SlothSort: Gemma proposes Telegram folders, I approve the changes

Description and use

SlothSort uses a local Gemma 3 4B model to automatically propose Telegram chat folders, presenting them for user approval via a web UI, thereby reducing manual folder management.

Problem

Manually sorting Telegram chats into folders is repetitive and time‑consuming.

Outcome

SlothSort reduces manual folder decisions by proposing and applying folder assignments only after user approval, leveraging a local Gemma model to avoid sending private messages to external APIs.

Workflow

• Read accessible Telegram dialogs via Telethon/MTProto
• Send dialog metadata (title, description, optional recent text samples) to local Gemma 3 4B via Ollama
• Receive structured JSON proposals with reasons and confidence scores
• Validate proposals against current Telegram state and existing folders
• Present review screen to user
• On user approval, create or update Telegram folders via MTProto
• Allow undo of the latest creation if folders remain unchanged

Evidence

It reads the Telegram dialogs my account can access, asks a local Gemma 3 model to propose folders, and shows a review screen before changing anything.

Source

Only my explicit approval creates Telegram's native chat folders through MTProto.

Source

With local Ollama, inference stays on my machine.

Source