@code-yeongyu/senpi-ai 2026.10.1 → 2026.10.3

package/dist/providers/data/zai-coding-cn.json

@@ -1 +1 @@
-{"openai-completions":{"chat:glm-4.6v":{"id":"glm-4.6v","name":"GLM-4.6V","api":"openai-completions","provider":"zai-coding-cn","baseUrl":"https://open.bigmodel.cn/api/coding/paas/v4","reasoning":true,"input":["text","image"],"cost":{"input":0.3,"output":0.9,"cacheRead":0,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":128000,"maxTokens":32768,"inputLimits":{"images":{"resize":{"maxWidth":2000,"maxHeight":2000,"maxBytes":4718592,"jpegQuality":80}}},"type":"chat"},"chat:glm-5.3":{"id":"glm-5.3","name":"GLM-5.3","api":"openai-completions","provider":"zai-coding-cn","baseUrl":"https://open.bigmodel.cn/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":null,"minimal":null,"low":"low","medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text"],"cost":{"input":1.4,"output":4.4,"cacheRead":0.26,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"type":"chat"},"chat:glm-5.3-flash":{"id":"glm-5.3-flash","name":"GLM-5.3-Flash","api":"openai-completions","provider":"zai-coding-cn","baseUrl":"https://open.bigmodel.cn/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":null,"minimal":null,"low":"low","medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text","image"],"cost":{"input":0.15,"output":0.5,"cacheRead":0.03,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"inputLimits":{"images":{"resize":{"maxWidth":2000,"maxHeight":2000,"maxBytes":4718592,"jpegQuality":80}}},"type":"chat"},"chat:glm-5.3-highspeed":{"id":"glm-5.3-highspeed","name":"GLM-5.3 Highspeed","api":"openai-completions","provider":"zai-coding-cn","baseUrl":"https://open.bigmodel.cn/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":null,"minimal":null,"low":"low","medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"type":"chat"}}}
+{"openai-completions":{"chat:glm-4.6v":{"id":"glm-4.6v","name":"GLM-4.6V","api":"openai-completions","provider":"zai-coding-cn","baseUrl":"https://open.bigmodel.cn/api/coding/paas/v4","reasoning":true,"input":["text","image"],"cost":{"input":0.3,"output":0.9,"cacheRead":0,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":128000,"maxTokens":32768,"thinkingLevelMap":{"minimal":null,"low":null,"medium":null,"xhigh":null,"max":null},"inputLimits":{"images":{"resize":{"maxWidth":2000,"maxHeight":2000,"maxBytes":4718592,"jpegQuality":80}}},"type":"chat"},"chat:glm-5.3":{"id":"glm-5.3","name":"GLM-5.3","api":"openai-completions","provider":"zai-coding-cn","baseUrl":"https://open.bigmodel.cn/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":null,"minimal":null,"low":"low","medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text"],"cost":{"input":1.4,"output":4.4,"cacheRead":0.26,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"type":"chat"},"chat:glm-5.3-flash":{"id":"glm-5.3-flash","name":"GLM-5.3-Flash","api":"openai-completions","provider":"zai-coding-cn","baseUrl":"https://open.bigmodel.cn/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":null,"minimal":null,"low":"low","medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text","image"],"cost":{"input":0.15,"output":0.5,"cacheRead":0.03,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"inputLimits":{"images":{"resize":{"maxWidth":2000,"maxHeight":2000,"maxBytes":4718592,"jpegQuality":80}}},"type":"chat"},"chat:glm-5.3-highspeed":{"id":"glm-5.3-highspeed","name":"GLM-5.3 Highspeed","api":"openai-completions","provider":"zai-coding-cn","baseUrl":"https://open.bigmodel.cn/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":null,"minimal":null,"low":"low","medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"type":"chat"}}}

package/dist/providers/data/zai.json

@@ -1 +1 @@
-{"openai-completions":{"chat:glm-4.7":{"id":"glm-4.7","name":"GLM-4.7","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"input":["text"],"cost":{"input":0.6,"output":2.2,"cacheRead":0.11,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":204800,"maxTokens":131072,"type":"chat"},"chat:glm-5-turbo":{"id":"glm-5-turbo","name":"GLM-5-Turbo","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"input":["text"],"cost":{"input":1.2,"output":4,"cacheRead":0.24,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":200000,"maxTokens":131072,"type":"chat"},"chat:glm-5.2":{"id":"glm-5.2","name":"GLM-5.2","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":"none","minimal":null,"low":null,"medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text"],"cost":{"input":1.4,"output":4.4,"cacheRead":0.26,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"type":"chat"},"chat:glm-5.2-highspeed":{"id":"glm-5.2-highspeed","name":"GLM-5.2 Highspeed","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":"none","minimal":null,"low":null,"medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"type":"chat"},"chat:glm-5.3":{"id":"glm-5.3","name":"GLM-5.3","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":null,"minimal":null,"low":"low","medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text"],"cost":{"input":1.4,"output":4.4,"cacheRead":0.26,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"type":"chat"},"chat:glm-5.3-flash":{"id":"glm-5.3-flash","name":"GLM-5.3-Flash","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":null,"minimal":null,"low":"low","medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text","image"],"cost":{"input":0.15,"output":0.5,"cacheRead":0.03,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"inputLimits":{"images":{"resize":{"maxWidth":2000,"maxHeight":2000,"maxBytes":4718592,"jpegQuality":80}}},"type":"chat"},"chat:glm-5.3-highspeed":{"id":"glm-5.3-highspeed","name":"GLM-5.3 Highspeed","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":null,"minimal":null,"low":"low","medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"type":"chat"}}}
+{"openai-completions":{"chat:glm-4.7":{"id":"glm-4.7","name":"GLM-4.7","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"input":["text"],"cost":{"input":0.6,"output":2.2,"cacheRead":0.11,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":204800,"maxTokens":131072,"thinkingLevelMap":{"minimal":null,"low":null,"medium":null,"xhigh":null,"max":null},"type":"chat"},"chat:glm-5-turbo":{"id":"glm-5-turbo","name":"GLM-5-Turbo","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"input":["text"],"cost":{"input":1.2,"output":4,"cacheRead":0.24,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":200000,"maxTokens":131072,"thinkingLevelMap":{"minimal":null,"low":null,"medium":null,"xhigh":null,"max":null},"type":"chat"},"chat:glm-5.2":{"id":"glm-5.2","name":"GLM-5.2","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":"none","minimal":null,"low":null,"medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text"],"cost":{"input":1.4,"output":4.4,"cacheRead":0.26,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"type":"chat"},"chat:glm-5.2-highspeed":{"id":"glm-5.2-highspeed","name":"GLM-5.2 Highspeed","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":"none","minimal":null,"low":null,"medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"type":"chat"},"chat:glm-5.3":{"id":"glm-5.3","name":"GLM-5.3","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":null,"minimal":null,"low":"low","medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text"],"cost":{"input":1.4,"output":4.4,"cacheRead":0.26,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"type":"chat"},"chat:glm-5.3-flash":{"id":"glm-5.3-flash","name":"GLM-5.3-Flash","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":null,"minimal":null,"low":"low","medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text","image"],"cost":{"input":0.15,"output":0.5,"cacheRead":0.03,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"inputLimits":{"images":{"resize":{"maxWidth":2000,"maxHeight":2000,"maxBytes":4718592,"jpegQuality":80}}},"type":"chat"},"chat:glm-5.3-highspeed":{"id":"glm-5.3-highspeed","name":"GLM-5.3 Highspeed","api":"openai-completions","provider":"zai","baseUrl":"https://api.z.ai/api/coding/paas/v4","reasoning":true,"thinkingLevelMap":{"off":null,"minimal":null,"low":"low","medium":null,"high":"high","xhigh":null,"max":"max"},"input":["text"],"cost":{"input":0,"output":0,"cacheRead":0,"cacheWrite":0},"compat":{"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":true,"maxTokensField":"max_tokens","thinkingFormat":"zai","supportsStrictMode":true,"zaiToolStream":true},"contextWindow":1000000,"maxTokens":131072,"type":"chat"}}}

package/dist/providers/opengateway-catalog.d.ts

@@ -0,0 +1,43 @@
+import type { ModelCost, ModelCostTier } from "../types.ts";
+export declare const OPENGATEWAY_BASE_URL = "https://apis.opengateway.ai/v1";
+export declare const OPENGATEWAY_MODELS_URL = "https://apis.opengateway.ai/v1/models";
+export declare const OPENGATEWAY_PRICES_URL = "https://opengateway.ai/api/model-prices";
+export interface OpenGatewayListedModel {
+    id: string;
+    status?: string;
+    inputModalities: readonly string[];
+    endpoints: readonly string[];
+    /** Provider route ids in the gateway's preference order. */
+    routes: readonly string[];
+    contextWindow?: number;
+    maxOutputTokens?: number;
+}
+/** Published per-million-token prices; a field is absent when the gateway does not publish it. */
+export interface OpenGatewayPrice {
+    input?: number;
+    output?: number;
+    cacheRead?: number;
+    cacheWrite?: number;
+    tiers: ModelCostTier[];
+}
+interface PriceRoute {
+    route: string;
+    modelId: string;
+    price: OpenGatewayPrice;
+}
+export type OpenGatewayPriceTable = readonly PriceRoute[];
+export declare function parseOpenGatewayListing(value: unknown): OpenGatewayListedModel[];
+/** Chat-completions models that can still be called. Retired models are listed but rejected. */
+export declare function isServableChatModel(model: OpenGatewayListedModel): boolean;
+/** `z-ai/glm-5.3-ultrafast` -> its base `z-ai/glm-5.3` plus the tier label, or undefined. */
+export declare function servingTierBase(id: string): {
+    baseId: string;
+    label: string;
+} | undefined;
+export declare function parseOpenGatewayPriceTable(value: unknown): OpenGatewayPriceTable;
+/** The price billed for a listed model: its preferred route's price, else any route's. */
+export declare function openGatewayPrice(model: OpenGatewayListedModel, table: OpenGatewayPriceTable): OpenGatewayPrice | undefined;
+/** Overlay published gateway prices on a fallback cost; unpublished fields keep the fallback. */
+export declare function withGatewayPrice(fallback: ModelCost, price: OpenGatewayPrice | undefined): ModelCost;
+export {};
+//# sourceMappingURL=opengateway-catalog.d.ts.map

package/dist/providers/opengateway-catalog.js

@@ -0,0 +1,151 @@
+// OpenGateway's public catalog data, shared by the build-time generator
+// (scripts/generate-models-opengateway.ts) and the runtime refresh
+// (opengateway-refresh.ts). Both endpoints are public and need no API key.
+//
+// - GET https://apis.opengateway.ai/v1/models lists every model the gateway
+//   serves: lifecycle status, modalities, endpoints, the provider routes in
+//   preference order, and (for most models) context window and max output.
+// - GET https://opengateway.ai/api/model-prices is the price table the
+//   gateway bills from, keyed per provider route, with the effective
+//   (discounted) per-token price and long-context tiers.
+export const OPENGATEWAY_BASE_URL = "https://apis.opengateway.ai/v1";
+export const OPENGATEWAY_MODELS_URL = `${OPENGATEWAY_BASE_URL}/models`;
+export const OPENGATEWAY_PRICES_URL = "https://opengateway.ai/api/model-prices";
+const SERVING_TIER_SUFFIXES = [
+    { suffix: "-ultrafast", label: "Ultrafast" },
+];
+function isRecord(value) {
+    return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+function positiveInteger(value) {
+    return typeof value === "number" && Number.isInteger(value) && value > 0 ? value : undefined;
+}
+function stringList(value) {
+    return Array.isArray(value) ? value.filter((entry) => typeof entry === "string") : [];
+}
+export function parseOpenGatewayListing(value) {
+    if (!isRecord(value) || !Array.isArray(value.data)) {
+        throw new Error("OpenGateway model listing is not a { data: [...] } list");
+    }
+    const models = [];
+    for (const entry of value.data) {
+        if (!isRecord(entry) || typeof entry.id !== "string" || !entry.id.includes("/"))
+            continue;
+        const modalities = isRecord(entry.modalities) ? entry.modalities : {};
+        const routes = Array.isArray(entry.providers)
+            ? entry.providers.flatMap((route) => (isRecord(route) && typeof route.id === "string" ? [route.id] : []))
+            : [];
+        models.push({
+            id: entry.id,
+            status: typeof entry.status === "string" ? entry.status : undefined,
+            inputModalities: stringList(modalities.input),
+            endpoints: stringList(entry.endpoints),
+            routes,
+            contextWindow: positiveInteger(entry.context_window),
+            maxOutputTokens: positiveInteger(entry.max_output_tokens),
+        });
+    }
+    if (models.length === 0)
+        throw new Error("OpenGateway model listing has no models");
+    return models;
+}
+/** Chat-completions models that can still be called. Retired models are listed but rejected. */
+export function isServableChatModel(model) {
+    return model.endpoints.includes("chat_completions") && model.status !== "retired";
+}
+/** `z-ai/glm-5.3-ultrafast` -> its base `z-ai/glm-5.3` plus the tier label, or undefined. */
+export function servingTierBase(id) {
+    for (const { suffix, label } of SERVING_TIER_SUFFIXES) {
+        if (id.endsWith(suffix) && id.length > suffix.length)
+            return { baseId: id.slice(0, -suffix.length), label };
+    }
+    return undefined;
+}
+function perMillion(value) {
+    if (typeof value !== "number" || !Number.isFinite(value) || value < 0)
+        return undefined;
+    // Per-token decimals (2e-7) scale to float noise (0.19999999999999998); 12 significant digits remove it.
+    return Number((value * 1_000_000).toPrecision(12));
+}
+function parseTiers(axisPrices) {
+    if (!isRecord(axisPrices))
+        return [];
+    const tiers = [];
+    for (const [key, value] of Object.entries(axisPrices)) {
+        // Keys look like "AxisKey(threshold=272000, tier=null)"; only plain context thresholds are input tiers.
+        const match = /^AxisKey\(threshold=(\d+), tier=null\)$/.exec(key);
+        if (!match || !isRecord(value) || value.isEmpty === true)
+            continue;
+        const input = perMillion(value.input);
+        const output = perMillion(value.output);
+        if (input === undefined || output === undefined)
+            continue;
+        tiers.push({
+            inputTokensAbove: Number(match[1]),
+            input,
+            output,
+            cacheRead: perMillion(value.cacheRead) ?? 0,
+            cacheWrite: perMillion(value.cacheCreation) ?? 0,
+        });
+    }
+    return tiers.sort((left, right) => left.inputTokensAbove - right.inputTokensAbove);
+}
+export function parseOpenGatewayPriceTable(value) {
+    if (!isRecord(value))
+        throw new Error("OpenGateway price table is not an object");
+    const routes = [];
+    for (const entry of Object.values(value)) {
+        if (!isRecord(entry) || typeof entry.provider !== "string")
+            continue;
+        if (typeof entry.modelOwner !== "string" || typeof entry.modelName !== "string")
+            continue;
+        const pricing = isRecord(entry.pricing) && isRecord(entry.pricing.current) ? entry.pricing.current : undefined;
+        // The effective price already carries the gateway's discount; the list price is the fallback.
+        const effective = pricing && isRecord(pricing.effectivePrice) ? pricing.effectivePrice : entry;
+        routes.push({
+            route: entry.provider,
+            modelId: `${entry.modelOwner}/${entry.modelName}`,
+            price: {
+                input: perMillion(effective.inputCostPerToken),
+                output: perMillion(effective.outputCostPerToken),
+                cacheRead: perMillion(effective.cacheReadInputTokenCost),
+                cacheWrite: perMillion(effective.cacheCreationInputTokenCost),
+                tiers: parseTiers(effective.axisPrices),
+            },
+        });
+    }
+    if (routes.length === 0)
+        throw new Error("OpenGateway price table has no priced routes");
+    return routes;
+}
+/** The price billed for a listed model: its preferred route's price, else any route's. */
+export function openGatewayPrice(model, table) {
+    const candidates = table.filter((entry) => entry.modelId === model.id);
+    for (const route of model.routes) {
+        const preferred = candidates.find((entry) => entry.route === route);
+        if (preferred)
+            return preferred.price;
+    }
+    return candidates[0]?.price;
+}
+/** Overlay published gateway prices on a fallback cost; unpublished fields keep the fallback. */
+export function withGatewayPrice(fallback, price) {
+    if (!price)
+        return fallback;
+    const cost = {
+        input: price.input ?? fallback.input,
+        output: price.output ?? fallback.output,
+        cacheRead: price.cacheRead ?? fallback.cacheRead,
+        cacheWrite: price.cacheWrite ?? fallback.cacheWrite,
+    };
+    // A published route price owns its tiers: no gateway tiers means flat billing.
+    if (price.input !== undefined && price.output !== undefined) {
+        if (price.tiers.length > 0)
+            cost.tiers = price.tiers;
+    }
+    else if (fallback.tiers) {
+        cost.tiers = fallback.tiers;
+    }
+    return cost;
+}
+//# sourceMappingURL=opengateway-catalog.js.map

package/dist/providers/opengateway-refresh.d.ts

@@ -0,0 +1,20 @@
+import type { RefreshModelsContext } from "../models.ts";
+import type { Model } from "../types.ts";
+import { type OpenGatewayListedModel, type OpenGatewayPriceTable } from "./opengateway-catalog.ts";
+type OpenGatewayModel = Model<"openai-completions">;
+export declare const OPENGATEWAY_REFRESH_INTERVAL_MS: number;
+/** Servable gateway models the shipped catalog lacks; only these need the price table. */
+export declare function unshippedServableModels(shipped: readonly OpenGatewayModel[], listing: readonly OpenGatewayListedModel[]): OpenGatewayListedModel[];
+export declare function overlayOpenGatewayCatalog(shipped: readonly OpenGatewayModel[], listing: readonly OpenGatewayListedModel[], prices: OpenGatewayPriceTable): OpenGatewayModel[];
+/**
+ * Catalog state for the provider: `getModels()` is synchronous, `refresh()` restores the
+ * persisted list and revalidates it against the gateway at most hourly. A persisted list
+ * older than the shipped catalog is ignored, so an upgrade never resurrects metadata the
+ * new release corrected.
+ */
+export declare function createOpenGatewayCatalog(shipped: readonly OpenGatewayModel[], shippedGeneratedAt: number | undefined): {
+    getModels: () => readonly OpenGatewayModel[];
+    refresh: (context: RefreshModelsContext) => Promise<void>;
+};
+export {};
+//# sourceMappingURL=opengateway-refresh.d.ts.map

package/dist/providers/opengateway-refresh.js

@@ -0,0 +1,129 @@
+// Runtime refresh for the OpenGateway provider. A successful gateway listing is
+// authoritative for availability; the shipped catalog (generated from the gateway
+// plus models.dev) supplies metadata and is the offline fallback:
+//
+// - a servable model the shipped catalog lacks is added, built from its shipped
+//   serving-tier base when there is one, priced from the gateway price table;
+// - a shipped model the gateway retired, or no longer lists, is removed;
+// - shipped rows keep their generated metadata (input caps, thinking maps,
+//   prices). Correcting those is the scheduled catalog regeneration's job.
+//
+// Any fetch or parse failure keeps the last good list and surfaces as a refresh error.
+import { isModelType } from "../utils/model-operations.js";
+import { applyOpenAiInputCap } from "../utils/openai-input-cap.js";
+import { isServableChatModel, OPENGATEWAY_BASE_URL, OPENGATEWAY_MODELS_URL, OPENGATEWAY_PRICES_URL, openGatewayPrice, parseOpenGatewayListing, parseOpenGatewayPriceTable, servingTierBase, withGatewayPrice, } from "./opengateway-catalog.js";
+export const OPENGATEWAY_REFRESH_INTERVAL_MS = 60 * 60 * 1000;
+const REQUEST_TIMEOUT_MS = 15_000;
+/** Max output for an added model the gateway publishes no limit for and that has no shipped base. */
+const UNPUBLISHED_MAX_OUTPUT_TOKENS = 32768;
+const ZERO_COST = { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 };
+function addedModel(item, shippedById, prices) {
+    const price = openGatewayPrice(item, prices);
+    // An unpriced model would bill as free in usage accounting; wait for the regeneration instead.
+    if (price?.input === undefined || price.output === undefined)
+        return undefined;
+    const tier = servingTierBase(item.id);
+    const base = tier ? shippedById.get(tier.baseId) : undefined;
+    const contextWindow = item.contextWindow ?? base?.contextWindow;
+    if (contextWindow === undefined)
+        return undefined;
+    const template = base
+        ? { ...base, name: `${base.name} ${tier?.label}` }
+        : {
+            id: item.id,
+            name: item.id,
+            api: "openai-completions",
+            provider: "opengateway",
+            baseUrl: OPENGATEWAY_BASE_URL,
+            compat: { supportsDeveloperRole: false },
+            reasoning: false,
+            input: ["text"],
+            cost: ZERO_COST,
+            contextWindow,
+            maxTokens: UNPUBLISHED_MAX_OUTPUT_TOKENS,
+        };
+    const model = {
+        ...template,
+        id: item.id,
+        input: item.inputModalities.includes("image") ? ["text", "image"] : ["text"],
+        cost: withGatewayPrice(ZERO_COST, price),
+        contextWindow,
+        maxTokens: Math.min(item.maxOutputTokens ?? template.maxTokens, contextWindow),
+    };
+    applyOpenAiInputCap(model);
+    return model;
+}
+/** Servable gateway models the shipped catalog lacks; only these need the price table. */
+export function unshippedServableModels(shipped, listing) {
+    const shippedIds = new Set(shipped.map((model) => model.id));
+    return listing.filter((item) => isServableChatModel(item) && !shippedIds.has(item.id));
+}
+export function overlayOpenGatewayCatalog(shipped, listing, prices) {
+    const servable = new Set(listing.filter(isServableChatModel).map((item) => item.id));
+    const shippedById = new Map(shipped.map((model) => [model.id, model]));
+    const added = unshippedServableModels(shipped, listing).flatMap((item) => addedModel(item, shippedById, prices) ?? []);
+    return [...shipped.filter((model) => servable.has(model.id)), ...added];
+}
+async function fetchJson(url, signal) {
+    const response = await fetch(url, { headers: { accept: "application/json" }, signal });
+    if (!response.ok)
+        throw new Error(`OpenGateway catalog request failed: ${url} returned ${response.status}`);
+    return response.json();
+}
+async function fetchRefreshedCatalog(shipped, signal) {
+    const listing = parseOpenGatewayListing(await fetchJson(OPENGATEWAY_MODELS_URL, signal));
+    const prices = unshippedServableModels(shipped, listing).length > 0
+        ? parseOpenGatewayPriceTable(await fetchJson(OPENGATEWAY_PRICES_URL, signal))
+        : [];
+    const refreshed = overlayOpenGatewayCatalog(shipped, listing, prices);
+    if (refreshed.length === 0)
+        throw new Error("OpenGateway listing has no servable chat models; keeping the last good catalog");
+    return refreshed;
+}
+function isOpenGatewayChatModel(model) {
+    return model.provider === "opengateway" && isModelType(model, "chat") && model.api === "openai-completions";
+}
+/**
+ * Catalog state for the provider: `getModels()` is synchronous, `refresh()` restores the
+ * persisted list and revalidates it against the gateway at most hourly. A persisted list
+ * older than the shipped catalog is ignored, so an upgrade never resurrects metadata the
+ * new release corrected.
+ */
+export function createOpenGatewayCatalog(shipped, shippedGeneratedAt) {
+    let current = shipped;
+    return {
+        getModels: () => current,
+        refresh: async (context) => {
+            const stored = context.stored;
+            const storedCheckedAt = stored?.checkedAt;
+            const usable = stored !== undefined &&
+                storedCheckedAt !== undefined &&
+                (shippedGeneratedAt === undefined || storedCheckedAt > shippedGeneratedAt);
+            if (usable && stored) {
+                const restored = stored.models.filter(isOpenGatewayChatModel);
+                if (!(await context.publish({
+                    update: () => {
+                        current = restored;
+                    },
+                })))
+                    return;
+            }
+            if (!context.allowNetwork || context.signal.aborted)
+                return;
+            const age = Date.now() - (storedCheckedAt ?? 0);
+            if (!context.force && usable && age >= 0 && age < OPENGATEWAY_REFRESH_INTERVAL_MS)
+                return;
+            const signal = AbortSignal.any([context.signal, AbortSignal.timeout(REQUEST_TIMEOUT_MS)]);
+            const refreshed = await fetchRefreshedCatalog(shipped, signal);
+            if (context.signal.aborted)
+                return;
+            await context.publish({
+                persist: { models: refreshed, checkedAt: Date.now() },
+                update: () => {
+                    current = refreshed;
+                },
+            });
+        },
+    };
+}
+//# sourceMappingURL=opengateway-refresh.js.map

package/dist/providers/opengateway.js

@@ -1,15 +1,26 @@
 import { openAICompletionsApi } from "../api/openai-completions.lazy.js";
 import { envApiKeyAuth } from "../auth/helpers.js";
 import { createProvider } from "../models.js";
+import modelDataManifest from "./data/.manifest.json" with { type: "json" };
 import { OPENGATEWAY_MODELS } from "./opengateway.models.js";
+import { createOpenGatewayCatalog } from "./opengateway-refresh.js";
 export function opengatewayProvider() {
-    return createProvider({
+    const shipped = Object.values(OPENGATEWAY_MODELS);
+    const generatedAt = Date.parse(modelDataManifest.generatedAt);
+    const catalog = createOpenGatewayCatalog(shipped, Number.isNaN(generatedAt) ? undefined : generatedAt);
+    const provider = createProvider({
         id: "opengateway",
         name: "OpenGateway",
         baseUrl: "https://apis.opengateway.ai/v1",
         auth: { apiKey: envApiKeyAuth("OpenGateway API key", ["OPENGATEWAY_API_KEY"]) },
-        models: Object.values(OPENGATEWAY_MODELS),
+        models: shipped,
         api: openAICompletionsApi(),
     });
+    return {
+        ...provider,
+        getModels: catalog.getModels,
+        getAllModels: catalog.getModels,
+        refreshModels: catalog.refresh,
+    };
 }
 //# sourceMappingURL=opengateway.js.map

package/dist/utils/oauth-page.js

@@ -1,4 +1,4 @@
-const LOGO_SVG = `<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 800 800" aria-hidden="true"><path fill="#fff" fill-rule="evenodd" d="M165.29 165.29 H517.36 V400 H400 V517.36 H282.65 V634.72 H165.29 Z M282.65 282.65 V400 H400 V282.65 Z"/><path fill="#fff" d="M517.36 400 H634.72 V634.72 H517.36 Z"/></svg>`;
+const LOGO_SVG = `<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 800 800" aria-hidden="true"><path fill="#F09082" d="M165.29 165.29H517.36V400H400V282.65H165.29Z"/><path fill="#4D9ABF" d="M165.29 282.65H282.65V400H400V517.36H282.65V634.72H165.29Z"/><path fill="#F1BE58" d="M517.36 400H634.72V634.72H517.36Z"/></svg>`;
 function escapeHtml(value) {
     return value
         .replaceAll("&", "&amp;")

package/dist/utils/openai-input-cap.d.ts

@@ -0,0 +1,4 @@
+import type { Api, Model } from "../types.ts";
+/** OpenAI's input/output split follows GPT-5/6 models across gateway providers. */
+export declare function applyOpenAiInputCap(model: Model<Api>): void;
+//# sourceMappingURL=openai-input-cap.d.ts.map

package/dist/utils/openai-input-cap.js

@@ -0,0 +1,18 @@
+const DOCUMENTED_INPUT_CAPS = new Map([
+    [400000, 272000],
+    [1050000, 922000],
+]);
+const GATEWAY_ID_PREFIX = /^(?:[a-z]{2}\.)?(?:global\.)?openai[./]/;
+/** OpenAI's input/output split follows GPT-5/6 models across gateway providers. */
+export function applyOpenAiInputCap(model) {
+    const bare = model.id.replace(GATEWAY_ID_PREFIX, "");
+    if (!/^gpt-(?:5|6)(?:[.-]|$)/.test(bare))
+        return;
+    // Historical models.dev output metadata duplicated GPT-5 Pro's input sub-limit.
+    if (bare === "gpt-5-pro" && model.maxTokens === 272000)
+        model.maxTokens = 128000;
+    if (model.maxTokens === 128000) {
+        model.contextWindow = DOCUMENTED_INPUT_CAPS.get(model.contextWindow) ?? model.contextWindow;
+    }
+}
+//# sourceMappingURL=openai-input-cap.js.map

package/dist/utils/overflow.js

@@ -29,7 +29,7 @@
  * - kiro-lb gateways: "Request payload is 1095225 bytes, over the 1085435 byte limit Kiro accepts." / "Request payload is N tokens, over the M token limit Kiro accepts." (HTTP 400 local payload guard)
  * - Kiro upstream via kiro-lb: "Model context limit reached. Conversation size exceeds model capacity." (CONTENT_LENGTH_EXCEEDS_THRESHOLD token overflow)
  * - Mistral: "Prompt contains X tokens ... too large for model with Y maximum context length"
- * - z.ai: `{"code":"1261","message":"Prompt too long"}` or silent overflow via usage.input > contextWindow
+ * - z.ai: `{"code":"1261","message":"Prompt too long"}`, `{"code":"1261","message":"Prompt exceeds max length"}` (CN endpoint), or silent overflow via usage.input > contextWindow
  * - Xiaomi MiMo: Truncates input to fill contextWindow exactly, then returns finish_reason "length"
  *   with output=0 (no room left to generate). Detected via stopReason "length" + zero output +
  *   input filling the context window.
@@ -42,6 +42,7 @@ const OVERFLOW_PATTERNS = [
     /^Context window exhausted: /, // pi-ai pre-flight guard: no answer room left, provider never called
     /^The conversation is too long to resend \(about \d+ tokens, limit \d+\)/, // anthropic-subscription cold-seed budget: re-send refused before dispatch
     /prompt (?:is )?too long/i, // Anthropic and z.ai token overflow
+    /prompt exceeds max length/i, // z.ai CN endpoint token overflow
     /request_too_large/i, // Anthropic request byte-size overflow (HTTP 413)
     /input is too long for requested model/i, // Amazon Bedrock
     /exceeds (?:(?:the|this) )?(?:model'?s )?context window/i, // OpenAI (Completions & Responses API)

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@code-yeongyu/senpi-ai",
-	"version": "2026.10.1",
+	"version": "2026.10.3",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"type": "module",
 	"main": "./dist/index.js",
@@ -15,6 +15,10 @@
 			"types": "./dist/index.d.ts",
 			"import": "./dist/index.js"
 		},
+		"./models": {
+			"types": "./dist/models.d.ts",
+			"import": "./dist/models.js"
+		},
 		"./compat": {
 			"types": "./dist/compat.d.ts",
 			"import": "./dist/compat.js"
@@ -84,7 +88,7 @@
 	"dependencies": {
 		"@anthropic-ai/sdk": "0.127.0",
 		"@aws-sdk/client-bedrock-runtime": "3.1136.0",
-		"@earendil-works/pi-telemetry": "npm:@code-yeongyu/senpi-telemetry@2026.10.1",
+		"@earendil-works/pi-telemetry": "npm:@code-yeongyu/senpi-telemetry@2026.10.3",
 		"@google/genai": "2.23.0",
 		"@smithy/node-http-handler": "4.12.1",
 		"http-proxy-agent": "9.1.0",