Gemini CLI server setup and Telegram bridge

This tutorial sets up Gemini CLI as a coding assistant on a server, then connects it to a Telegram bot. The speaker compares it with Claude Code and Codex, emphasizing free access and claiming a two-million-token context window. Those claims describe the service at the time of the recording, rather than guaranteed current terms.

The walkthrough uses 12 prepared Markdown files to guide installation, Google account authorization, autostart and project folder creation. One account fails to respond; switching to the speaker's main account succeeds. He suspects a regional restriction, but does not establish the cause. The working setup uses account access rather than a Gemini API key.

The Telegram bridge needs a bot token and user identifier. Later steps configure voice messages, photos and documents, followed by memory layers using SQLite, agent memory and Obsidian. The speaker also adds context compression, a task scheduler and a command menu. When commands fail to appear in Telegram, he asks the agent to correct the menu configuration.

Optional external tools receive a brief explanation, but the speaker skips connecting them in this demonstration. A final agent self-check reports an active scheduler and no critical log errors. The setup places the agent process on the speaker's server; the tutorial does not demonstrate how to run models locally.