@wenathlan/getry 2.0.165 → 2.0.167

package/gateway.ts

@@ -0,0 +1,1306 @@
+/**
+ * gateway — consolidated gateway logics of the getry family
+ * @module gateway
+ * @family wave/logicfold3
+ * One keyed file per family: the former variant arms (gatewayconfig, gatewayviewconfig)
+ * retired into this seat with every export preserved under the same name.
+ */
+
+import type { gatewaydefinition, modeldef, thinkinglevel } from "./types.ts";
+import { listlevels, providerchain, type thinkingproviderrow } from "@wenathlan/getry/thinking";
+
+/**
+ * gateway.ts — the gateway versions domain of getry (root layer).
+ *
+ * the five provider gateways the app serves (v1 through v5), the seven
+ * OpenAI-compatible route kinds each one answers, and the canonical
+ * parser pattern the streaming routes follow. the rows below are the
+ * in-memory catalog of the route map: the self-hosted gateway answers
+ * the real traffic over HTTPS, the static build falls back to this
+ * catalog without ever touching the visitor machine.
+ */
+
+/** the seven OpenAI-compatible route kinds every gateway version carries. */
+export const GATEWAYROUTES = [
+  "chat/completions",
+  "completions",
+  "embeddings",
+  "keys",
+  "messages",
+  "models",
+  "responses",
+] as const;
+
+/** one route kind of the gateway surface. */
+export type GatewayRoute = (typeof GATEWAYROUTES)[number];
+
+/** the lifecycle badge a gateway version carries. */
+export type GatewayBadge = "live" | "paused";
+
+/** one provider gateway of the family (v1 through v5). */
+export type GatewayVersion = {
+  v: string;
+  provider: string;
+  pattern: string;
+  badge: GatewayBadge;
+  models: string;
+  routes: GatewayRoute[];
+  note: string;
+};
+
+/** the in-memory catalog the DB layer persists on first run. */
+export const GATEWAYVERSIONS: GatewayVersion[] = [
+  {
+    v: "v1",
+    provider: "zai",
+    pattern: "passthrough retransmit via z ai web dev sdk",
+    badge: "live",
+    models:
+      "glm-5.3 glm-5.3-flash glm-5.3-fast glm-5.3-air glm-5.2 glm-5.1 glm-5 glm-5v glm-5-turbo glm-4-plus glm-4-flash devthink",
+    routes: [...GATEWAYROUTES],
+    note: "no token chat id required — sdk has internal config — individual model calling plus devthink meta over glm-5.3",
+  },
+  {
+    v: "v2",
+    provider: "babel town",
+    pattern: "rebuild plus byte tee passthrough paused fallback v1",
+    badge: "paused",
+    models: "babel-glm-5.2 devthink",
+    routes: [...GATEWAYROUTES],
+    note: "service is paused returns 503 fallback to v1 zai — glm-5.2 only",
+  },
+  {
+    v: "v3",
+    provider: "nvidia nim",
+    pattern: "rebuild per family reasoning control pure byte passthrough",
+    badge: "live",
+    models: "17 chat models kimi-k3 deepseek-v4 nemotron-3 muse-glimmer laguna gpt-oss gemma-4 mistral llama-3.2",
+    routes: [...GATEWAYROUTES],
+    note: "22 keys round robin devthink meta model rotates every 6 messages",
+  },
+  {
+    v: "v4",
+    provider: "opencode zen plus kilo",
+    pattern: "dynamic free discovery anonymous no key",
+    badge: "live",
+    models: "discovered live -free and :free tagged models from opencode zen plus kilo",
+    routes: [...GATEWAYROUTES],
+    note: "zero hardcoded model names — universal free-tag filter plus devthink context-window math",
+  },
+  {
+    v: "v5",
+    provider: "openrouter",
+    pattern: "dynamic free discovery openrouter :free tagged",
+    badge: "live",
+    models: "discovered live :free tagged models from openrouter",
+    routes: [...GATEWAYROUTES],
+    note: "requires free openrouter api key — zero hardcoded model names plus devthink context-window math",
+  },
+];
+
+/** the canonical parser pattern of the streaming routes, in order. */
+export const PARSERSTEPS: string[] = [
+  "readablestream with safeenqueue safeclose plus closed flag",
+  "setinterval keepalive 200ms only when silent greater than 1 second",
+  "upstream call fetch with abortsignal timeout 2147483647 for stream",
+  "parse sse parsesseframes split on double newline never partial frame",
+  "discard heartbeat lines starting with colon or event ping",
+  "mask model name to devthink via regex no json parse",
+  "re-emit normalized chunks delta reasoning content before content",
+  "finally emit finish reason stop plus single done plus savemsg",
+];
+
+/**
+ * lists the gateway versions in serve order (v1 first).
+ *
+ * @returns the version rows.
+ */
+export function listversions(): GatewayVersion[] {
+  return GATEWAYVERSIONS;
+}
+
+/**
+ * finds one gateway version by its version tag.
+ *
+ * @param v the version tag (v1 through v5).
+ * @returns the version row or null.
+ */
+export function findversion(v: string): GatewayVersion | null {
+  return GATEWAYVERSIONS.find((version) => version.v === v) ?? null;
+}
+
+/**
+ * counts the routes the whole gateway serves (versions times route kinds).
+ *
+ * @returns the total route count.
+ */
+export function totalroutes(): number {
+  return GATEWAYVERSIONS.reduce((sum, version) => sum + version.routes.length, 0);
+}
+
+/**
+ * lists the versions that answer live traffic right now.
+ *
+ * @returns the live version rows.
+ */
+export function liveversions(): GatewayVersion[] {
+  return GATEWAYVERSIONS.filter((version) => version.badge === "live");
+}
+
+/**
+ * builds the cross-provider thinking chain for one level over the catalog
+ * this module owns: every gateway version's provider row (the
+ * GATEWAYVERSIONS lifecycle badges) wired with the model rows of the
+ * matching gatewayViewConfig version, routed through the thinking seat's
+ * providerchain — the tier routing (the level selects the model tier), the
+ * budget fit and the live-first order all answer from the seat (the one
+ * owner of the routing math), the rows from this catalog (the one owner of
+ * the data). the family consumers (devthink and siblings) call this
+ * through the @wenathlan/getry/gatewayconfig face and never rebuild the
+ * chain themselves.
+ *
+ * @law the level routes the chain — the data rides the catalog rows, the
+ *      math rides the thinking seat, nothing duplicates
+ * @param level the thinking level to route.
+ * @returns the ordered provider/model candidates (live providers first,
+ *          tier-exact models first within each provider).
+ */
+export function thinkingchain(level: thinkinglevel): { provider: string; model: modeldef }[] {
+  const catalog: thinkingproviderrow[] = GATEWAYVERSIONS.map((version) => ({
+    provider: version.provider,
+    badge: version.badge,
+    models: gatewayViewConfig.versions[version.v]?.models ?? [],
+  }));
+  return providerchain(catalog, level).map((candidate) => ({
+    provider: candidate.provider,
+    model: candidate.model as modeldef,
+  }));
+}
+
+/* ===== block: gatewayconfig ===== */
+/**
+ * gatewayconfig.ts — the gateway definition of the devthink app (root layer).
+ *
+ * The embedded gateway serves the OpenAI-compatible contract the chat speaks:
+ * `/v1/chat/completions` + `/v1/models` (and the legacy `/api/v1/*` prefix the
+ * server always carried). The v1 row rides the z.ai SDK transport — keyless,
+ * the SDK authenticates itself — exactly the family doctrine: the chat talks to
+ * the gateway, the gateway talks to the model, and no side-service ever ships.
+ */
+
+
+export const gatewayConfig: gatewaydefinition = {
+  name: "devthink gateway",
+  description: "the embedded gateway of the devthink app — v1 zai sdk keyless",
+  versions: {
+    v1: {
+      id: "v1",
+      name: "zai sdk",
+      providername: "zai",
+      note: "passthrough retransmit via z ai web dev sdk — no token chat id required",
+      transport: { type: "zai-sdk" },
+      upstreams: [],
+      auth: { mode: "keylesssdk" },
+      defaultmodel: "devthink",
+      models: [
+        { id: "devthink", context: 200000, maxoutput: 32768, reasoning: true },
+        { id: "glm-5.3", context: 200000, maxoutput: 32768, reasoning: true, family: "glm" },
+        { id: "glm-5.3-flash", context: 128000, maxoutput: 16384, reasoning: true, family: "glm", free: true },
+        { id: "glm-5.3-fast", context: 128000, maxoutput: 16384, reasoning: true, family: "glm" },
+        { id: "glm-5.3-air", context: 128000, maxoutput: 16384, reasoning: true, family: "glm", free: true },
+        { id: "glm-5.2", context: 200000, maxoutput: 32768, reasoning: true, family: "glm" },
+        { id: "glm-5.1", context: 128000, maxoutput: 32768, reasoning: true, family: "glm" },
+        { id: "glm-5", context: 128000, maxoutput: 32768, reasoning: true, family: "glm" },
+        { id: "glm-5v", context: 128000, maxoutput: 16384, reasoning: false, vision: true, family: "glm" },
+        { id: "glm-5-turbo", context: 128000, maxoutput: 16384, reasoning: false, family: "glm", free: true },
+        { id: "glm-4-plus", context: 128000, maxoutput: 16384, reasoning: false, family: "glm" },
+        { id: "glm-4-flash", context: 128000, maxoutput: 8192, reasoning: false, family: "glm", free: true },
+      ],
+      defaults: {
+        primary: "devthink",
+        maxtokens: 8192,
+        temperature: [0, 2, 0.7],
+      },
+      thinking: {
+        defaultlevel: "high",
+        param: "thinking",
+        twocalls: { threshold: 98304, freshresponseguarantee: true },
+      },
+      metamodel: {
+        id: "devthink",
+        contextoverride: 200000,
+        maxoutput: 32768,
+        maskupstreammodel: true,
+        alwaysdisplay: true,
+        listedcreated: 1700000000,
+        pattern: "fusion meta model over the zai sdk transport",
+      },
+      context: {
+        restoresessionhistory: false,
+        truncate: true,
+        historylimit: 40,
+        truncatemargin: 4096,
+      },
+    },
+  },
+};
+
+/* (the gatewayconfig arm once answered a dual default export beside this
+ * named seat — the logicfold3 fold converted it to the named export alone,
+ * the default retired with the variant file.) */
+
+/* ===== block: gatewayviewconfig ===== */
+/**
+ * gatewayviewconfig — the user customization layer of the embedded gateway
+ * one file one responsibility — only version definitions live here
+ * this is the hardcoded v1 v5 configuration the gateway ships with
+ *
+ * the catalog lives at the repository root beside the engine modules it
+ * configures (the family law: page folders carry only the page anchor and
+ * the visual components beside it — engine config is root module matter)
+ *
+ * the library engine at root is universal and dry — it reads a config file
+ * every value here is what the user customizes:
+ *   baseurls — which upstream apis to call
+ *   models — which models to expose
+ *   auth — how keys are resolved per method
+ *   rotation — model rotation strategies
+ *   retry — backoff and status handling
+ *   context — session history and truncation
+ *   thinking — two-calls pattern and budgets
+ *
+ * create as many versions as you want — v1 through v9 and beyond
+ * each version gets all 7 routes generated automatically:
+ *   chat/completions completions messages responses embeddings keys models
+ *
+ * the engine config loader (gateway-configloader.ts) probes the standard
+ * sol/config.mjs|ts|js locations of the working directory for the definition
+ * a deployment loads; this catalog is the shipped example the gateway
+ * console page renders — copy it to the location your serve reads, or
+ * use the cli to scaffold a fresh config (gateway init) and edit the values
+ */
+
+/* the gatewaydefinition family moved to the gateway types with the
+ * wave/ownership.getry split: the catalog imports the kind from the engine
+ * owner (types.ts of this app root) — the os gateway lane keeps its own
+ * web-check projection and never leaks into the engine import graph. */
+
+
+/** the live thinking budgets consumed from the getry engine (pr16): the
+ * ladder is owned by getry and every budget map of this catalog derives
+ * from the owner's registered rungs — never a second literal copy, never a
+ * fixed 7. */
+const ladderbudgets = Object.fromEntries(listlevels().map((rung) => [rung.level, rung.budget])) as Record<
+  thinkinglevel,
+  number
+>;
+
+export const gatewayViewConfig: gatewaydefinition = {
+  name: "DevThink Gateway",
+  description: "multi provider ai gateway v1 v5 with thinking rotation retry and fusion context",
+
+  versions: {
+    // -----------------------------------------------------------------------
+    // v1 — zai sdk keyless transport with 2-calls thinking pattern
+    // -----------------------------------------------------------------------
+    v1: {
+      id: "v1",
+      name: "V1 ZAI",
+      providername: "zai",
+      note: "v1 replica retains token plus chat id functionality calls internal api z ai directly",
+      transport: {
+        type: "zai-sdk",
+      },
+      upstreams: [{ name: "zai", baseurl: "https://internal-api.z.ai/v1" }],
+      auth: {
+        mode: "keylesssdk",
+        required: false,
+      },
+      models: [
+        {
+          id: "glm-5.3",
+          context: 1048576,
+          maxoutput: 131072,
+          vision: false,
+          reasoning: true,
+          free: false,
+          rank: { score: 56, tier: 1 },
+        },
+        {
+          id: "glm-5.3-flash",
+          context: 1048576,
+          maxoutput: 131072,
+          vision: true,
+          reasoning: true,
+          free: false,
+          rank: { score: 55, tier: 1 },
+        },
+        {
+          id: "glm-5.2",
+          context: 1048576,
+          maxoutput: 131072,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 53, tier: 1 },
+        },
+        {
+          id: "glm-5.1",
+          context: 131072,
+          maxoutput: 16384,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 40, tier: 2 },
+        },
+        {
+          id: "glm-5v",
+          context: 131072,
+          maxoutput: 16384,
+          vision: true,
+          reasoning: true,
+          free: true,
+          rank: { score: 38, tier: 2 },
+        },
+        {
+          id: "glm-4-plus",
+          context: 131072,
+          maxoutput: 4096,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 25, tier: 3 },
+        },
+        {
+          id: "glm-4-flash",
+          context: 131072,
+          maxoutput: 4096,
+          vision: false,
+          reasoning: false,
+          free: true,
+          rank: { score: 15, tier: 4 },
+        },
+      ],
+      defaultmodel: "glm-5.3",
+      metamodel: {
+        id: "devthink",
+        contextoverride: 2300000,
+        fallbackcontext: 1048576,
+        maxoutput: 98304,
+        maskupstreammodel: false,
+        alwaysdisplay: false,
+        pattern: "2-calls thinking over glm-5.2 backend",
+      },
+      thinking: {
+        defaultlevel: "high",
+        budgets: ladderbudgets,
+        twocalls: {
+          threshold: 98304,
+          freshresponseguarantee: true,
+        },
+        param: "thinking",
+      },
+      bodybuild: {
+        optionalparamspolicy: "always",
+        clampmaxtokens: true,
+      },
+      routes: {
+        embeddings: "disabled",
+        embeddingsmessage: "zai has no embeddings endpoint",
+      },
+      headers: {
+        useragent: "DevThink-Gateway/1.0",
+      },
+    },
+
+    // -----------------------------------------------------------------------
+    // v2 — babel town paused with fallback to v1
+    // -----------------------------------------------------------------------
+    v2: {
+      id: "v2",
+      name: "V2 Babel",
+      providername: "babel",
+      note: "service is paused returns 503 fallback to v1 zai",
+      paused: true,
+      upstreams: [{ name: "babel", baseurl: "https://api.babel.town/v1" }],
+      auth: {
+        mode: "apikeyheader",
+        headername: "x-token",
+        keysources: ["ephemeralurl"],
+        keyurl: "https://glm.babel.town/api/get_api_key",
+        keyexpiryminutes: 60,
+        ipbound: true,
+        required: false,
+      },
+      models: [
+        {
+          id: "babel-glm-5.3",
+          upstream: "glm-5.3",
+          context: 1048576,
+          maxoutput: 131072,
+          free: true,
+        },
+        {
+          id: "babel-glm-5.3-flash",
+          upstream: "glm-5.3-flash",
+          context: 1048576,
+          maxoutput: 131072,
+          free: true,
+        },
+        {
+          id: "babel-glm-5.2",
+          upstream: "glm-5.2",
+          context: 1048576,
+          maxoutput: 131072,
+          free: true,
+        },
+        {
+          id: "babel-devthink",
+          upstream: "devthink",
+          context: 1048576,
+          maxoutput: 131072,
+          free: true,
+        },
+      ],
+      defaultmodel: "babel-glm-5.3",
+      metamodel: {
+        id: "babel-devthink",
+        fallbackcontext: 1048576,
+        maxoutput: 131072,
+        maskupstreammodel: false,
+        alwaysdisplay: false,
+      },
+      routes: {
+        embeddings: "disabled",
+        embeddingsmessage: "babel has no embeddings endpoint",
+        pausedstatus: 503,
+        pausedmessage: "babel town is paused — use v1 zai instead",
+        fallbackroute: "/v1/chat/completions",
+      },
+    },
+
+    // -----------------------------------------------------------------------
+    // v3 — nvidia nim with 4-backend rotation db keys retry backoff
+    // -----------------------------------------------------------------------
+    v3: {
+      id: "v3",
+      name: "V3 NVIDIA",
+      providername: "nvidia",
+      note: "nvidia keys round robin devthink meta model rotates",
+      upstreams: [{ name: "nvidia", baseurl: "https://integrate.api.nvidia.com/v1" }],
+      auth: {
+        mode: "bearer",
+        keysources: ["db", "env"],
+        dbmodel: "nvidiaKey",
+        dbtake: 50,
+        envvar: "NVIDIA_API_KEY",
+        envseparator: ",",
+        keyrotation: "roundrobin",
+        keylabelformat: "nvapi-key-{idx}",
+        required: true,
+        minkeylength: 10,
+      },
+      models: [
+        // rotation backends — the devthink meta rotates across these 4
+        {
+          id: "moonshotai/kimi-k3",
+          context: 262144,
+          maxoutput: 68000,
+          vision: true,
+          reasoning: true,
+          free: true,
+          rank: { score: 60, tier: 1 },
+          family: "kimi",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "deepseek-ai/deepseek-v4-pro-0813",
+          context: 1000000,
+          maxoutput: 68000,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 55, tier: 1 },
+          family: "deepseek",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "meta/muse-glimmer-30b",
+          context: 262144,
+          maxoutput: 68000,
+          vision: true,
+          reasoning: true,
+          free: true,
+          rank: { score: 50, tier: 1 },
+          family: "muse",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "poolside/laguna-xs-2.1",
+          context: 262144,
+          maxoutput: 68000,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 0, tier: 4 },
+          family: "laguna",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        // individual catalog
+        {
+          id: "deepseek-ai/deepseek-v4-flash-0731",
+          context: 1000000,
+          maxoutput: 68000,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 52, tier: 1 },
+          family: "deepseek",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "nvidia/nemotron-3-ultra-550b-a55b",
+          context: 1000000,
+          maxoutput: 68000,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 38, tier: 2 },
+          family: "nemotron",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "nvidia/nemotron-3.5-lightning-30b-a3b",
+          context: 262144,
+          maxoutput: 68000,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 24, tier: 2 },
+          family: "nemotron",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "google/gemma-4-31b-it",
+          context: 262144,
+          maxoutput: 68000,
+          vision: true,
+          reasoning: true,
+          free: true,
+          rank: { score: 30, tier: 2 },
+          family: "gemma",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "nvidia/nemotron-3-super-120b-a12b",
+          context: 1000000,
+          maxoutput: 68000,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 26, tier: 2 },
+          family: "nemotron",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "openai/gpt-oss-120b",
+          context: 131072,
+          maxoutput: 68000,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 24, tier: 2 },
+          family: "gptoss",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "nvidia/nemotron-3-nano-30b-a3b",
+          context: 1000000,
+          maxoutput: 68000,
+          vision: false,
+          reasoning: true,
+          free: true,
+          family: "nemotron",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning",
+          context: 262144,
+          maxoutput: 68000,
+          vision: true,
+          reasoning: true,
+          free: true,
+          family: "nemotron",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "openai/gpt-oss-20b",
+          context: 131072,
+          maxoutput: 68000,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 15, tier: 3 },
+          family: "gptoss",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "mistralai/mistral-nemotron",
+          context: 262144,
+          maxoutput: 68000,
+          vision: false,
+          reasoning: true,
+          free: true,
+          family: "mistral",
+          chattemplatekwargs: { enable_thinking: true },
+        },
+        {
+          id: "meta/llama-3.2-90b-vision-instruct",
+          context: 131072,
+          maxoutput: 68000,
+          vision: true,
+          reasoning: false,
+          free: true,
+          family: "llama",
+          chattemplatekwargs: {},
+        },
+        {
+          id: "meta/llama-3.2-11b-vision-instruct",
+          context: 131072,
+          maxoutput: 68000,
+          vision: true,
+          reasoning: false,
+          free: true,
+          family: "llama",
+          chattemplatekwargs: {},
+        },
+      ],
+      defaultmodel: "devthink",
+      rotation: {
+        mode: "persession",
+        models: [
+          "moonshotai/kimi-k3",
+          "deepseek-ai/deepseek-v4-pro-0813",
+          "meta/muse-glimmer-30b",
+          "poolside/laguna-xs-2.1",
+        ],
+        everynmessages: 6,
+        maxmodelrotations: 4,
+        rotateonstatus: [529, 404, 410],
+        rotateontimeout: true,
+        staggernewsessions: false,
+        sessionstore: "memory",
+      },
+      retry: {
+        maxretries: 3,
+        backoffbasems: 400,
+        backoffcapms: 8000,
+        jitter: 0.2,
+        statuses: [429, 502, 503, 504],
+        nonretryable: [400],
+        fallback: "none",
+      },
+      timeout: {
+        modelswitchintervalms: 60000,
+        requestms: 300000,
+        streamms: 2147483647,
+      },
+      context: {
+        restoresessionhistory: true,
+        historylimit: 1000,
+        truncatemargin: 4096,
+        percallfallback: 1048576,
+        markshared: true,
+      },
+      thinking: {
+        defaultlevel: "high",
+        param: "chattemplatekwargs",
+      },
+      bodybuild: {
+        optionalparamspolicy: "omit-unspecified",
+        clampmaxtokens: true,
+        maxtokenscap: 68000,
+      },
+      metamodel: {
+        id: "devthink",
+        fallbackcontext: 1048576,
+        maxoutput: 272000,
+        maskupstreammodel: true,
+        alwaysdisplay: false,
+        pattern: "rotation over 4 nvidia reasoning models 2-calls pattern",
+      },
+      routes: {
+        embeddings: "disabled",
+        embeddingsmessage: "nvidia has no embeddings endpoint on this gateway",
+      },
+      headers: {
+        useragent: "DevThink-Gateway/3.0",
+      },
+    },
+
+    // -----------------------------------------------------------------------
+    // v4 — opencode zen plus kilo anonymous free models cross-provider fallback
+    // -----------------------------------------------------------------------
+    v4: {
+      id: "v4",
+      name: "V4 OpenCode Kilo",
+      providername: "opencode",
+      note: "anonymous free tier no api key required model masking to devthink",
+      upstreams: [
+        { name: "opencode", baseurl: "https://opencode.ai/zen/v1", fallback: "kilo" },
+        { name: "kilo", baseurl: "https://api.kilo.ai/api/gateway", fallback: "opencode" },
+      ],
+      auth: {
+        mode: "anonymous",
+        required: false,
+      },
+      models: [
+        // opencode zen free models
+        {
+          id: "muse-spark-1.2-contributor-free",
+          context: 131072,
+          maxoutput: 16384,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 57, tier: 1 },
+          endpoint: "responses",
+        },
+        {
+          id: "deepseek-v4-flash-free",
+          context: 1048576,
+          maxoutput: 16384,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 52, tier: 1 },
+        },
+        {
+          id: "mimo-v2.5-free",
+          context: 1048576,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 43, tier: 1 },
+        },
+        {
+          id: "nemotron-3-ultra-free",
+          context: 1048576,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 38, tier: 2 },
+          nvidiarouted: true,
+        },
+        {
+          id: "ling-3.0-flash-fin-free",
+          context: 262144,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 38, tier: 2 },
+        },
+        {
+          id: "nemotron-3.5-lightning-free",
+          context: 1048576,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 24, tier: 2 },
+          nvidiarouted: true,
+        },
+        {
+          id: "laguna-s-2.1-free",
+          context: 262144,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 0, tier: 4 },
+        },
+        // kilo free models
+        {
+          id: "thinkingmachines/inkling:free",
+          context: 1048576,
+          maxoutput: 262144,
+          vision: true,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "tencent/hy3:free",
+          context: 262144,
+          maxoutput: 128000,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "thinkingmachines/inkling-small:free",
+          context: 1048576,
+          maxoutput: 262144,
+          vision: true,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "inclusionai/ling-3.0-flash-fin:free",
+          context: 262144,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "nvidia/nemotron-3-ultra-550b-a55b:free",
+          context: 1048576,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+          nvidiarouted: true,
+        },
+        {
+          id: "minimax/minimax-m3:free",
+          context: 1048576,
+          maxoutput: 16384,
+          vision: true,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "stepfun/step-3.7-flash:free",
+          context: 262144,
+          maxoutput: 262144,
+          vision: true,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "nvidia/nemotron-3.5-lightning:free",
+          context: 1048576,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+          nvidiarouted: true,
+        },
+        {
+          id: "nvidia/nemotron-3-super-120b-a12b:free",
+          context: 1048576,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+          nvidiarouted: true,
+        },
+        {
+          id: "cohere/north-mini-code:free",
+          context: 256000,
+          maxoutput: 64000,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free",
+          context: 256000,
+          maxoutput: 65536,
+          vision: true,
+          reasoning: true,
+          free: true,
+          nvidiarouted: true,
+        },
+        {
+          id: "liquid/lfm-2.5-2.6b:free",
+          context: 65536,
+          maxoutput: 8192,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "minimax/minimax-m2.7:free",
+          context: 196608,
+          maxoutput: 196608,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "dots-studio/dots-3-note-preview:free",
+          context: 512000,
+          maxoutput: 460800,
+          vision: true,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "poolside/laguna-s-2.1:free",
+          context: 262144,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "poolside/laguna-xs-2.1:free",
+          context: 262144,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "nvidia/nemotron-3.5-content-safety:free",
+          context: 128000,
+          maxoutput: 8192,
+          vision: true,
+          reasoning: false,
+          free: true,
+          nvidiarouted: true,
+        },
+      ],
+      defaultmodel: "devthink",
+      rotation: {
+        mode: "persession",
+        models: [
+          "muse-spark-1.2-contributor-free",
+          "deepseek-v4-flash-free",
+          "mimo-v2.5-free",
+          "nemotron-3-ultra-free",
+          "ling-3.0-flash-fin-free",
+          "nemotron-3.5-lightning-free",
+          "laguna-s-2.1-free",
+          "thinkingmachines/inkling:free",
+          "tencent/hy3:free",
+          "thinkingmachines/inkling-small:free",
+          "inclusionai/ling-3.0-flash-fin:free",
+          "nvidia/nemotron-3-ultra-550b-a55b:free",
+          "minimax/minimax-m3:free",
+          "stepfun/step-3.7-flash:free",
+          "nvidia/nemotron-3.5-lightning:free",
+          "nvidia/nemotron-3-super-120b-a12b:free",
+          "cohere/north-mini-code:free",
+          "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free",
+          "liquid/lfm-2.5-2.6b:free",
+          "minimax/minimax-m2.7:free",
+          "dots-studio/dots-3-note-preview:free",
+          "poolside/laguna-s-2.1:free",
+          "poolside/laguna-xs-2.1:free",
+          "nvidia/nemotron-3.5-content-safety:free",
+        ],
+        everynmessages: 6,
+        maxmodelrotations: 1,
+        rotateonstatus: [],
+        rotateontimeout: false,
+        staggernewsessions: true,
+        sessionstore: "memory",
+      },
+      retry: {
+        maxretries: 0,
+        backoffbasems: 0,
+        statuses: [],
+        nonretryable: [400],
+        fallbackstatuses: [404, 429, 500, 502, 503, 529],
+        fallback: "crossprovider",
+      },
+      timeout: {
+        requestms: 300000,
+        streamms: 2147483647,
+      },
+      context: {
+        restoresessionhistory: true,
+        historylimit: 1000,
+        truncatemargin: 4096,
+        percallfallback: 1048576,
+        markshared: true,
+      },
+      thinking: {
+        defaultlevel: "high",
+        param: "chattemplatekwargs",
+      },
+      bodybuild: {
+        optionalparamspolicy: "omit-unspecified",
+        clampmaxtokens: false,
+      },
+      metamodel: {
+        id: "devthink",
+        fallbackcontext: 1048576,
+        maxoutput: 32768,
+        maskupstreammodel: true,
+        alwaysdisplay: true,
+        pattern: "24 free models across opencode zen and kilo with cross provider fallback",
+      },
+      routes: {
+        embeddings: "disabled",
+        embeddingsmessage: "v4 free gateways have no embeddings endpoint",
+      },
+      headers: {
+        useragent: "DevThink-Gateway/4.0",
+      },
+    },
+
+    // -----------------------------------------------------------------------
+    // v5 — openrouter free models single env key per-request rotation
+    // -----------------------------------------------------------------------
+    v5: {
+      id: "v5",
+      name: "V5 OpenRouter",
+      providername: "openrouter",
+      note: "openrouter free tier 19 models per-request round-robin",
+      upstreams: [{ name: "openrouter", baseurl: "https://openrouter.ai/api/v1" }],
+      auth: {
+        mode: "bearer",
+        keysources: ["env"],
+        envvar: "OPENROUTER_API_KEY",
+        keyrotation: "single",
+        required: true,
+        minkeylength: 10,
+        signupurl: "https://openrouter.ai",
+        extraheaders: {
+          "http-referer": "https://devthink.ai",
+          "x-title": "DevThink Gateway",
+        },
+      },
+      models: [
+        {
+          id: "z-ai/glm-5.2:free",
+          context: 1048576,
+          maxoutput: 131072,
+          vision: false,
+          reasoning: true,
+          free: true,
+          rank: { score: 53, tier: 1 },
+          exclusive: true,
+        },
+        {
+          id: "minimax/minimax-m3:free",
+          context: 1048576,
+          maxoutput: 16384,
+          vision: true,
+          reasoning: true,
+          free: true,
+          rank: { score: 45, tier: 1 },
+        },
+        {
+          id: "thinkingmachines/inkling:free",
+          context: 1048576,
+          maxoutput: 262144,
+          vision: true,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "tencent/hy3:free",
+          context: 262144,
+          maxoutput: 128000,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "thinkingmachines/inkling-small:free",
+          context: 1048576,
+          maxoutput: 262144,
+          vision: true,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "inclusionai/ling-3.0-flash-fin:free",
+          context: 262144,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "nvidia/nemotron-3-ultra-550b-a55b:free",
+          context: 1048576,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+          nvidiarouted: true,
+        },
+        {
+          id: "stepfun/step-3.7-flash:free",
+          context: 262144,
+          maxoutput: 262144,
+          vision: true,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "google/gemma-4-31b-it:free",
+          context: 262144,
+          maxoutput: 32768,
+          vision: true,
+          reasoning: true,
+          free: true,
+          exclusive: true,
+        },
+        {
+          id: "google/gemma-4-26b-a4b-it:free",
+          context: 262144,
+          maxoutput: 32768,
+          vision: true,
+          reasoning: true,
+          free: true,
+          exclusive: true,
+        },
+        {
+          id: "nvidia/nemotron-3-super-120b-a12b:free",
+          context: 1048576,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+          nvidiarouted: true,
+        },
+        {
+          id: "nvidia/nemotron-3.5-lightning:free",
+          context: 1048576,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+          nvidiarouted: true,
+        },
+        {
+          id: "cohere/north-mini-code:free",
+          context: 256000,
+          maxoutput: 64000,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free",
+          context: 256000,
+          maxoutput: 65536,
+          vision: true,
+          reasoning: true,
+          free: true,
+          nvidiarouted: true,
+        },
+        {
+          id: "liquid/lfm-2.5-2.6b:free",
+          context: 65536,
+          maxoutput: 8192,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "minimax/minimax-m2.7:free",
+          context: 196608,
+          maxoutput: 196608,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "dots-studio/dots-3-note-preview:free",
+          context: 512000,
+          maxoutput: 460800,
+          vision: true,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "poolside/laguna-s-2.1:free",
+          context: 262144,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+        {
+          id: "poolside/laguna-xs-2.1:free",
+          context: 262144,
+          maxoutput: 32768,
+          vision: false,
+          reasoning: true,
+          free: true,
+        },
+      ],
+      defaultmodel: "devthink",
+      rotation: {
+        mode: "perrequest",
+        models: [
+          "z-ai/glm-5.2:free",
+          "minimax/minimax-m3:free",
+          "thinkingmachines/inkling:free",
+          "tencent/hy3:free",
+          "thinkingmachines/inkling-small:free",
+          "inclusionai/ling-3.0-flash-fin:free",
+          "nvidia/nemotron-3-ultra-550b-a55b:free",
+          "stepfun/step-3.7-flash:free",
+          "google/gemma-4-31b-it:free",
+          "google/gemma-4-26b-a4b-it:free",
+          "nvidia/nemotron-3-super-120b-a12b:free",
+          "nvidia/nemotron-3.5-lightning:free",
+          "cohere/north-mini-code:free",
+          "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free",
+          "liquid/lfm-2.5-2.6b:free",
+          "minimax/minimax-m2.7:free",
+          "dots-studio/dots-3-note-preview:free",
+          "poolside/laguna-s-2.1:free",
+          "poolside/laguna-xs-2.1:free",
+        ],
+        everynmessages: 1,
+        maxmodelrotations: 1,
+        staggernewsessions: true,
+        sessionstore: "memory",
+      },
+      retry: {
+        maxretries: 0,
+        statuses: [],
+        nonretryable: [400],
+        fallback: "none",
+      },
+      timeout: {
+        requestms: 300000,
+        streamms: 2147483647,
+      },
+      context: {
+        restoresessionhistory: false,
+        truncate: false,
+      },
+      thinking: {
+        defaultlevel: "high",
+        param: "chattemplatekwargs",
+      },
+      bodybuild: {
+        optionalparamspolicy: "omit-unspecified",
+        clampmaxtokens: false,
+      },
+      metamodel: {
+        id: "devthink",
+        fallbackcontext: 1048576,
+        maxoutput: 32768,
+        maskupstreammodel: true,
+        alwaysdisplay: true,
+        pattern: "19 openrouter free models per-request round-robin",
+      },
+      routes: {
+        embeddings: "disabled",
+        embeddingsmessage: "openrouter has no embeddings endpoint on this gateway",
+      },
+      headers: {
+        useragent: "DevThink-Gateway/5.0",
+      },
+    },
+  },
+};
+
+/* (the gatewayviewconfig arm answered the same dual default export shape —
+ * the logicfold3 fold converted it to the named gatewayViewConfig seat
+ * above, the default retired with the variant file.) */
+