fix: filter telegram binary caption text (#66663) (thanks @joelnishanth)

* Telegram: filter binary content from msg.caption to prevent token explosion (#66647)

When a user sends a binary document (e.g. .mobi, .epub) via Telegram, raw
binary bytes can leak into msg.caption. getTelegramTextParts() passes this
through to the LLM prompt, causing catastrophic token explosion (~460K tokens).

Add isBinaryContent() that detects non-printable control characters (0x00-0x08,
0x0E-0x1F) and use it to sanitize the text in getTelegramTextParts() before it
reaches the prompt pipeline. When binary content is detected, the text and
entities are both replaced with empty values so the message is still processed
(media placeholder still works) but the binary junk is dropped.

Made-with: Cursor

* fix: distill telegram binary caption filtering

* fix: filter telegram binary caption text (#66663) (thanks @joelnishanth)

---------

Co-authored-by: Ayaan Zaidi <hi@obviy.us>

CHANGELOG.md

@@ -16,6 +16,7 @@ Docs: https://docs.openclaw.ai
 - Cron/agents: forward embedded-run tool policy and internal event params into the attempt layer so `--tools` allowlists, cron-owned message-tool suppression, explicit message targeting, and command-path internal events all take effect at runtime again. (#62675) Thanks @hexsprite.
 - Setup/providers: guard preferred-provider lookup during setup so malformed plugin metadata with a missing provider id no longer crashes the wizard with `Cannot read properties of undefined (reading 'trim')`. (#66649) Thanks @Tianworld.
 - Matrix/security: normalize sandboxed profile avatar params, preserve `mxc://` avatar URLs, and surface gmail watcher stop failures during reload. (#64701) Thanks @slepybear.
+- Telegram/documents: drop leaked binary caption bytes from inbound Telegram text handling so document uploads like `.mobi` or `.epub` no longer explode prompt token counts. (#66663) Thanks @joelnishanth.
 
 ## 2026.4.14
 

extensions/telegram/src/bot-handlers.runtime.ts

@@ -212,7 +212,7 @@ export const registerTelegramHandlers = ({
       entry.debounceLane === "forward" ? FORWARD_BURST_DEBOUNCE_MS : debounceMs,
     buildKey: (entry) => entry.debounceKey,
     shouldDebounce: (entry) => {
-      const text = entry.msg.text ?? entry.msg.caption ?? "";
+      const text = getTelegramTextParts(entry.msg).text;
       const hasDebounceableText = shouldDebounceTextInbound({
         text,
         cfg,
@@ -248,7 +248,7 @@ export const registerTelegramHandlers = ({
         return;
       }
       const combinedText = entries
-        .map((entry) => entry.msg.text ?? entry.msg.caption ?? "")
+        .map((entry) => getTelegramTextParts(entry.msg).text)
         .filter(Boolean)
         .join("\n");
       const combinedMedia = entries.flatMap((entry) => entry.allMedia);
@@ -836,7 +836,7 @@ export const registerTelegramHandlers = ({
       // for reactions, we cannot determine if the reaction came from a topic, so block all
       // reactions if requireTopic is enabled for this DM.
       if (!isGroup) {
-        const requireTopic = (eventAuthContext.groupConfig as TelegramDirectConfig | undefined)
+        const requireTopic = (eventAuthContext.groupConfig)
           ?.requireTopic;
         if (requireTopic === true) {
           logVerbose(

extensions/telegram/src/bot/body-helpers.ts

@@ -92,14 +92,25 @@ export function buildSenderLabel(msg: Message, senderId?: number | string) {
 
 export type TelegramTextEntity = NonNullable<Message["entities"]>[number];
 
+export function isBinaryContent(text: string): boolean {
+  for (let i = 0; i < text.length; i++) {
+    const code = text.charCodeAt(i);
+    if (code <= 0x1f && code !== 0x09 && code !== 0x0a && code !== 0x0d) {
+      return true;
+    }
+  }
+  return false;
+}
+
 export function getTelegramTextParts(
   msg: Pick<Message, "text" | "caption" | "entities" | "caption_entities">,
 ): {
   text: string;
   entities: TelegramTextEntity[];
 } {
-  const text = msg.text ?? msg.caption ?? "";
-  const entities = msg.entities ?? msg.caption_entities ?? [];
+  const raw = msg.text ?? msg.caption ?? "";
+  const text = isBinaryContent(raw) ? "" : raw;
+  const entities = text ? (msg.entities ?? msg.caption_entities ?? []) : [];
   return { text, entities };
 }
 

extensions/telegram/src/bot/helpers.test.ts

@@ -7,6 +7,7 @@ import {
   expandTextLinks,
   getTelegramTextParts,
   hasBotMention,
+  isBinaryContent,
   normalizeForwardedContext,
   resolveTelegramDirectPeerId,
   resolveTelegramForumFlag,
@@ -444,6 +445,67 @@ describe("describeReplyTarget", () => {
   });
 });
 
+describe("isBinaryContent", () => {
+  it("returns false for normal user text", () => {
+    expect(isBinaryContent("Hello, world!")).toBe(false);
+  });
+
+  it("returns false for text with common whitespace (tabs, newlines)", () => {
+    expect(isBinaryContent("line one\nline two\ttab")).toBe(false);
+  });
+
+  it("returns true for string containing null bytes", () => {
+    expect(isBinaryContent("PK\x00\x03\x04")).toBe(true);
+  });
+
+  it("returns true for typical binary file header bytes", () => {
+    const mobiBinarySnippet = "\x00\x00\x00\x01BOOKMOBI\x00\x00\x02\x0E";
+    expect(isBinaryContent(mobiBinarySnippet)).toBe(true);
+  });
+
+  it("returns false for empty string", () => {
+    expect(isBinaryContent("")).toBe(false);
+  });
+});
+
+describe("getTelegramTextParts — binary caption filtering (#66647)", () => {
+  it("strips binary caption content to prevent token explosion", () => {
+    const binaryCaption = "PK\x03\x04\x14\x00\x08binary-ebook-data";
+    const result = getTelegramTextParts({
+      caption: binaryCaption,
+      caption_entities: [{ type: "mention", offset: 0, length: 5 }],
+      chat: { id: 1, type: "private" },
+      date: 1,
+      message_id: 1,
+    } as any);
+    expect(result.text).toBe("");
+    expect(result.entities).toEqual([]);
+  });
+
+  it("preserves normal caption text", () => {
+    const result = getTelegramTextParts({
+      caption: "Here is my document",
+      caption_entities: [],
+      chat: { id: 1, type: "private" },
+      date: 1,
+      message_id: 1,
+    } as any);
+    expect(result.text).toBe("Here is my document");
+  });
+
+  it("strips binary content in msg.text as well", () => {
+    const result = getTelegramTextParts({
+      text: "\x00\x01\x02 binary junk",
+      entities: [{ type: "bold", offset: 0, length: 3 }],
+      chat: { id: 1, type: "private" },
+      date: 1,
+      message_id: 1,
+    } as any);
+    expect(result.text).toBe("");
+    expect(result.entities).toEqual([]);
+  });
+});
+
 describe("hasBotMention", () => {
   it("prefers caption text and caption entities when message text is absent", () => {
     expect(

extensions/telegram/src/bot/helpers.ts

@@ -18,6 +18,7 @@ import {
   extractTelegramLocation,
   getTelegramTextParts,
   hasBotMention,
+  isBinaryContent,
   normalizeForwardedContext,
   resolveTelegramMediaPlaceholder,
   type TelegramForwardedContext,
@@ -32,6 +33,7 @@ export {
   extractTelegramLocation,
   getTelegramTextParts,
   hasBotMention,
+  isBinaryContent,
   normalizeForwardedContext,
   resolveTelegramMediaPlaceholder,
 };