Handle long caption in onPhoto

src/handlers/onPhoto.ts

@@ -6,6 +6,7 @@ import { recognizeImageText } from "../helpers/vision.ts";
 import { log } from "../helpers";
 import { useConfig } from "../config.ts";
 import { sendTelegramMessage } from "../telegram/send.ts";
+import { createNewContext } from "../telegram/context.ts";
 
 // Type guard to check if update has a message
 function isMessageUpdate(update: Update): update is Update.MessageUpdate {
@@ -32,6 +33,26 @@ export default async function onPhoto(ctx: Context) {
     role: "user",
   });
 
+  if (msg.caption && msg.caption.length > 100) {
+    log({
+      msg: `[photo] caption too long, skip ocr: ${msg.caption.length}`,
+      logLevel: "info",
+      chatId: msg.chat.id,
+      chatTitle,
+    });
+
+    const newMsg = {
+      ...msg,
+      text: msg.caption,
+      entities: [],
+    } as const;
+
+    const contextWithCaption = createNewContext(ctx, newMsg);
+
+    await onTextMessage(contextWithCaption);
+    return;
+  }
+
   const config = useConfig();
   const model = config?.vision?.model || "";
   if (!model)
@@ -42,13 +63,13 @@ export default async function onPhoto(ctx: Context) {
 
   // Create a new message object with the recognized text
   const processPhoto = async () => {
-    let text = '';
+    let text = "";
     try {
       text = await recognizeImageText(msg, chat);
     } catch (error) {
       await sendTelegramMessage(
         msg.chat.id,
-        `Ошибка при распознавании текста: ${error instanceof Error ? error.message : 'Неизвестная ошибка'}`
+        `Ошибка при распознавании текста: ${error instanceof Error ? error.message : "Неизвестная ошибка"}`,
       );
     }
     const caption = msg.caption ? `${msg.caption}\n` : "";
@@ -69,15 +90,7 @@ export default async function onPhoto(ctx: Context) {
     } as const;
 
     // Create a new context by extending the original context
-    const contextWithNewMessage = Object.create(Object.getPrototypeOf(ctx), {
-      ...Object.getOwnPropertyDescriptors(ctx),
-      message: { value: newMsg, writable: true, configurable: true },
-      update: {
-        value: { ...ctx.update, message: newMsg },
-        writable: true,
-        configurable: true,
-      },
-    });
+    const contextWithNewMessage = createNewContext(ctx, newMsg);
 
     await onTextMessage(contextWithNewMessage);
   };

src/telegram/context.ts

@@ -142,3 +142,15 @@ export function getCtxChatMsg(ctx: Context): {
 
   return { chat, msg };
 }
+
+export function createNewContext(ctx: Context, newMsg: Message) {
+  return Object.create(Object.getPrototypeOf(ctx), {
+    ...Object.getOwnPropertyDescriptors(ctx),
+    message: { value: newMsg, writable: true, configurable: true },
+    update: {
+      value: { ...ctx.update, message: newMsg },
+      writable: true,
+      configurable: true,
+    },
+  });
+}

tests/handlers/onPhotoMain.test.ts

@@ -73,4 +73,21 @@ describe("onPhoto main flow", () => {
     expect(calledCtx.message.text).toBe("cap\nocr");
     expect(mockSendTelegramMessage).not.toHaveBeenCalled();
   });
+
+  it("skips ocr when caption is long", async () => {
+    const caption = "a".repeat(101);
+    const msg = {
+      chat: { id: 1, type: "private", title: "t" },
+      photo: [{ file_id: "f" }],
+      caption,
+    } as Message.PhotoMessage;
+    mockCheckAccessLevel.mockResolvedValue({ msg, chat: {} as ConfigChatType });
+    mockUseConfig.mockReturnValue({ vision: { model: "m" } });
+    const ctx = createCtx(msg);
+    await onPhoto(ctx);
+    expect(mockRecognizeImageText).not.toHaveBeenCalled();
+    expect(mockOnTextMessage).toHaveBeenCalled();
+    const calledCtx = mockOnTextMessage.mock.calls[0][0];
+    expect(calledCtx.message.text).toBe(caption);
+  });
 });

tests/telegram/context.test.ts

@@ -1,7 +1,8 @@
 import { jest, describe, it, expect, beforeEach } from "@jest/globals";
-import type { Context, Update } from "telegraf";
-import type { Message } from "telegraf/types";
+import type { Context } from "telegraf";
+import type { Message, Update } from "telegraf/types";
 import type { ConfigChatType } from "../../src/types";
+import { createNewContext } from "../../src/telegram/context";
 
 const mockUseConfig = jest.fn();
 const mockLog = jest.fn();
@@ -33,6 +34,14 @@ function createCtx(update?: Partial<Update>, botName = "bot") {
   } as unknown as Context;
 }
 
+function createMsg(username: string): Message.TextMessage {
+  return {
+    chat: { id: 123, type: "private", username },
+    from: { username },
+    text: "hi",
+  } as unknown as Message.TextMessage;
+}
+
 describe("getActionUserMsg", () => {
   it("returns user and message from callback query", () => {
     const ctx = createCtx({
@@ -61,14 +70,6 @@ describe("getCtxChatMsg", () => {
     toolParams: {},
   } as ConfigChatType;
 
-  function createMsg(username: string): Message.TextMessage {
-    return {
-      chat: { id: 123, type: "private", username },
-      from: { username },
-      text: "hi",
-    } as unknown as Message.TextMessage;
-  }
-
   it("returns chat when user allowed", () => {
     mockUseConfig.mockReturnValue({ chats: [baseChat], privateUsers: [] });
     const ctx = createCtx({ message: createMsg("u") });
@@ -114,3 +115,13 @@ describe("getCtxChatMsg", () => {
     expect(chat).toMatchObject(baseChat);
   });
 });
+
+describe("createNewContext", () => {
+  it("creates new context with new message", () => {
+    const ctx = createCtx({ message: createMsg("u") });
+    const newMsg = createMsg("new");
+    const newCtx = createNewContext(ctx, newMsg);
+    expect(newCtx.message).toEqual(newMsg);
+    expect(newCtx.update).toEqual({ message: newMsg });
+  });
+});