Tool · Active t.me/@befeedbackbot

Telegram Feedback Bot

A Telegram bot + web dashboard for collecting structured player feedback at scale. Bot intake with privacy-preserving pseudonyms → searchable web dashboard with filtering → AI-generated daily and weekly summaries surfacing sentiment shifts and recurring issues. Directly feeds product backlog prioritization.

Stack: Node.js (Telegraf) · FastAPI · PostgreSQL · Redis AI: GPT-4o-mini (daily summaries + ad-hoc analysis) Integrations: Telegram · Slack · Google Sheets
Directly feeds product backlog prioritization — the team uses it to decide what to fix and what to build next

Feedback Collection

  • Structured categories: Ideas, Fixes & Improvements, Questions/AMA, and special limited-time categories.
  • Multilingual support: Russian, English, Ukrainian, Polish — with language selection per user.
  • Privacy-preserving pseudonym system (~10,700 unique combinations like "lazy_duck", "magic_dolphin") — admins never see real Telegram IDs.
  • Anonymity choice: each user decides whether to appear as anonymous or by @username.
  • Daily limits (4 feedbacks/day), 30-minute edit window, text validation (15–3200 chars).

Web Dashboard

  • Main dashboard with stat cards (total, ideas, fixes, today count) and real-time feedback list.
  • Filtering by category, date range, user pseudonym, and text search — with pagination.
  • Dedicated AMA page for questions category.
  • AI Daily Summary section with bilingual toggle (EN/RU).
  • Summary history archive: browse all generated summaries by date and time slot.
  • Per-feedback moderation: ban user or delete entry directly from the dashboard.

AI-Powered Analytics

  • Automatic daily summaries generated at 09:00, 13:00, 17:00 UTC — bilingual (EN + RU).
  • Weekly summaries sent to Slack with threaded discussion.
  • Ad-hoc analysis: ask arbitrary questions about feedback and get sourced answers referencing specific feedback IDs.
  • Strict prompt engineering: every claim must cite a real feedback entry — no hallucination.
  • Supports filtered analysis: narrow by category, date range, user, or text before asking.

Moderation & Admin

  • Ban users by pseudonym or Telegram ID — without revealing real identities.
  • Blacklist management: web UI and bot commands (/ban, /unban, /blacklist).
  • Ban reason tracking and persistence across PostgreSQL + Redis fallback.
  • Web blacklist page with one-click unban.

Architecture

Three-layer system: Bot layer (Node.js/Telegraf) handles Telegram message intake, command processing, inline keyboards, and anonymity/pseudonym management. Web layer (Python/FastAPI with Jinja2 templates) provides the dashboard, REST API, scheduled AI summary generation, and moderation actions. Data layer (PostgreSQL + Redis) stores feedback, user mappings, blacklist, and daily summaries — with Redis caching state, daily counters, and pending edits.

Slack integration sends real-time feedback notifications (Block Kit formatting) and weekly AI summary threads. Optional Google Sheets backup for feedback log and user mapping. Containerized with Docker Compose.

Operational impact

Before this tool, player feedback arrived unstructured through Discord, social media, and support tickets — requiring manual triage to surface patterns. The Telegram Feedback Bot structures intake with categories and pseudonyms, surfaces sentiment shifts through AI-generated daily summaries, and makes the full archive searchable.

The product team uses the AI summaries and the searchable dashboard directly for backlog prioritization — deciding what to fix, what to build next, and which player pain points are growing. Weekly Slack summaries keep stakeholders aligned without manual report writing.