refactor: normalize voice-call runtime defaults

extensions/voice-call/src/config.test.ts

@@ -1,5 +1,10 @@
 import { afterEach, beforeEach, describe, expect, it } from "vitest";
-import { validateProviderConfig, resolveVoiceCallConfig, type VoiceCallConfig } from "./config.js";
+import {
+  validateProviderConfig,
+  normalizeVoiceCallConfig,
+  resolveVoiceCallConfig,
+  type VoiceCallConfig,
+} from "./config.js";
 import { createVoiceCallBaseConfig } from "./test-fixtures.js";
 
 function createBaseConfig(provider: "telnyx" | "twilio" | "plivo" | "mock"): VoiceCallConfig {
@@ -166,3 +171,22 @@ describe("validateProviderConfig", () => {
     });
   });
 });
+
+describe("normalizeVoiceCallConfig", () => {
+  it("fills nested runtime defaults from a partial config boundary", () => {
+    const normalized = normalizeVoiceCallConfig({
+      enabled: true,
+      provider: "mock",
+      streaming: {
+        enabled: true,
+        streamPath: "/custom-stream",
+      },
+    });
+
+    expect(normalized.serve.path).toBe("/voice/webhook");
+    expect(normalized.streaming.streamPath).toBe("/custom-stream");
+    expect(normalized.streaming.sttModel).toBe("gpt-4o-transcribe");
+    expect(normalized.tunnel.provider).toBe("none");
+    expect(normalized.webhookSecurity.allowedHosts).toEqual([]);
+  });
+});

extensions/voice-call/src/config.ts

@@ -350,17 +350,53 @@ export const VoiceCallConfigSchema = z
   .strict();
 
 export type VoiceCallConfig = z.infer<typeof VoiceCallConfigSchema>;
+type DeepPartial<T> =
+  T extends Array<infer U>
+    ? DeepPartial<U>[]
+    : T extends object
+      ? { [K in keyof T]?: DeepPartial<T[K]> }
+      : T;
+export type VoiceCallConfigInput = DeepPartial<VoiceCallConfig>;
 
 // -----------------------------------------------------------------------------
 // Configuration Helpers
 // -----------------------------------------------------------------------------
 
+const DEFAULT_VOICE_CALL_CONFIG = VoiceCallConfigSchema.parse({});
+
+function cloneDefaultVoiceCallConfig(): VoiceCallConfig {
+  return structuredClone(DEFAULT_VOICE_CALL_CONFIG);
+}
+
+export function normalizeVoiceCallConfig(config: VoiceCallConfigInput): VoiceCallConfig {
+  const defaults = cloneDefaultVoiceCallConfig();
+  return {
+    ...defaults,
+    ...config,
+    allowFrom: config.allowFrom ?? defaults.allowFrom,
+    outbound: { ...defaults.outbound, ...config.outbound },
+    serve: { ...defaults.serve, ...config.serve },
+    tailscale: { ...defaults.tailscale, ...config.tailscale },
+    tunnel: { ...defaults.tunnel, ...config.tunnel },
+    webhookSecurity: {
+      ...defaults.webhookSecurity,
+      ...config.webhookSecurity,
+      allowedHosts: config.webhookSecurity?.allowedHosts ?? defaults.webhookSecurity.allowedHosts,
+      trustedProxyIPs:
+        config.webhookSecurity?.trustedProxyIPs ?? defaults.webhookSecurity.trustedProxyIPs,
+    },
+    streaming: { ...defaults.streaming, ...config.streaming },
+    stt: { ...defaults.stt, ...config.stt },
+    tts: config.tts ?? defaults.tts,
+  };
+}
+
 /**
  * Resolves the configuration by merging environment variables into missing fields.
  * Returns a new configuration object with environment variables applied.
  */
-export function resolveVoiceCallConfig(config: VoiceCallConfig): VoiceCallConfig {
+export function resolveVoiceCallConfig(config: VoiceCallConfigInput): VoiceCallConfig {
-  const resolved = JSON.parse(JSON.stringify(config)) as VoiceCallConfig;
+  const resolved = normalizeVoiceCallConfig(config);
 
   // Telnyx
   if (resolved.provider === "telnyx") {
@@ -405,7 +441,7 @@ export function resolveVoiceCallConfig(config: VoiceCallConfig): VoiceCallConfig
     resolved.webhookSecurity.trustForwardingHeaders ?? false;
   resolved.webhookSecurity.trustedProxyIPs = resolved.webhookSecurity.trustedProxyIPs ?? [];
 
-  return resolved;
+  return normalizeVoiceCallConfig(resolved);
 }
 
 /**

extensions/voice-call/src/providers/tts-openai.test.ts

@@ -3,6 +3,8 @@ import type { OpenAITTSConfig } from "./tts-openai.js";
 import { OpenAITTSProvider } from "./tts-openai.js";
 
 type ProviderInternals = {
+  model: string;
+  voice: string;
   speed: number;
 };
 
@@ -27,4 +29,15 @@ describe("OpenAITTSProvider constructor defaults", () => {
 
     expect(provider.speed).toBe(1.0);
   });
+
+  it("treats blank model and voice overrides as unset", () => {
+    const provider = readProviderInternals({
+      apiKey: "sk-test", // pragma: allowlist secret
+      model: "   ",
+      voice: "",
+    });
+
+    expect(provider.model).toBe("gpt-4o-mini-tts");
+    expect(provider.voice).toBe("coral");
+  });
 });

extensions/voice-call/src/providers/tts-openai.ts

@@ -66,6 +66,11 @@ export const OPENAI_TTS_VOICES = [
 
 export type OpenAITTSVoice = (typeof OPENAI_TTS_VOICES)[number];
 
+function trimToUndefined(value: string | undefined): string | undefined {
+  const trimmed = value?.trim();
+  return trimmed ? trimmed : undefined;
+}
+
 /**
  * OpenAI TTS Provider for generating speech audio.
  */
@@ -77,13 +82,14 @@ export class OpenAITTSProvider {
   private instructions?: string;
 
   constructor(config: OpenAITTSConfig = {}) {
-    this.apiKey = config.apiKey || process.env.OPENAI_API_KEY || "";
+    this.apiKey =
+      trimToUndefined(config.apiKey) ?? trimToUndefined(process.env.OPENAI_API_KEY) ?? "";
     // Default to gpt-4o-mini-tts for intelligent realtime applications
-    this.model = config.model || "gpt-4o-mini-tts";
+    this.model = trimToUndefined(config.model) ?? "gpt-4o-mini-tts";
     // Default to coral - good balance of quality and natural tone
-    this.voice = (config.voice as OpenAITTSVoice) || "coral";
+    this.voice = (trimToUndefined(config.voice) as OpenAITTSVoice | undefined) ?? "coral";
     this.speed = config.speed ?? 1.0;
-    this.instructions = config.instructions;
+    this.instructions = trimToUndefined(config.instructions);
 
     if (!this.apiKey) {
       throw new Error("OpenAI API key required (set OPENAI_API_KEY or pass apiKey)");
@@ -105,7 +111,7 @@ export class OpenAITTSProvider {
     };
 
     // Add instructions if using gpt-4o-mini-tts model
-    const effectiveInstructions = instructions || this.instructions;
+    const effectiveInstructions = trimToUndefined(instructions) ?? this.instructions;
     if (effectiveInstructions && this.model.includes("gpt-4o-mini-tts")) {
       body.instructions = effectiveInstructions;
     }

extensions/voice-call/src/webhook.test.ts

@@ -274,6 +274,32 @@ describe("VoiceCallWebhookServer replay handling", () => {
   });
 });
 
+describe("VoiceCallWebhookServer response normalization", () => {
+  it("preserves explicit empty provider response bodies", async () => {
+    const responseProvider: VoiceCallProvider = {
+      ...provider,
+      parseWebhookEvent: () => ({
+        events: [],
+        statusCode: 204,
+        providerResponseBody: "",
+      }),
+    };
+    const { manager } = createManager([]);
+    const config = createConfig({ serve: { port: 0, bind: "127.0.0.1", path: "/voice/webhook" } });
+    const server = new VoiceCallWebhookServer(config, manager, responseProvider);
+
+    try {
+      const baseUrl = await server.start();
+      const response = await postWebhookForm(server, baseUrl, "CallSid=CA123&SpeechResult=hello");
+
+      expect(response.status).toBe(204);
+      expect(await response.text()).toBe("");
+    } finally {
+      await server.stop();
+    }
+  });
+});
+
 describe("VoiceCallWebhookServer start idempotency", () => {
   it("returns existing URL when start() is called twice without stop()", async () => {
     const { manager } = createManager([]);

extensions/voice-call/src/webhook.ts

@@ -5,7 +5,7 @@ import {
   readRequestBodyWithLimit,
   requestBodyErrorToText,
 } from "openclaw/plugin-sdk/voice-call";
-import type { VoiceCallConfig } from "./config.js";
+import { normalizeVoiceCallConfig, type VoiceCallConfig } from "./config.js";
 import type { CoreConfig } from "./core-bridge.js";
 import type { CallManager } from "./manager.js";
 import type { MediaStreamConfig } from "./media-stream.js";
@@ -24,6 +24,26 @@ type WebhookResponsePayload = {
   headers?: Record<string, string>;
 };
 
+function buildRequestUrl(
+  requestUrl: string | undefined,
+  requestHost: string | undefined,
+  fallbackHost = "localhost",
+): URL {
+  return new URL(requestUrl ?? "/", `http://${requestHost ?? fallbackHost}`);
+}
+
+function normalizeWebhookResponse(parsed: {
+  statusCode?: number;
+  providerResponseHeaders?: Record<string, string>;
+  providerResponseBody?: string;
+}): WebhookResponsePayload {
+  return {
+    statusCode: parsed.statusCode ?? 200,
+    headers: parsed.providerResponseHeaders,
+    body: parsed.providerResponseBody ?? "OK",
+  };
+}
+
 /**
  * HTTP server for receiving voice call webhooks from providers.
  * Supports WebSocket upgrades for media streams when streaming is enabled.
@@ -46,13 +66,13 @@ export class VoiceCallWebhookServer {
     provider: VoiceCallProvider,
     coreConfig?: CoreConfig,
   ) {
-    this.config = config;
+    this.config = normalizeVoiceCallConfig(config);
     this.manager = manager;
     this.provider = provider;
     this.coreConfig = coreConfig ?? null;
 
     // Initialize media stream handler if streaming is enabled
-    if (config.streaming?.enabled) {
+    if (this.config.streaming.enabled) {
       this.initializeMediaStreaming();
     }
   }
@@ -68,7 +88,8 @@ export class VoiceCallWebhookServer {
    * Initialize media streaming with OpenAI Realtime STT.
    */
   private initializeMediaStreaming(): void {
-    const apiKey = this.config.streaming?.openaiApiKey || process.env.OPENAI_API_KEY;
+    const streaming = this.config.streaming;
+    const apiKey = streaming.openaiApiKey ?? process.env.OPENAI_API_KEY;
 
     if (!apiKey) {
       console.warn("[voice-call] Streaming enabled but no OpenAI API key found");
@@ -77,17 +98,17 @@ export class VoiceCallWebhookServer {
 
     const sttProvider = new OpenAIRealtimeSTTProvider({
       apiKey,
-      model: this.config.streaming?.sttModel,
+      model: streaming.sttModel,
-      silenceDurationMs: this.config.streaming?.silenceDurationMs,
+      silenceDurationMs: streaming.silenceDurationMs,
-      vadThreshold: this.config.streaming?.vadThreshold,
+      vadThreshold: streaming.vadThreshold,
     });
 
     const streamConfig: MediaStreamConfig = {
       sttProvider,
-      preStartTimeoutMs: this.config.streaming?.preStartTimeoutMs,
+      preStartTimeoutMs: streaming.preStartTimeoutMs,
-      maxPendingConnections: this.config.streaming?.maxPendingConnections,
+      maxPendingConnections: streaming.maxPendingConnections,
-      maxPendingConnectionsPerIp: this.config.streaming?.maxPendingConnectionsPerIp,
+      maxPendingConnectionsPerIp: streaming.maxPendingConnectionsPerIp,
-      maxConnections: this.config.streaming?.maxConnections,
+      maxConnections: streaming.maxConnections,
       shouldAcceptStream: ({ callId, token }) => {
         const call = this.manager.getCallByProviderCallId(callId);
         if (!call) {
@@ -190,7 +211,7 @@ export class VoiceCallWebhookServer {
    */
   async start(): Promise<string> {
     const { port, bind, path: webhookPath } = this.config.serve;
-    const streamPath = this.config.streaming?.streamPath || "/voice/stream";
+    const streamPath = this.config.streaming.streamPath;
 
     // Guard: if a server is already listening, return the existing URL.
     // This prevents EADDRINUSE when start() is called more than once on the
@@ -280,8 +301,7 @@ export class VoiceCallWebhookServer {
 
   private getUpgradePathname(request: http.IncomingMessage): string | null {
     try {
-      const host = request.headers.host || "localhost";
+      return buildRequestUrl(request.url, request.headers.host).pathname;
-      return new URL(request.url || "/", `http://${host}`).pathname;
     } catch {
       return null;
     }
@@ -322,7 +342,7 @@ export class VoiceCallWebhookServer {
     req: http.IncomingMessage,
     webhookPath: string,
   ): Promise<WebhookResponsePayload> {
-    const url = new URL(req.url || "/", `http://${req.headers.host}`);
+    const url = buildRequestUrl(req.url, req.headers.host);
 
     if (url.pathname === "/voice/hold-music") {
       return {
@@ -360,7 +380,7 @@ export class VoiceCallWebhookServer {
     const ctx: WebhookContext = {
       headers: req.headers as Record<string, string | string[] | undefined>,
       rawBody: body,
-      url: `http://${req.headers.host}${req.url}`,
+      url: url.toString(),
       method: "POST",
       query: Object.fromEntries(url.searchParams),
       remoteAddress: req.socket.remoteAddress ?? undefined,
@@ -386,11 +406,7 @@ export class VoiceCallWebhookServer {
       this.processParsedEvents(parsed.events);
     }
 
-    return {
+    return normalizeWebhookResponse(parsed);
-      statusCode: parsed.statusCode || 200,
-      headers: parsed.providerResponseHeaders,
-      body: parsed.providerResponseBody || "OK",
-    };
   }
 
   private processParsedEvents(events: NormalizedEvent[]): void {