fix(*): images do work

backend/src/bot/handlers/message/handler.py

@@ -1,4 +1,5 @@
 import asyncio
+import base64
 import contextlib
 import io
 import time
@@ -26,6 +27,14 @@ EDIT_THROTTLE_SECONDS = 1.0
 TELEGRAM_MAX_LENGTH = 4096
 
 
+async def fetch_chat_images(chat_id: str) -> list[ImageData]:
+    chat_images = await convex.query("messages:getChatImages", {"chatId": chat_id})
+    return [
+        ImageData(data=base64.b64decode(img["base64"]), media_type=img["mediaType"])
+        for img in (chat_images or [])
+    ]
+
+
 def make_follow_up_keyboard(options: list[str]) -> ReplyKeyboardMarkup:
     buttons = [[KeyboardButton(text=opt)] for opt in options]
     return ReplyKeyboardMarkup(
@@ -114,7 +123,7 @@ async def send_long_message(
         )
 
 
-async def process_message_from_web(  # noqa: C901, PLR0915
+async def process_message_from_web(  # noqa: C901, PLR0912, PLR0915
     convex_user_id: str, text: str, bot: Bot, convex_chat_id: str
 ) -> None:
     user = await convex.query("users:getById", {"userId": convex_user_id})
@@ -178,7 +187,11 @@ async def process_message_from_web(  # noqa: C901, PLR0915
             prompt_text = text
             hist = history[:-1]
 
-        final_answer = await stream_response(text_agent, prompt_text, hist, on_chunk)
+        chat_images = await fetch_chat_images(convex_chat_id)
+
+        final_answer = await stream_response(
+            text_agent, prompt_text, hist, on_chunk, images=chat_images
+        )
 
         if state:
             await state.flush()
@@ -204,6 +217,21 @@ async def process_message_from_web(  # noqa: C901, PLR0915
             },
         )
 
+        if is_summarize:
+            await convex.mutation(
+                "chats:clear", {"chatId": convex_chat_id, "preserveImages": True}
+            )
+            await convex.mutation(
+                "messages:create",
+                {
+                    "chatId": convex_chat_id,
+                    "role": "assistant",
+                    "content": final_answer,
+                    "source": "web",
+                    "followUpOptions": follow_ups,
+                },
+            )
+
         if tg_chat_id and processing_msg:
             with contextlib.suppress(Exception):
                 await processing_msg.delete()
@@ -229,7 +257,7 @@ async def process_message_from_web(  # noqa: C901, PLR0915
 
 
 async def process_message(
-    user_id: int, text: str, bot: Bot, chat_id: int, image: ImageData | None = None
+    user_id: int, text: str, bot: Bot, chat_id: int, *, skip_user_message: bool = False
 ) -> None:
     user = await convex.query(
         "users:getByTelegramId", {"telegramId": ConvexInt64(user_id)}
@@ -251,15 +279,16 @@ async def process_message(
     api_key = user["geminiApiKey"]
     model_name = user.get("model", "gemini-3-pro-preview")
 
-    await convex.mutation(
-        "messages:create",
-        {
-            "chatId": active_chat_id,
-            "role": "user",
-            "content": text,
-            "source": "telegram",
-        },
-    )
+    if not skip_user_message:
+        await convex.mutation(
+            "messages:create",
+            {
+                "chatId": active_chat_id,
+                "role": "user",
+                "content": text,
+                "source": "telegram",
+            },
+        )
 
     assistant_message_id = await convex.mutation(
         "messages:create",
@@ -293,8 +322,10 @@ async def process_message(
                 {"messageId": assistant_message_id, "content": content},
             )
 
+        chat_images = await fetch_chat_images(active_chat_id)
+
         final_answer = await stream_response(
-            text_agent, text, history[:-2], on_chunk, image=image
+            text_agent, text, history[:-2], on_chunk, images=chat_images
         )
 
         await state.flush()
@@ -305,7 +336,7 @@ async def process_message(
         follow_up_agent = create_follow_up_agent(
             api_key=api_key, model_name=follow_up_model, system_prompt=follow_up_prompt
         )
-        follow_ups = await get_follow_ups(follow_up_agent, full_history, image=image)
+        follow_ups = await get_follow_ups(follow_up_agent, full_history)
 
         await state.stop_typing()
 
@@ -380,6 +411,14 @@ async def on_photo_message(message: types.Message, bot: Bot) -> None:
         },
     )
 
+    user = await convex.query(
+        "users:getByTelegramId", {"telegramId": ConvexInt64(message.from_user.id)}
+    )
+
+    if not user or not user.get("activeChatId"):
+        await message.answer("Use /new first to create a chat.")
+        return
+
     caption = message.caption or "Process the image according to your task"
     photo = message.photo[-1]
 
@@ -391,11 +430,24 @@ async def on_photo_message(message: types.Message, bot: Bot) -> None:
     buffer = io.BytesIO()
     await bot.download_file(file.file_path, buffer)
     image_bytes = buffer.getvalue()
+    image_base64 = base64.b64encode(image_bytes).decode()
 
     ext = file.file_path.rsplit(".", 1)[-1].lower()
     media_type = f"image/{ext}" if ext in ("png", "gif", "webp") else "image/jpeg"
-    image = ImageData(data=image_bytes, media_type=media_type)
+
+    active_chat_id = user["activeChatId"]
+    await convex.mutation(
+        "messages:create",
+        {
+            "chatId": active_chat_id,
+            "role": "user",
+            "content": caption,
+            "source": "telegram",
+            "imageBase64": image_base64,
+            "imageMediaType": media_type,
+        },
+    )
 
     await process_message(
-        message.from_user.id, caption, bot, message.chat.id, image=image
+        message.from_user.id, caption, bot, message.chat.id, skip_user_message=True
     )

frontend/src/lib/convex/_generated/api.d.ts

@@ -9,7 +9,6 @@
  */
 
 import type * as chats from "../chats.js";
-import type * as http from "../http.js";
 import type * as messages from "../messages.js";
 import type * as pendingGenerations from "../pendingGenerations.js";
 import type * as users from "../users.js";
@@ -22,7 +21,6 @@ import type {
 
 declare const fullApi: ApiFromModules<{
   chats: typeof chats;
-  http: typeof http;
   messages: typeof messages;
   pendingGenerations: typeof pendingGenerations;
   users: typeof users;

frontend/src/lib/convex/chats.ts

@@ -43,7 +43,7 @@ export const clear = mutation({
 			.collect();
 
 		for (const message of messages) {
-			if (args.preserveImages && message.imageStorageId) {
+			if (args.preserveImages && message.imageBase64) {
 				continue;
 			}
 			await ctx.db.delete(message._id);

frontend/src/lib/convex/http.ts

@@ -1,40 +0,0 @@
-import { httpRouter } from 'convex/server';
-import { httpAction } from './_generated/server';
-import { internal } from './_generated/api';
-import type { Id } from './_generated/dataModel';
-
-const http = httpRouter();
-
-http.route({
-	path: '/upload-image',
-	method: 'POST',
-	handler: httpAction(async (ctx, req) => {
-		const chatId = req.headers.get('X-Chat-Id');
-		const mediaType = req.headers.get('Content-Type') || 'image/jpeg';
-		const caption = req.headers.get('X-Caption') || '';
-
-		if (!chatId) {
-			return new Response(JSON.stringify({ error: 'Missing X-Chat-Id header' }), {
-				status: 400,
-				headers: { 'Content-Type': 'application/json' }
-			});
-		}
-
-		const blob = await req.blob();
-		const storageId = await ctx.storage.store(blob);
-
-		await ctx.runMutation(internal.messages.createWithImage, {
-			chatId: chatId as Id<'chats'>,
-			content: caption,
-			imageStorageId: storageId,
-			imageMediaType: mediaType
-		});
-
-		return new Response(JSON.stringify({ storageId }), {
-			status: 200,
-			headers: { 'Content-Type': 'application/json' }
-		});
-	})
-});
-
-export default http;

frontend/src/lib/convex/messages.ts

@@ -1,5 +1,5 @@
 import { v } from 'convex/values';
-import { internalMutation, mutation, query } from './_generated/server';
+import { mutation, query } from './_generated/server';
 
 export const listByChat = query({
 	args: { chatId: v.id('chats') },
@@ -10,7 +10,7 @@ export const listByChat = query({
 			chatId: v.id('chats'),
 			role: v.union(v.literal('user'), v.literal('assistant')),
 			content: v.string(),
-			imageStorageId: v.optional(v.id('_storage')),
+			imageBase64: v.optional(v.string()),
 			imageMediaType: v.optional(v.string()),
 			followUpOptions: v.optional(v.array(v.string())),
 			source: v.union(v.literal('telegram'), v.literal('web')),
@@ -33,7 +33,7 @@ export const create = mutation({
 		role: v.union(v.literal('user'), v.literal('assistant')),
 		content: v.string(),
 		source: v.union(v.literal('telegram'), v.literal('web')),
-		imageStorageId: v.optional(v.id('_storage')),
+		imageBase64: v.optional(v.string()),
 		imageMediaType: v.optional(v.string()),
 		followUpOptions: v.optional(v.array(v.string())),
 		isStreaming: v.optional(v.boolean())
@@ -45,7 +45,7 @@ export const create = mutation({
 			role: args.role,
 			content: args.content,
 			source: args.source,
-			imageStorageId: args.imageStorageId,
+			imageBase64: args.imageBase64,
 			imageMediaType: args.imageMediaType,
 			followUpOptions: args.followUpOptions,
 			createdAt: Date.now(),
@@ -132,7 +132,7 @@ export const getLastAssistantMessage = query({
 			chatId: v.id('chats'),
 			role: v.union(v.literal('user'), v.literal('assistant')),
 			content: v.string(),
-			imageStorageId: v.optional(v.id('_storage')),
+			imageBase64: v.optional(v.string()),
 			imageMediaType: v.optional(v.string()),
 			followUpOptions: v.optional(v.array(v.string())),
 			source: v.union(v.literal('telegram'), v.literal('web')),
@@ -152,21 +152,12 @@ export const getLastAssistantMessage = query({
 	}
 });
 
-export const generateUploadUrl = mutation({
-	args: {},
-	returns: v.string(),
-	handler: async (ctx) => {
-		return await ctx.storage.generateUploadUrl();
-	}
-});
-
-export const getImageUrls = query({
+export const getChatImages = query({
 	args: { chatId: v.id('chats') },
 	returns: v.array(
 		v.object({
-			storageId: v.id('_storage'),
-			mediaType: v.string(),
-			url: v.union(v.string(), v.null())
+			base64: v.string(),
+			mediaType: v.string()
 		})
 	),
 	handler: async (ctx, args) => {
@@ -175,41 +166,11 @@ export const getImageUrls = query({
 			.withIndex('by_chat_id', (q) => q.eq('chatId', args.chatId))
 			.collect();
 
-		const imageMessages = messages.filter((m) => m.imageStorageId && m.imageMediaType);
-		const results = [];
-
-		for (const msg of imageMessages) {
-			if (msg.imageStorageId && msg.imageMediaType) {
-				const url = await ctx.storage.getUrl(msg.imageStorageId);
-				results.push({
-					storageId: msg.imageStorageId,
-					mediaType: msg.imageMediaType,
-					url
-				});
-			}
-		}
-
-		return results;
-	}
-});
-
-export const createWithImage = internalMutation({
-	args: {
-		chatId: v.id('chats'),
-		content: v.string(),
-		imageStorageId: v.id('_storage'),
-		imageMediaType: v.string()
-	},
-	returns: v.id('messages'),
-	handler: async (ctx, args) => {
-		return await ctx.db.insert('messages', {
-			chatId: args.chatId,
-			role: 'user' as const,
-			content: args.content,
-			source: 'telegram' as const,
-			imageStorageId: args.imageStorageId,
-			imageMediaType: args.imageMediaType,
-			createdAt: Date.now()
-		});
+		return messages
+			.filter((m) => m.imageBase64 && m.imageMediaType)
+			.map((m) => ({
+				base64: m.imageBase64!,
+				mediaType: m.imageMediaType!
+			}));
 	}
 });

frontend/src/lib/convex/schema.ts

@@ -23,8 +23,9 @@ export default defineSchema({
 		chatId: v.id('chats'),
 		role: v.union(v.literal('user'), v.literal('assistant')),
 		content: v.string(),
-		imageStorageId: v.optional(v.id('_storage')),
+		imageBase64: v.optional(v.string()),
 		imageMediaType: v.optional(v.string()),
+		imageStorageId: v.optional(v.id('_storage')),
 		followUpOptions: v.optional(v.array(v.string())),
 		source: v.union(v.literal('telegram'), v.literal('web')),
 		createdAt: v.number(),