Pizza Brand Telegram Bot

• What it is: Multilingual (ru/uz/en) Telegram e-commerce bot for the Restaurant chain (UZ) pizzeria chain (choparpizza.uz). Menu/cart, delivery/pickup, terminal selection, stop-lists, authorization, reviews, mass broadcasts. Frontend to an external Laravel API (api.choparpizza.uz-style). Same repo owner (ShahzodSalimsakov) and architecture as Restaurant chain (UZ)Bot.
• Type / status / role: bot / maintained / contributor (2–3 developers; the user has 174 of 317 commits, ~55%).
• Active period: 2021-07-31 → 2026-05-18, ~5 years. Alive: 2025–2026 show a fresh wave of features with clean conventional-commits (some — with Claude).

• Languages: JavaScript (Node.js, CommonJS). TypeScript only in devDeps (unused).
• Frameworks/libraries: Telegraf 4.15 (newer than Restaurant chain (UZ)Bot's 4.4; Scenes/WizardScene), telegraf-i18n (ru/uz/en), Fastify 3 + fastify-telegraf, Bull 4 + @bull-board, ioredis, luxon, hashids, bwip-js/canvas (barcodes), axios.
• Infra/deploy: PM2, Sentry, @inspector-apm/inspector-nodejs (APM monitoring — the only notable dependency difference from Restaurant chain (UZ)Bot), choparpizza.uz / Contabo server. Bun (bun.lockb present) + yarn.lock — package manager migration.
• Data: MongoDB (sessions + event_systems collection), Redis (Bull + pub/sub).
• Notable tooling: agenda/node-cron/sfn-scheduler (partly legacy/commented out).

Early history — junk messages ("asd", "dsd", "fix"); from 2025 onward — clean conventional commits (some generated by Claude).

• Menu/cart — controllers/menu/index.js >2755 lines (larger than Restaurant chain (UZ)Bot's): pagination, cart, modifiers, half-products, promotions (threeSaleProducts, 8246873).
• Per-terminal stop-lists — f241cf3 fix: Enforce stop-list check on product variants + b0697e4 (answerCbQuery when blocked).
• Scheduled broadcasts — bot.js + util/queue.js: Bull cron jobs, progress reported to the API, message deletion; cb9c9f3 feat: endpoint to remove scheduled events from MongoDB, 0dac3ce feat: Emit 'removed' event on job removal.
• send_notify with reply — 30928f3 refactor: Add support for replying to a message.
• Business updates: cities (94653c7 trying to fix cities issue), close time, delivery copy.
• Final refactor (`5880f35`, 2026-05-18): replaced /api/configs/public with a static env config (same refactor as Restaurant chain (UZ)Bot — a shared update across both bots).
• Volume: 317 commits over ~5 years, 174 by the user. Activity in waves; recently revived with AI assistance.

• Stop-list at the variant/modifier level (controllers/menu/index.js, addToCart, commit f241cf3): previously only the parent product ID was validated → out-of-stock pizza sizes (40 cm) slipped into the cart. Fix: GET terminals/get_stock?terminal_id=..., build blockedIds from [selectedProdId, rightProduct.id, ...selectedModifiers.map(id)], check intersection with the stop-list, block with productIsNotInStock. → A real production bug in domain logic; correctly extending the check to all product components.
• Removing scheduled messages from MongoDB (bot.js, cb9c9f3): endpoint /api/schedule_event/remove_within_mongo — cursor over event_systems by jobId, bot.telegram.deleteMessage(chatId, message_id) + document deletion, with detailed logging and try/catch per message. → Managing broadcast state (also has a bug: a reference to queueJob.id outside the scope — a potential ReferenceError).
• Custom Bull queue with Redis pub/sub (util/queue.js) — same as in Restaurant chain (UZ)Bot: cross-process cancellation of running jobs (a shared infra pattern of the family).
• APM instrumentation (@inspector-apm/inspector-nodejs) — production performance monitoring of the bot.

• Sessions found: 1 (within the local Claude Code sessions directory, 0f95e959-...jsonl, ~3.2 MB).
• What was done with AI: sample of the first prompt — "commit and push" (minimal manual input); actual AI edits visible in commits with Co-Authored-By: Claude Opus 4.7 (1M context): variant-level stop-list (f241cf3), logging for sending/deleting messages, env-config refactor (5880f35).
• AI-workflow patterns: using Claude Code for targeted bugfixes and refactoring inside a large legacy file (menu/index.js, 2.7k+ lines), with clean conventional-commit messages generated by AI. Delegating routine ("commit and push").
• Unlike alfaclean — here AI was used pointedly on legacy code, not for greenfield architecture.

• Real production bot for the Restaurant chain (UZ) pizzeria chain; 3 languages; integration with Laravel API/MongoDB/Redis/Sentry/APM.
• 174 user commits out of 317 over ~5 years — a long-lived product.
• menu/index.js >2755 lines — the largest controller; rich domain logic (modifiers, half-products, stop-lists, promotions).
• No exact business metrics in the code.