fix(whatsapp): trim leading whitespace in direct outbound sends (#43539)

Trim leading whitespace from direct WhatsApp text and media caption sends.

Also guard empty text-only web sends after trimming.

src/channels/plugins/outbound/whatsapp.sendpayload.test.ts

@@ -1,4 +1,4 @@
-import { describe, vi } from "vitest";
+import { describe, expect, it, vi } from "vitest";
 import type { ReplyPayload } from "../../../auto-reply/types.js";
 import {
   installSendPayloadContractSuite,
@@ -34,4 +34,92 @@ describe("whatsappOutbound sendPayload", () => {
     chunking: { mode: "split", longTextLength: 5000, maxChunkLength: 4000 },
     createHarness,
   });
+
+  it("trims leading whitespace for direct text sends", async () => {
+    const sendWhatsApp = vi.fn(async () => ({ messageId: "wa-1", toJid: "jid" }));
+
+    await whatsappOutbound.sendText!({
+      cfg: {},
+      to: "5511999999999@c.us",
+      text: "\n \thello",
+      deps: { sendWhatsApp },
+    });
+
+    expect(sendWhatsApp).toHaveBeenCalledWith("5511999999999@c.us", "hello", {
+      verbose: false,
+      cfg: {},
+      accountId: undefined,
+      gifPlayback: undefined,
+    });
+  });
+
+  it("trims leading whitespace for direct media captions", async () => {
+    const sendWhatsApp = vi.fn(async () => ({ messageId: "wa-1", toJid: "jid" }));
+
+    await whatsappOutbound.sendMedia!({
+      cfg: {},
+      to: "5511999999999@c.us",
+      text: "\n \tcaption",
+      mediaUrl: "/tmp/test.png",
+      deps: { sendWhatsApp },
+    });
+
+    expect(sendWhatsApp).toHaveBeenCalledWith("5511999999999@c.us", "caption", {
+      verbose: false,
+      cfg: {},
+      mediaUrl: "/tmp/test.png",
+      mediaLocalRoots: undefined,
+      accountId: undefined,
+      gifPlayback: undefined,
+    });
+  });
+
+  it("trims leading whitespace for sendPayload text and caption delivery", async () => {
+    const sendWhatsApp = vi.fn(async () => ({ messageId: "wa-1", toJid: "jid" }));
+
+    await whatsappOutbound.sendPayload!({
+      cfg: {},
+      to: "5511999999999@c.us",
+      text: "",
+      payload: { text: "\n\nhello" },
+      deps: { sendWhatsApp },
+    });
+    await whatsappOutbound.sendPayload!({
+      cfg: {},
+      to: "5511999999999@c.us",
+      text: "",
+      payload: { text: "\n\ncaption", mediaUrl: "/tmp/test.png" },
+      deps: { sendWhatsApp },
+    });
+
+    expect(sendWhatsApp).toHaveBeenNthCalledWith(1, "5511999999999@c.us", "hello", {
+      verbose: false,
+      cfg: {},
+      accountId: undefined,
+      gifPlayback: undefined,
+    });
+    expect(sendWhatsApp).toHaveBeenNthCalledWith(2, "5511999999999@c.us", "caption", {
+      verbose: false,
+      cfg: {},
+      mediaUrl: "/tmp/test.png",
+      mediaLocalRoots: undefined,
+      accountId: undefined,
+      gifPlayback: undefined,
+    });
+  });
+
+  it("skips whitespace-only text payloads", async () => {
+    const sendWhatsApp = vi.fn();
+
+    const result = await whatsappOutbound.sendPayload!({
+      cfg: {},
+      to: "5511999999999@c.us",
+      text: "",
+      payload: { text: "\n \t" },
+      deps: { sendWhatsApp },
+    });
+
+    expect(result).toEqual({ channel: "whatsapp", messageId: "" });
+    expect(sendWhatsApp).not.toHaveBeenCalled();
+  });
 });

src/channels/plugins/outbound/whatsapp.ts

@@ -5,6 +5,10 @@ import { resolveWhatsAppOutboundTarget } from "../../../whatsapp/resolve-outboun
 import type { ChannelOutboundAdapter } from "../types.js";
 import { sendTextMediaPayload } from "./direct-text-media.js";
 
+function trimLeadingWhitespace(text: string | undefined): string {
+  return text?.trimStart() ?? "";
+}
+
 export const whatsappOutbound: ChannelOutboundAdapter = {
   deliveryMode: "gateway",
   chunker: chunkText,
@@ -13,12 +17,32 @@ export const whatsappOutbound: ChannelOutboundAdapter = {
   pollMaxOptions: 12,
   resolveTarget: ({ to, allowFrom, mode }) =>
     resolveWhatsAppOutboundTarget({ to, allowFrom, mode }),
-  sendPayload: async (ctx) =>
-    await sendTextMediaPayload({ channel: "whatsapp", ctx, adapter: whatsappOutbound }),
+  sendPayload: async (ctx) => {
+    const text = trimLeadingWhitespace(ctx.payload.text);
+    const hasMedia = Boolean(ctx.payload.mediaUrl) || (ctx.payload.mediaUrls?.length ?? 0) > 0;
+    if (!text && !hasMedia) {
+      return { channel: "whatsapp", messageId: "" };
+    }
+    return await sendTextMediaPayload({
+      channel: "whatsapp",
+      ctx: {
+        ...ctx,
+        payload: {
+          ...ctx.payload,
+          text,
+        },
+      },
+      adapter: whatsappOutbound,
+    });
+  },
   sendText: async ({ cfg, to, text, accountId, deps, gifPlayback }) => {
+    const normalizedText = trimLeadingWhitespace(text);
+    if (!normalizedText) {
+      return { channel: "whatsapp", messageId: "" };
+    }
     const send =
       deps?.sendWhatsApp ?? (await import("../../../web/outbound.js")).sendMessageWhatsApp;
-    const result = await send(to, text, {
+    const result = await send(to, normalizedText, {
       verbose: false,
       cfg,
       accountId: accountId ?? undefined,
@@ -27,9 +51,10 @@ export const whatsappOutbound: ChannelOutboundAdapter = {
     return { channel: "whatsapp", ...result };
   },
   sendMedia: async ({ cfg, to, text, mediaUrl, mediaLocalRoots, accountId, deps, gifPlayback }) => {
+    const normalizedText = trimLeadingWhitespace(text);
     const send =
       deps?.sendWhatsApp ?? (await import("../../../web/outbound.js")).sendMessageWhatsApp;
-    const result = await send(to, text, {
+    const result = await send(to, normalizedText, {
       verbose: false,
       cfg,
       mediaUrl,

src/web/outbound.test.ts

@@ -48,6 +48,34 @@ describe("web outbound", () => {
     expect(sendMessage).toHaveBeenCalledWith("+1555", "hi", undefined, undefined);
   });
 
+  it("trims leading whitespace before sending text and captions", async () => {
+    await sendMessageWhatsApp("+1555", "\n \thello", { verbose: false });
+    expect(sendMessage).toHaveBeenLastCalledWith("+1555", "hello", undefined, undefined);
+
+    const buf = Buffer.from("img");
+    loadWebMediaMock.mockResolvedValueOnce({
+      buffer: buf,
+      contentType: "image/jpeg",
+      kind: "image",
+    });
+    await sendMessageWhatsApp("+1555", "\n \tcaption", {
+      verbose: false,
+      mediaUrl: "/tmp/pic.jpg",
+    });
+    expect(sendMessage).toHaveBeenLastCalledWith("+1555", "caption", buf, "image/jpeg");
+  });
+
+  it("skips whitespace-only text sends without media", async () => {
+    const result = await sendMessageWhatsApp("+1555", "\n \t", { verbose: false });
+
+    expect(result).toEqual({
+      messageId: "",
+      toJid: "1555@s.whatsapp.net",
+    });
+    expect(sendComposingTo).not.toHaveBeenCalled();
+    expect(sendMessage).not.toHaveBeenCalled();
+  });
+
   it("throws a helpful error when no active listener exists", async () => {
     setActiveWebListener(null);
     await expect(

src/web/outbound.ts

@@ -26,7 +26,11 @@ export async function sendMessageWhatsApp(
     accountId?: string;
   },
 ): Promise<{ messageId: string; toJid: string }> {
-  let text = body;
+  let text = body.trimStart();
+  const jid = toWhatsappJid(to);
+  if (!text && !options.mediaUrl) {
+    return { messageId: "", toJid: jid };
+  }
   const correlationId = generateSecureUuid();
   const startedAt = Date.now();
   const { listener: active, accountId: resolvedAccountId } = requireActiveWebListener(
@@ -51,7 +55,6 @@ export async function sendMessageWhatsApp(
     to: redactedTo,
   });
   try {
-    const jid = toWhatsappJid(to);
     const redactedJid = redactIdentifier(jid);
     let mediaBuffer: Buffer | undefined;
     let mediaType: string | undefined;