fix(images): cache Telegram photos by unique_id and pass base64 to LLM commands
Stop reading image files inside gemini/mistral commands; use pre-encoded image data from message parts Rework loadImagesIfExists / loadImagesFromFileIds to return cached file_unique_ids and download only missing sizes Encode cached images to base64 when assembling the reply chain Make getPhotoMaxSize synchronous (returns PhotoSize), and map to URL only when needed Await MessageStore.put / UserStore.put and prefetch single-image downloads on message receipt
This commit is contained in:
@@ -13,7 +13,6 @@ import {
|
||||
oldReplyToMessage,
|
||||
startIntervalEditor
|
||||
} from "../util/utils";
|
||||
import fs from "node:fs";
|
||||
|
||||
export class GeminiChat extends ChatCommand {
|
||||
command = "gemini";
|
||||
@@ -66,10 +65,9 @@ export class GeminiChat extends ChatCommand {
|
||||
const images = messageParts[0].images;
|
||||
|
||||
images.forEach(image => {
|
||||
const base64Image = Buffer.from(fs.readFileSync(image)).toString("base64");
|
||||
input.push({
|
||||
type: "image",
|
||||
data: base64Image,
|
||||
data: image,
|
||||
mime_type: "image/png"
|
||||
});
|
||||
});
|
||||
|
||||
@@ -7,13 +7,11 @@ import {
|
||||
escapeMarkdownV2Text,
|
||||
logError,
|
||||
oldReplyToMessage,
|
||||
photoPathByUniqueId,
|
||||
startIntervalEditor
|
||||
} from "../util/utils";
|
||||
import {Environment} from "../common/environment";
|
||||
import {bot, mistralAi} from "../index";
|
||||
import {MessageStore} from "../common/message-store";
|
||||
import fs from "node:fs";
|
||||
|
||||
export class MistralChat extends ChatCommand {
|
||||
command = "mistral";
|
||||
@@ -46,10 +44,9 @@ export class MistralChat extends ChatCommand {
|
||||
});
|
||||
|
||||
for (const image of part.images) {
|
||||
const base64Image = Buffer.from(fs.readFileSync(photoPathByUniqueId(image))).toString("base64");
|
||||
content.push({
|
||||
type: "image_url",
|
||||
imageUrl: "data:image/jpeg;base64," + base64Image
|
||||
imageUrl: "data:image/jpeg;base64," + image
|
||||
});
|
||||
}
|
||||
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
import {StoredMessage} from "../model/stored-message";
|
||||
import {Message} from "typescript-telegram-bot-api";
|
||||
import {extractTextMessage, isStoredMessage} from "../util/utils";
|
||||
import {extractTextMessage, getPhotoMaxSize, isStoredMessage} from "../util/utils";
|
||||
import {messageDao} from "../index";
|
||||
|
||||
export class MessageStore {
|
||||
@@ -22,6 +22,7 @@ export class MessageStore {
|
||||
fromId: m.from.id,
|
||||
text: extractTextMessage(m),
|
||||
date: m.date ?? 0,
|
||||
photoMaxSizeFilePath: m.photo ? [getPhotoMaxSize(m.photo).file_unique_id] : null
|
||||
};
|
||||
|
||||
this.map.set(this.key(msg.chatId, msg.id), msg);
|
||||
|
||||
@@ -9,7 +9,7 @@ export class UserStore {
|
||||
return this.map;
|
||||
}
|
||||
|
||||
static async put(u: User) {
|
||||
static async put(u: User): Promise<StoredUser> {
|
||||
const user: StoredUser = {
|
||||
id: u.id,
|
||||
isBot: u.is_bot,
|
||||
@@ -22,6 +22,7 @@ export class UserStore {
|
||||
this.map.set(u.id, user);
|
||||
|
||||
await userDao.insert(userDao.mapTo([u]));
|
||||
return user;
|
||||
}
|
||||
|
||||
static async get(id: number): Promise<StoredUser | null> {
|
||||
|
||||
+53
-24
@@ -497,9 +497,18 @@ export async function loadImagesIfExists(msg: Message | StoredMessage): Promise<
|
||||
return msg.photoMaxSizeFilePath;
|
||||
}
|
||||
|
||||
if (!msg.photo?.length) return;
|
||||
|
||||
const imageFilePaths: string[] = [];
|
||||
|
||||
const maxSize = await getPhotoMaxSize(msg.photo);
|
||||
for (const size of msg.photo) {
|
||||
const exists = fs.existsSync(photoPathByUniqueId(size.file_unique_id));
|
||||
if (exists) {
|
||||
return [size.file_unique_id];
|
||||
}
|
||||
}
|
||||
|
||||
const maxSize = await mapPhotoSizeToMax(getPhotoMaxSize(msg.photo));
|
||||
if (maxSize) {
|
||||
const imagePath = path.join(Environment.DATA_PATH, "temp");
|
||||
if (!fs.existsSync(imagePath)) {
|
||||
@@ -527,14 +536,23 @@ export async function loadImagesIfExists(msg: Message | StoredMessage): Promise<
|
||||
return imageFilePaths;
|
||||
}
|
||||
|
||||
export async function loadImagesFromFileIds(maxSizes: PhotoMaxSize[]): Promise<string[] | null> {
|
||||
if (!maxSizes?.length) return null;
|
||||
export async function loadImagesFromFileIds(sizes: PhotoSize[]): Promise<string[] | null> {
|
||||
if (!sizes?.length) return null;
|
||||
|
||||
const dataPath = path.join(Environment.DATA_PATH, "temp");
|
||||
if (!fs.existsSync(dataPath)) {
|
||||
fs.mkdirSync(dataPath);
|
||||
}
|
||||
|
||||
const existing =
|
||||
sizes.filter(s => fs.existsSync(photoPathByUniqueId(s.file_unique_id)))
|
||||
.map(s => s.file_unique_id);
|
||||
|
||||
const promises = sizes.filter(s => !fs.existsSync(photoPathByUniqueId(s.file_unique_id)))
|
||||
.map(s => mapPhotoSizeToMax(s));
|
||||
|
||||
const maxSizes = await Promise.all(promises);
|
||||
|
||||
const imagePromises = maxSizes.map((size) => {
|
||||
return axios.get<ArrayBuffer>(size.url, {responseType: "arraybuffer"});
|
||||
});
|
||||
@@ -542,17 +560,19 @@ export async function loadImagesFromFileIds(maxSizes: PhotoMaxSize[]): Promise<s
|
||||
const responses = await Promise.all(imagePromises);
|
||||
const paths = responses.map((res, index) => {
|
||||
try {
|
||||
const imageFilePath = path.join(dataPath, maxSizes[index].unique_file_id + ".jpg");
|
||||
const uniqueFileId = maxSizes[index].unique_file_id;
|
||||
const imageFilePath = path.join(dataPath, uniqueFileId + ".jpg");
|
||||
const src = Buffer.from(res.data);
|
||||
fs.writeFileSync(imageFilePath, src);
|
||||
return imageFilePath;
|
||||
return uniqueFileId;
|
||||
} catch (e) {
|
||||
logError(e);
|
||||
return null;
|
||||
}
|
||||
});
|
||||
|
||||
return paths.filter(p => p);
|
||||
const finalPaths = paths.filter(p => p);
|
||||
finalPaths.unshift(...existing);
|
||||
return finalPaths;
|
||||
}
|
||||
|
||||
export async function collectReplyChainText(triggerMsg: Message | StoredMessage, limit: number = 40, includeTrigger = true, cutPrefix: boolean = true): Promise<MessagePart[]> {
|
||||
@@ -561,15 +581,20 @@ export async function collectReplyChainText(triggerMsg: Message | StoredMessage,
|
||||
const pushPart = async (msg: Message | StoredMessage, textRequired: boolean = false) => {
|
||||
const rawText = extractTextMessage(msg);
|
||||
const cleanText = cutPrefix ? cutPrefixes(rawText) : rawText;
|
||||
const images = await loadImagesIfExists(msg);
|
||||
const imageNames = await loadImagesIfExists(msg);
|
||||
|
||||
if (!cleanText && textRequired) return;
|
||||
if (!cleanText && !images?.length) return;
|
||||
if (!cleanText && !imageNames?.length) return;
|
||||
|
||||
const fromId = isStoredMessage(msg) ? msg.fromId : msg.from.id;
|
||||
const firstName = isStoredMessage(msg) ?
|
||||
(await UserStore.get(msg.fromId))?.firstName : msg.from.first_name;
|
||||
|
||||
const images = imageNames ? imageNames.map(n => {
|
||||
const filePath = photoPathByUniqueId(n);
|
||||
return Buffer.from(fs.readFileSync(filePath)).toString("base64");
|
||||
}) : null;
|
||||
|
||||
parts.push({
|
||||
bot: fromId === botUser.id,
|
||||
content: cleanText ? cleanText : "",
|
||||
@@ -932,7 +957,7 @@ export function getRuntimeInfo(): RuntimeInfo {
|
||||
|
||||
export type PhotoMaxSize = { width: number, height: number, url: string; file_id: string; unique_file_id: string; };
|
||||
|
||||
export async function getPhotoMaxSize(photos: PhotoSize[], target: number = Environment.MAX_PHOTO_SIZE): Promise<PhotoMaxSize | null> {
|
||||
export function getPhotoMaxSize(photos: PhotoSize[], target: number = Environment.MAX_PHOTO_SIZE): PhotoSize | null {
|
||||
if (!photos) return null;
|
||||
|
||||
photos = photos.filter(p => Math.max(p.width, p.height) <= target);
|
||||
@@ -940,7 +965,7 @@ export async function getPhotoMaxSize(photos: PhotoSize[], target: number = Envi
|
||||
if (photos.length === 0) return null;
|
||||
|
||||
if (photos.length === 1) {
|
||||
return mapPhotoSizeToMax(photos[0]);
|
||||
return photos[0];
|
||||
}
|
||||
|
||||
const max = photos.reduce((prev, cur) => {
|
||||
@@ -949,8 +974,7 @@ export async function getPhotoMaxSize(photos: PhotoSize[], target: number = Envi
|
||||
return cur.width * cur.height > prev.width * prev.height ? cur : prev;
|
||||
}, null);
|
||||
|
||||
if (!max) return null;
|
||||
return mapPhotoSizeToMax(max);
|
||||
return max;
|
||||
}
|
||||
|
||||
export async function mapPhotoSizeToMax(size: PhotoSize): Promise<PhotoMaxSize | null> {
|
||||
@@ -993,14 +1017,21 @@ export async function processNewMessage(msg: Message) {
|
||||
console.log("message", msg);
|
||||
|
||||
let storedMsg: StoredMessage | null = null;
|
||||
Promise.all([
|
||||
MessageStore.put(msg).then(r => {
|
||||
storedMsg = r;
|
||||
console.log("storedMsg", storedMsg);
|
||||
}),
|
||||
UserStore.put(msg.from)
|
||||
]
|
||||
).catch(logError);
|
||||
|
||||
try {
|
||||
const results = await Promise.all([
|
||||
MessageStore.put(msg),
|
||||
UserStore.put(msg.from)
|
||||
]
|
||||
);
|
||||
|
||||
storedMsg = results[0];
|
||||
if (!msg.media_group_id && storedMsg.photoMaxSizeFilePath) {
|
||||
await loadImagesIfExists(msg);
|
||||
}
|
||||
} catch (e) {
|
||||
logError(e);
|
||||
}
|
||||
|
||||
if ((msg.new_chat_members?.length || 0 > 0)) {
|
||||
await bot.sendMessage({chat_id: msg.chat.id, text: randomValue(Environment.ANSWERS.invite)}).catch(logError);
|
||||
@@ -1080,9 +1111,7 @@ async function processAlbum(groupId: string): Promise<string[]> {
|
||||
.map(m => m.photo);
|
||||
|
||||
const allPhotoMaxSizes = await Promise.all(allPhotos.map(photo => getPhotoMaxSize(photo)));
|
||||
const ids = allPhotoMaxSizes.map(p => p.unique_file_id);
|
||||
|
||||
await loadImagesFromFileIds(allPhotoMaxSizes);
|
||||
const ids = await loadImagesFromFileIds(allPhotoMaxSizes);
|
||||
|
||||
console.log(`Received album ${groupId} with ${ids.length} photos.`);
|
||||
console.log("File IDs:", ids);
|
||||
|
||||
Reference in New Issue
Block a user