST_SLUG, tt as clampEffort, vn as upstreamAllowH2, wn as withOneMSuffixForLead, xn as catalogAdvertises1M, yn as upstreamMaxConnections, z as isAdvisorRequested, zt as warnOnTokenPriceDrift } from "./peer-mcp-personas-D4bHdjOx.js";
 import { t as getPackageVersion } from "./version-C8x2hQrZ.js";
 import { i as ensurePaths, t as PATHS } from "./paths-IcZW8gZx.js";
 import { t as state } from "./state-B8adscCv.js";
 import { T as copilotHeaders, _ as forwardError, a as cacheCopilotVersion, c as filterBetaHeader, d as resolveModel, f as sleep, g as HTTPError, i as tryRefreshAndRetry, l as isNullish, n as setupGitHubAgentToken, o as cacheModels, r as setupGitHubToken, s as cacheVSCodeVersion, t as setupCopilotToken, w as copilotBaseUrl, y as fetchWithTransientRetry } from "./token-CbuFyx-B.js";
 import { t as getCopilotUsage } from "./get-copilot-usage-ByjQqhdi.js";
 import { a as resolveExecutable, n as killManagedTree, o as runCommandCapture, r as parseBoolEnv, s as runCommandVoid } from "./exec-y8C_MU8A.js";
-import { r as FAST_PROFILE_ADVISOR_MODEL } from "./fast-profile-contract-K3P58nj3.js";
-import { b as stripTrailingOneMSuffix, d as maxAdvisorPinIsValid, i as MAX_PROFILE_MODELS, l as maxAdvisorEffortForModel, m as maxOpusModel, n as MAX_PROFILE_ALLOWED_LEAD_MODEL_IDS, t as MAX_PROFILE_ADVISOR_INSTRUCTIONS, u as maxAdvisorModelFromPin } from "./max-profile-contract-BTXowh6k.js";
+import { r as FAST_PROFILE_ADVISOR_MODEL } from "./fast-profile-contract-OuZQaGif.js";
+import { d as maxAdvisorPinIsValid, i as MAX_PROFILE_MODELS, l as maxAdvisorEffortForModel, m as maxOpusModel, n as MAX_PROFILE_ALLOWED_LEAD_MODEL_IDS, t as MAX_PROFILE_ADVISOR_INSTRUCTIONS, u as maxAdvisorModelFromPin, x as stripTrailingOneMSuffix } from "./max-profile-contract-CiD5wyET.js";
 import consola from "consola";
 import * as fs$2 from "node:fs";
 import fs, { existsSync } from "node:fs";
@@ -1502,7 +1502,7 @@ function collectToolFieldKeys(body) {
 //#endregion
 //#region package.json
 var name = "github-router";
-var version = "0.3.302";
+var version = "0.3.304";
 //#endregion
 //#region src/lib/approval.ts
 const awaitApproval = async () => {
@@ -4367,12 +4367,12 @@ function identityPreflightErrorResponse(c, reason, path = "/v1/messages") {
 //#region src/lib/max-request-preprocess.ts
 function maxRequestError(result) {
 	if (result.rejectedAlias) return `Router-owned model alias ${JSON.stringify(result.rejectedAlias)} is valid only for an authenticated -m max launch.`;
-	if (result.rejectedModel) return `Model ${JSON.stringify(result.rejectedModel)} is outside the fixed -m max model set (Sol, Luna, Gemini 3.7 Flash, or Opus 5).`;
+	if (result.rejectedModel) return `Model ${JSON.stringify(result.rejectedModel)} is outside the fixed -m max model set (Sol, Luna, Gemini 3.8 Flash, or Opus 5).`;
 }
 const MAX_MODEL_EFFORTS = Object.freeze({
 	"gpt-5.6-sol": "high",
 	"gpt-5.6-luna": "high",
-	"gemini-3.7-flash": "high",
+	"gemini-3.8-flash": "high",
 	"claude-opus-5": "high"
 });
 function allowedModel(base) {
@@ -4427,14 +4427,14 @@ function preprocessMaxRequest(rawBody, launch, subagentRequest = false) {
 			modified: true
 		};
 	}
-	const allowedSubagentModel = subagentRequest && base === "grok-4.6";
+	const allowedSubagentModel = subagentRequest && (base === "grok-4.6" || base === "gpt-5.6-luna");
 	if (!allowedModel(base) && !allowedSubagentModel) return {
 		body: rawBody,
 		originalModel,
 		modified: false,
 		rejectedModel: originalModel
 	};
-	const effort = allowedSubagentModel ? "medium" : MAX_MODEL_EFFORTS[base];
+	const effort = allowedSubagentModel ? base === "gpt-5.6-luna" ? "max" : "high" : MAX_MODEL_EFFORTS[base];
 	if (!effort) return {
 		body: rawBody,
 		originalModel,
@@ -4475,6 +4475,12 @@ function preprocessFastRequest(rawBody, launch, subagentRequest = false) {
 		body: rawBody,
 		modified: false
 	};
+	if (isRetiredFastModelAlias(originalModel)) return {
+		body: rawBody,
+		originalModel,
+		modified: false,
+		retiredAlias: originalModel
+	};
 	const alias = resolveModelAlias(originalModel);
 	if (alias && (launch?.profileId !== "fast" || isMaxModelAlias(originalModel))) return {
 		body: rawBody,
@@ -4495,7 +4501,7 @@ function preprocessFastRequest(rawBody, launch, subagentRequest = false) {
 	} else if (bare === "gpt-5.6-luna") effort = "max";
 	else if (bare === "gpt-5.6-sol") effort = "high";
 	else if (bare === "grok-4.6") effort = "medium";
-	else if (bare === "gemini-3.7-flash") effort = "high";
+	else if (bare === "gemini-3.8-flash") effort = "high";
 	else if (bare === "claude-opus-5") effort = "high";
 	if (!effort && !alias) return {
 		body: rawBody,
@@ -5003,8 +5009,8 @@ async function handleCompletion(c) {
 		}
 	}
 	const fastPreprocess = preprocessFastRequest(rawBody, identity.launch, subagentRequest);
-	if (fastPreprocess.rejectedAlias || fastPreprocess.rejectedModel) {
-		const message = identity.launch?.profileId === "max" ? maxRequestError(fastPreprocess) ?? "Invalid max request" : fastPreprocess.rejectedAlias ? `Router-owned model alias ${JSON.stringify(fastPreprocess.rejectedAlias)} is valid only for an authenticated -m fast launch.` : `Model ${JSON.stringify(fastPreprocess.rejectedModel)} is outside the fixed -m fast model set.`;
+	if (fastPreprocess.retiredAlias || fastPreprocess.rejectedAlias || fastPreprocess.rejectedModel) {
+		const message = fastPreprocess.retiredAlias ? `Router-owned model alias ${JSON.stringify(fastPreprocess.retiredAlias)} belongs to a retired Fast role and is no longer valid. Relaunch or select a current Fast role.` : identity.launch?.profileId === "max" ? maxRequestError(fastPreprocess) ?? "Invalid max request" : fastPreprocess.rejectedAlias ? `Router-owned model alias ${JSON.stringify(fastPreprocess.rejectedAlias)} is valid only for an authenticated -m fast launch.` : `Model ${JSON.stringify(fastPreprocess.rejectedModel)} is outside the fixed -m fast model set.`;
 		return c.json({
 			type: "error",
 			error: {
@@ -5551,8 +5557,8 @@ async function handleCountTokens(c) {
 	const rawBody = await c.req.text();
 	const subagentRequest = Boolean(c.req.header("x-claude-code-agent-id"));
 	const fastPreprocess = preprocessFastRequest(rawBody, identity.launch, subagentRequest);
-	if (fastPreprocess.rejectedAlias || fastPreprocess.rejectedModel) {
-		const message = identity.launch?.profileId === "max" ? maxRequestError(fastPreprocess) ?? "Invalid max request" : fastPreprocess.rejectedAlias ? `Router-owned model alias ${JSON.stringify(fastPreprocess.rejectedAlias)} is valid only for an authenticated -m fast launch.` : `Model ${JSON.stringify(fastPreprocess.rejectedModel)} is outside the fixed -m fast model set.`;
+	if (fastPreprocess.retiredAlias || fastPreprocess.rejectedAlias || fastPreprocess.rejectedModel) {
+		const message = fastPreprocess.retiredAlias ? `Router-owned model alias ${JSON.stringify(fastPreprocess.retiredAlias)} belongs to a retired Fast role and is no longer valid. Relaunch or select a current Fast role.` : identity.launch?.profileId === "max" ? maxRequestError(fastPreprocess) ?? "Invalid max request" : fastPreprocess.rejectedAlias ? `Router-owned model alias ${JSON.stringify(fastPreprocess.rejectedAlias)} is valid only for an authenticated -m fast launch.` : `Model ${JSON.stringify(fastPreprocess.rejectedModel)} is outside the fixed -m fast model set.`;
 		return c.json({
 			type: "error",
 			error: {
@@ -6544,8 +6550,8 @@ function parseSharedArgs(args) {
 * Copilot `/responses` (gpt) or `/chat/completions` (gemini/grok).
 *
 * This list is EXACT and STATIC — no dynamic Gemini-review append (the
-* earlier `gemini-3.1-pro-preview`-preferred / `gemini-3.7-flash`-fallback
-* row is retired: `gemini-3.7-flash` is now a first-class row on its own,
+* earlier `gemini-3.1-pro-preview`-preferred / `gemini-3.8-flash`-fallback
+* row is retired: `gemini-3.8-flash` is now a first-class row on its own,
 * always at this fixed id). A model missing from the catalog is simply
 * omitted, never substituted — see `nativeSelectableModelsInCatalog`.
 *
@@ -6565,8 +6571,8 @@ const NATIVE_NON_CLAUDE_MODELS = [
 		displayName: "GPT-5.6 Luna"
 	},
 	{
-		id: "gemini-3.7-flash",
-		displayName: "Gemini 3.7 Flash"
+		id: "gemini-3.8-flash",
+		displayName: "Gemini 3.8 Flash"
 	},
 	{
 		id: "grok-4.6",
@@ -6621,7 +6627,7 @@ const MAX_NATIVE_MODELS = [
 	},
 	{
 		id: MAX_PROFILE_MODELS.gemini,
-		displayName: "Gemini 3.7 Flash"
+		displayName: "Gemini 3.8 Flash"
 	},
 	{
 		id: MAX_PROFILE_MODELS.opus,
@@ -6917,4 +6923,4 @@ function getCodexEnvVars(serverUrl) {
 //#endregion
 export { sharedServerArgs as a, stopKeepAwake as c, runSelfUpdate as d, checkClaudeVersion as f, setupAndServe as i, listModelsForEndpoint as l, getCodexEnvVars as n, LAUNCH_SECRET_HEADER as o, updateClaude as p, parseSharedArgs as r, startKeepAwake as s, getClaudeCodeEnvVars as t, enableFileLogging as u };
 
-//# sourceMappingURL=server-setup-Zn5wkosb.js.map
+//# sourceMappingURL=server-setup-CSkdFAPP.js.map