@openclaw/voice-call 2026.3.13 → 2026.5.2-beta.1

package/src/types.ts

@@ -1,11 +1,11 @@
-import { z } from "zod";
+import { z } from "openclaw/plugin-sdk/zod";
 import type { CallMode } from "./config.js";
 
 // -----------------------------------------------------------------------------
 // Provider Identifiers
 // -----------------------------------------------------------------------------
 
-export const ProviderNameSchema = z.enum(["telnyx", "twilio", "plivo", "mock"]);
+const ProviderNameSchema = z.enum(["telnyx", "twilio", "plivo", "mock"]);
 export type ProviderName = z.infer<typeof ProviderNameSchema>;
 
 // -----------------------------------------------------------------------------
@@ -16,13 +16,13 @@ export type ProviderName = z.infer<typeof ProviderNameSchema>;
 export type CallId = string;
 
 /** Provider-specific call identifier */
-export type ProviderCallId = string;
+type ProviderCallId = string;
 
 // -----------------------------------------------------------------------------
 // Call Lifecycle States
 // -----------------------------------------------------------------------------
 
-export const CallStateSchema = z.enum([
+const CallStateSchema = z.enum([
   // Non-terminal states
   "initiated",
   "ringing",
@@ -55,7 +55,7 @@ export const TerminalStates = new Set<CallState>([
   "voicemail",
 ]);
 
-export const EndReasonSchema = z.enum([
+const EndReasonSchema = z.enum([
   "completed",
   "hangup-user",
   "hangup-bot",
@@ -87,7 +87,7 @@ const BaseEventSchema = z.object({
   to: z.string().optional(),
 });
 
-export const NormalizedEventSchema = z.discriminatedUnion("type", [
+const NormalizedEventSchema = z.discriminatedUnion("type", [
   BaseEventSchema.extend({
     type: z.literal("call.initiated"),
   }),
@@ -134,14 +134,13 @@ export type NormalizedEvent = z.infer<typeof NormalizedEventSchema>;
 // Call Direction
 // -----------------------------------------------------------------------------
 
-export const CallDirectionSchema = z.enum(["outbound", "inbound"]);
-export type CallDirection = z.infer<typeof CallDirectionSchema>;
+const CallDirectionSchema = z.enum(["outbound", "inbound"]);
 
 // -----------------------------------------------------------------------------
 // Call Record
 // -----------------------------------------------------------------------------
 
-export const TranscriptEntrySchema = z.object({
+const TranscriptEntrySchema = z.object({
   timestamp: z.number(),
   speaker: z.enum(["bot", "user"]),
   text: z.string(),
@@ -212,8 +211,10 @@ export type InitiateCallInput = {
   to: string;
   webhookUrl: string;
   clientState?: Record<string, string>;
-  /** Inline TwiML to execute (skips webhook, used for notify mode) */
+  /** Inline TwiML to execute without fetching webhook TwiML. */
   inlineTwiml?: string;
+  /** TwiML to serve once before normal webhook-driven call handling resumes. */
+  preConnectTwiml?: string;
 };
 
 export type InitiateCallResult = {
@@ -227,6 +228,11 @@ export type HangupCallInput = {
   reason: EndReason;
 };
 
+export type AnswerCallInput = {
+  callId: CallId;
+  providerCallId: ProviderCallId;
+};
+
 export type PlayTtsInput = {
   callId: CallId;
   providerCallId: ProviderCallId;
@@ -235,6 +241,12 @@ export type PlayTtsInput = {
   locale?: string;
 };
 
+export type SendDtmfInput = {
+  callId: CallId;
+  providerCallId: ProviderCallId;
+  digits: string;
+};
+
 export type StartListeningInput = {
   callId: CallId;
   providerCallId: ProviderCallId;
@@ -274,31 +286,6 @@ export type OutboundCallOptions = {
   message?: string;
   /** Call mode (overrides config default) */
   mode?: CallMode;
-};
-
-// -----------------------------------------------------------------------------
-// Tool Result Types
-// -----------------------------------------------------------------------------
-
-export type InitiateCallToolResult = {
-  success: boolean;
-  callId?: string;
-  status?: "initiated" | "queued" | "no-answer" | "busy" | "failed";
-  error?: string;
-};
-
-export type ContinueCallToolResult = {
-  success: boolean;
-  transcript?: string;
-  error?: string;
-};
-
-export type SpeakToUserToolResult = {
-  success: boolean;
-  error?: string;
-};
-
-export type EndCallToolResult = {
-  success: boolean;
-  error?: string;
+  /** DTMF digits to send after the call is connected */
+  dtmfSequence?: string;
 };

package/src/utils.test.ts

@@ -0,0 +1,17 @@
+import os from "node:os";
+import path from "node:path";
+import { describe, expect, it } from "vitest";
+import { resolveUserPath } from "./utils.js";
+
+describe("resolveUserPath", () => {
+  it("returns trimmed empty input unchanged", () => {
+    expect(resolveUserPath("   ")).toBe("");
+  });
+
+  it("expands tildes and resolves relative paths", () => {
+    expect(resolveUserPath("~/voice-call/config.json")).toBe(
+      path.resolve(os.homedir(), "voice-call/config.json"),
+    );
+    expect(resolveUserPath("./voice-call")).toBe(path.resolve("./voice-call"));
+  });
+});

package/src/voice-mapping.test.ts

@@ -0,0 +1,34 @@
+import { describe, expect, it } from "vitest";
+import {
+  DEFAULT_POLLY_VOICE,
+  escapeXml,
+  getOpenAiVoiceNames,
+  isOpenAiVoice,
+  mapVoiceToPolly,
+} from "./voice-mapping.js";
+
+describe("voice mapping", () => {
+  it("escapes xml-special characters", () => {
+    expect(escapeXml(`5 < 6 & "quote" 'apostrophe' > 4`)).toBe(
+      "5 &lt; 6 &amp; &quot;quote&quot; &apos;apostrophe&apos; &gt; 4",
+    );
+  });
+
+  it("maps openai voices, passes through provider voices, and falls back to default", () => {
+    expect(mapVoiceToPolly("alloy")).toBe("Polly.Joanna");
+    expect(mapVoiceToPolly("ECHO")).toBe("Polly.Matthew");
+    expect(mapVoiceToPolly("Polly.Brian")).toBe("Polly.Brian");
+    expect(mapVoiceToPolly("Google.en-US-Standard-C")).toBe("Google.en-US-Standard-C");
+    expect(mapVoiceToPolly("unknown")).toBe(DEFAULT_POLLY_VOICE);
+    expect(mapVoiceToPolly(undefined)).toBe(DEFAULT_POLLY_VOICE);
+  });
+
+  it("detects known openai voices and lists them", () => {
+    expect(isOpenAiVoice("nova")).toBe(true);
+    expect(isOpenAiVoice("NOVA")).toBe(true);
+    expect(isOpenAiVoice("Polly.Joanna")).toBe(false);
+    expect(getOpenAiVoiceNames()).toEqual(
+      expect.arrayContaining(["alloy", "echo", "fable", "nova", "onyx", "shimmer"]),
+    );
+  });
+});

package/src/voice-mapping.ts

@@ -1,6 +1,7 @@
 /**
  * Voice mapping and XML utilities for voice call providers.
  */
+import { normalizeLowercaseStringOrEmpty } from "openclaw/plugin-sdk/text-runtime";
 
 /**
  * Escape XML special characters for TwiML and other XML responses.
@@ -49,14 +50,14 @@ export function mapVoiceToPolly(voice: string | undefined): string {
   }
 
   // Map OpenAI voices to Polly equivalents
-  return OPENAI_TO_POLLY_MAP[voice.toLowerCase()] || DEFAULT_POLLY_VOICE;
+  return OPENAI_TO_POLLY_MAP[normalizeLowercaseStringOrEmpty(voice)] || DEFAULT_POLLY_VOICE;
 }
 
 /**
  * Check if a voice name is a known OpenAI voice.
  */
 export function isOpenAiVoice(voice: string): boolean {
-  return voice.toLowerCase() in OPENAI_TO_POLLY_MAP;
+  return normalizeLowercaseStringOrEmpty(voice) in OPENAI_TO_POLLY_MAP;
 }
 
 /**