@neutrome/open-ai-router 0.9.3 → 0.9.6

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@neutrome/open-ai-router",
-  "version": "0.9.3",
+  "version": "0.9.6",
   "type": "module",
   "files": [
     "src",
@@ -17,7 +17,7 @@
     "posthog-node": "^5.41.0",
     "zod": "^4.2.1",
     "@neutrome/lil-engine": "0.6.1",
-    "@neutrome/lilsdk": "0.6.3"
+    "@neutrome/lilsdk": "0.6.5"
   },
   "devDependencies": {
     "@types/node": "^25.9.3",

package/src/app/factory.test.ts

@@ -10,6 +10,7 @@ const config: RouterConfig = {
     openrouter: {
       apiBaseUrl: "https://openrouter.ai/api/v1",
       style: "chat-completions",
+      timeoutMs: 120_000,
       exports: ["*"],
     },
   },
@@ -31,6 +32,7 @@ describe("createRouterApp trace finalization", () => {
           google: {
             apiBaseUrl: "https://google.example.test/v1beta/openai",
             style: "chat-completions",
+            timeoutMs: 120_000,
             exports: ["*"],
           },
         },

package/src/app/factory.ts

@@ -300,9 +300,6 @@ export function createRouterApp(options: RouterAppOptions) {
         onNamedInvocation(result) {
           usage.recordInvocation(result);
         },
-        ...(options.upstreamTimeoutMs !== undefined
-          ? { upstreamTimeoutMs: options.upstreamTimeoutMs }
-          : {}),
         ...(options.remoteMcpClientFactory
           ? { remoteMcpClientFactory: options.remoteMcpClientFactory }
           : {}),

package/src/app/types.ts

@@ -39,7 +39,6 @@ export type RouterAppOptions = {
   observe?: ExecutionRuntimeOptions["observe"];
   requestIdFactory?: ExecutionRuntimeOptions["requestIdFactory"];
   executionIdFactory?: ExecutionRuntimeOptions["executionIdFactory"];
-  upstreamTimeoutMs?: number;
   telemetryDistinctId?: string;
   remoteMcpClientFactory?: RemoteMcpClientFactory;
 };

package/src/router/config.ts

@@ -23,6 +23,7 @@ export type ProviderAuthDriverConfig =
 export type ProviderTargetConfig = {
   apiBaseUrl: string;
   style: ProviderApiStyle;
+  timeoutMs: number;
   exports?: string[];
   headers?: Record<string, string>;
   auth?: ProviderAuthDriverConfig[];

package/src/router/execute.test.ts

@@ -20,6 +20,7 @@ describe("router execution", () => {
           openrouter: {
             apiBaseUrl: "https://openrouter.ai/api/v1",
             style: "chat-completions",
+            timeoutMs: 120_000,
             exports: ["*"],
           },
         },
@@ -121,6 +122,7 @@ describe("router execution", () => {
           openrouter: {
             apiBaseUrl: "https://openrouter.ai/api/v1",
             style: "chat-completions",
+            timeoutMs: 120_000,
             exports: ["*"],
           },
         },
@@ -276,6 +278,7 @@ describe("router execution", () => {
           openai: {
             apiBaseUrl: "https://api.openai.com/v1",
             style: "responses",
+            timeoutMs: 120_000,
             exports: ["*"],
           },
         },
@@ -356,6 +359,7 @@ describe("router execution", () => {
           anthropic: {
             apiBaseUrl: "https://api.anthropic.com",
             style: "anthropic-messages",
+            timeoutMs: 120_000,
             exports: ["*"],
           },
         },
@@ -437,6 +441,7 @@ describe("router execution", () => {
           google: {
             apiBaseUrl: "https://generativelanguage.googleapis.com/v1beta",
             style: "google-genai",
+            timeoutMs: 120_000,
             exports: ["*"],
           },
         },
@@ -570,6 +575,73 @@ describe("router execution", () => {
     });
   });
 
+  it("applies configured suffix transforms to nested executor invocations", async () => {
+    let transformApplied = false;
+    const runtime = createRouterRuntime({
+      config: {
+        modelSuffixes: ["trim-input"],
+        providers: {},
+        modelNamespaces: {
+          namespace: {
+            exports: ["*"],
+            models: {
+              root: { target: { executor: "root" } },
+              model: { target: { executor: "model" } },
+            },
+          },
+        },
+      },
+    });
+    const ctx: RequestContext = {
+      request: new Request("http://local/v1/chat/completions", {
+        method: "POST",
+        headers: { "content-type": "application/json" },
+        body: JSON.stringify({
+          model: "namespace/root",
+          messages: [{ role: "user", content: "hi" }],
+        }),
+      }),
+      incomingAuth: new Map(),
+      env: {},
+    };
+
+    const response = await handleChatCompletions({
+      runtime,
+      ctx,
+      transforms: {
+        "trim-input": {
+          name: "trim-input",
+          capabilities: [],
+          apply(request) {
+            transformApplied = true;
+            return request;
+          },
+        },
+      },
+      executorImplementations: {
+        root: {
+          async execute(request, executorContext) {
+            return executorContext.invoke(
+              "namespace/model+trim-input",
+              request,
+            );
+          },
+          async *stream() {},
+        },
+        model: {
+          async execute() {
+            expect(transformApplied).toBe(true);
+            return appendAssistantMessage(createProgram(), "done");
+          },
+          async *stream() {},
+        },
+      },
+    });
+
+    expect(response.status).toBe(200);
+    expect(transformApplied).toBe(true);
+  });
+
   it("streams provider chunks as openai sse output", async () => {
     const runtime = createRouterRuntime({
       config: {
@@ -577,6 +649,7 @@ describe("router execution", () => {
           openrouter: {
             apiBaseUrl: "https://openrouter.ai/api/v1",
             style: "chat-completions",
+            timeoutMs: 120_000,
             exports: ["*"],
           },
         },
@@ -655,6 +728,7 @@ describe("router execution", () => {
           openrouter: {
             apiBaseUrl: "https://openrouter.ai/api/v1",
             style: "chat-completions",
+            timeoutMs: 1_000,
             exports: ["*"],
           },
         },
@@ -694,13 +768,12 @@ describe("router execution", () => {
     const response = await handleChatCompletions({
       runtime,
       ctx,
-      upstreamTimeoutMs: 5,
     });
 
     expect(response.status).toBe(504);
     expect(await response.json()).toEqual({
       error: {
-        message: "Upstream provider timed out after 5ms",
+        message: "Upstream provider timed out after 1000ms",
         type: "invalid_request_error",
         param: null,
         code: "upstream_timeout",
@@ -716,6 +789,7 @@ describe("router execution", () => {
           anthropic: {
             apiBaseUrl: "https://api.anthropic.com",
             style: "anthropic-messages",
+            timeoutMs: 120_000,
             exports: ["*"],
           },
         },

package/src/router/execute.ts

@@ -24,12 +24,6 @@ import {
   toBody,
   toProviderError,
 } from "./error-codec.ts";
-import {
-  createUpstreamSignal,
-  fetchProvider,
-  getProviderOrThrow,
-  normalizeUpstreamAbort,
-} from "./upstream-client.ts";
 import { emitProviderUsage, injectStreamUsage } from "./provider-telemetry.ts";
 import {
   applyRoutingHeaders,
@@ -66,8 +60,6 @@ import { countUserProgramTokens, protocolUsage } from "../usage/user-usage.ts";
 
 const encoder = new TextEncoder();
 const decoder = new TextDecoder();
-const DEFAULT_UPSTREAM_TIMEOUT_MS = 120_000;
-
 export type RouterExecutionOptions = Pick<
   ExecutionRuntimeOptions,
   | "cache"
@@ -78,7 +70,6 @@ export type RouterExecutionOptions = Pick<
 > & {
   executorImplementations?: Readonly<Record<string, Executor>>;
   transforms?: Readonly<Record<string, ProgramTransform>>;
-  upstreamTimeoutMs?: number;
   remoteMcpClientFactory?: RemoteMcpClientFactory;
   incomingExecutionId?: string;
   userUsage?: { inputTokens: number };
@@ -132,11 +123,7 @@ export function createRouterExecutionRuntime(
       }
       return resolveRequestTarget(runtime, request).target;
     },
-    providerInvoker: createFetchProviderInvoker(
-      runtime,
-      ctx,
-      options.upstreamTimeoutMs,
-    ),
+    providerInvoker: createFetchProviderInvoker(runtime, ctx),
   };
 
   if (options.executorImplementations) {

package/src/router/provider-invoker.ts

@@ -19,17 +19,17 @@ import type { RouterRuntime } from "./runtime.ts";
 import { observeTimedExecution } from "./execution-events.ts";
 import { normalizeChatRequest } from "../telemetry/chat-normalization.ts";
 
-const DEFAULT_UPSTREAM_TIMEOUT_MS = 120_000;
-
 export function createFetchProviderInvoker(
   runtime: RouterRuntime,
   ctx: RequestContext,
-  upstreamTimeoutMs = DEFAULT_UPSTREAM_TIMEOUT_MS,
 ): ProviderInvoker {
   return {
     async execute(request, providerCtx) {
       const provider = getProviderOrThrow(runtime, providerCtx.target.provider);
-      const timed = createUpstreamSignal(providerCtx.signal, upstreamTimeoutMs);
+      const timed = createUpstreamSignal(
+        providerCtx.signal,
+        provider.timeoutMs,
+      );
 
       try {
         const resolvedRequest = setModel(request, providerCtx.target.model);
@@ -126,13 +126,7 @@ export function createFetchProviderInvoker(
     },
 
     stream(request, providerCtx) {
-      return streamProvider(
-        runtime,
-        ctx,
-        upstreamTimeoutMs,
-        request,
-        providerCtx,
-      );
+      return streamProvider(runtime, ctx, request, providerCtx);
     },
   };
 }

package/src/router/provider-stream.ts

@@ -29,12 +29,11 @@ const decoder = new TextDecoder();
 export async function* streamProvider(
   runtime: RouterRuntime,
   ctx: RequestContext,
-  upstreamTimeoutMs: number,
   request: Program,
   providerCtx: ProviderInvocationContext,
 ): AsyncGenerator<Program> {
   const provider = getProviderOrThrow(runtime, providerCtx.target.provider);
-  const timed = createUpstreamSignal(providerCtx.signal, upstreamTimeoutMs);
+  const timed = createUpstreamSignal(providerCtx.signal, provider.timeoutMs);
   let reader: ReadableStreamDefaultReader<Uint8Array> | null = null;
   let buffer = "";
 

package/src/router/remote-mcp-execution.ts

@@ -11,7 +11,6 @@ import type { RouterRuntime } from "./runtime.ts";
 type RemoteMcpExecutionOptions = {
   executorImplementations?: Readonly<Record<string, Executor>>;
   transforms?: Readonly<Record<string, ProgramTransform>>;
-  upstreamTimeoutMs?: number;
   remoteMcpClientFactory?: RemoteMcpClientFactory;
   incomingExecutionId?: string;
 };

package/src/router/resolve.test.ts

@@ -16,6 +16,7 @@ describe("router resolution", () => {
           openrouter: {
             apiBaseUrl: "https://openrouter.ai/api/v1",
             style: "chat-completions",
+            timeoutMs: 120_000,
             exports: ["*"],
           },
         },
@@ -53,6 +54,7 @@ describe("router resolution", () => {
           openrouter: {
             apiBaseUrl: "https://openrouter.ai/api/v1",
             style: "chat-completions",
+            timeoutMs: 120_000,
             exports: ["google/*"],
           },
         },
@@ -80,6 +82,7 @@ describe("router resolution", () => {
           emma: {
             apiBaseUrl: "https://api.example.test/v1",
             style: "chat-completions",
+            timeoutMs: 120_000,
             exports: [],
           },
         },
@@ -107,6 +110,7 @@ describe("router resolution", () => {
           openrouter: {
             apiBaseUrl: "https://openrouter.ai/api/v1",
             style: "chat-completions",
+            timeoutMs: 120_000,
             exports: ["*"],
           },
         },
@@ -126,6 +130,7 @@ describe("router resolution", () => {
           openrouter: {
             apiBaseUrl: "https://openrouter.ai/api/v1",
             style: "chat-completions",
+            timeoutMs: 120_000,
             exports: ["openai/gpt-4.1-mini"],
           },
         },
@@ -163,6 +168,7 @@ describe("router resolution", () => {
           openrouter: {
             apiBaseUrl: "https://openrouter.ai/api/v1",
             style: "chat-completions",
+            timeoutMs: 120_000,
             exports: ["*"],
           },
         },
@@ -191,6 +197,7 @@ describe("router resolution", () => {
           openrouter: {
             apiBaseUrl: "https://openrouter.ai/api/v1",
             style: "chat-completions",
+            timeoutMs: 120_000,
             exports: ["*"],
           },
         },

package/src/router/runtime.test.ts

@@ -4,6 +4,7 @@ import { createRouterRuntime } from "./runtime.ts";
 const provider = {
   apiBaseUrl: "https://api.example.test/v1",
   style: "chat-completions" as const,
+  timeoutMs: 120_000,
 };
 
 describe("router runtime configuration", () => {
@@ -55,4 +56,17 @@ describe("router runtime configuration", () => {
     });
     expect(runtime.modelSuffixes).toEqual(new Set(["trim"]));
   });
+
+  it("requires provider timeouts from one second through fifteen minutes", () => {
+    for (const timeoutMs of [999, 900_001, 1.5]) {
+      expect(() =>
+        createRouterRuntime({
+          config: {
+            providers: { upstream: { ...provider, timeoutMs } },
+            modelNamespaces: {},
+          },
+        }),
+      ).toThrow("timeout must be an integer between 1000 and 900000ms");
+    }
+  });
 });

package/src/router/runtime.ts

@@ -12,6 +12,7 @@ export type ProviderRuntime = {
   name: string;
   style: ProviderTargetConfig["style"];
   apiBaseUrl: string;
+  timeoutMs: number;
   endpointPath: string;
   headers: Readonly<Record<string, string>>;
   auth?: ProviderTargetConfig["auth"];
@@ -90,6 +91,7 @@ function validateRouterConfig(config: RouterConfig): void {
   validateModelSuffixes(config.modelSuffixes);
   for (const [name, provider] of Object.entries(config.providers)) {
     validateHttpUrl(provider.apiBaseUrl, `Provider ${name}`);
+    validateProviderTimeout(provider.timeoutMs, `Provider ${name}`);
   }
   for (const [name, mcp] of Object.entries(config.mcps ?? {})) {
     validateHttpUrl(mcp.url, `MCP ${name}`);
@@ -120,6 +122,14 @@ function validateRouterConfig(config: RouterConfig): void {
   }
 }
 
+function validateProviderTimeout(value: number, label: string): void {
+  if (!Number.isInteger(value) || value < 1_000 || value > 900_000) {
+    throw new Error(
+      `${label} timeout must be an integer between 1000 and 900000ms`,
+    );
+  }
+}
+
 function validateModelSuffixes(suffixes: readonly string[] | undefined): void {
   if (!suffixes) return;
   const names = new Set<string>();
@@ -154,6 +164,7 @@ function buildProviderRuntime(
     name,
     style: config.style,
     apiBaseUrl: config.apiBaseUrl,
+    timeoutMs: config.timeoutMs,
     endpointPath: defaultEndpointPath(config.style),
     headers: config.headers ?? {},
     auth: config.auth,

package/src/router/upstream-client.test.ts

@@ -8,6 +8,7 @@ const provider = (
   name: "google",
   style: "google-genai",
   apiBaseUrl: "https://generativelanguage.googleapis.com/v1beta",
+  timeoutMs: 120_000,
   endpointPath: "",
   headers: {},
   auth: [{ kind: "env", env: "GEMINI_API_KEY", header: "x-goog-api-key" }],

package/src/worker.ts

@@ -11,6 +11,7 @@ const config: RouterConfig = {
     openrouter: {
       apiBaseUrl: "https://openrouter.ai/api/v1",
       style: "chat-completions",
+      timeoutMs: 120_000,
       exports: ["*:free"],
       auth: [
         {