n8n Telegram AI Workflow: Gemini Media Processing + Postgres

Turn Telegram media groups into intelligent, database-backed AI replies

This n8n Telegram AI workflow connects Gemini media processing with PostgreSQL to handle everything from text and documents to voice, video, and mixed media—while solving Telegram’s tricky media group timing and formatting challenges.

What this workflow does

Designed as an AI-powered Telegram assistant, the workflow can receive user messages and media, then process them into one coherent response—even when multiple files are sent together.

• Multi-file media group management using PostgreSQL tables: media_group, media_queue, and chat_histories.
• Document parsing for: CSV, HTML, ICS, JSON, ODS, PDF, RTF, TXT, XML, and spreadsheet formats (with an AI fallback when needed for PDF).
• Voice & video transcription so content can be analyzed and included in AI responses.
• Image, audio, and video description to provide richer context for Gemini-powered replies.
• Telegram-safe MarkdownV2 formatting, including auto-splitting for messages exceeding 4096 characters.
• Error fallback for unsupported file types.

Use cases

• Provide AI support for users who send documents and multiple attachments in one Telegram media group.
• Process customer-submitted voice notes and videos into actionable AI summaries.
• Run a lightweight Telegram AI assistant for SaaS operations where chat history and media grouping must be reliable.

Technical details

• PostgreSQL-driven grouping and storage via media_group, media_queue, and chat_histories.
• Workflow logic and transformations using: if, set, code, html, wait, and merge.

Inspired by: Ezema Gingsley Chibuzo’s initial concept for a multi-modal Telegram support bot (n8n workflow link provided in the source).