SlothSort: AI Telegram Folder Assistant Built for Hacktoberfest
A developer created SlothSort for Hacktoberfest 2026, using Gemma 3 to propose Telegram folder structures with required human approval.
Stock photo for illustration only, not from the actual event
- Developed during Hacktoberfest 2026 Week 3 Slow and Steady challenge
- Uses local Gemma 3 model via Ollama to propose chat categories
- Requires manual user review and approval before any Telegram folders are created
- Source code and setup instructions are fully available on GitHub
Managing the sidebar in messaging platforms like Telegram often turns into a tedious chore that people keep postponing. Deciding which chats belong together and manually maintaining folders consumes valuable time. This friction inspired a developer to build a dedicated assistant during the third week of the online Hacktoberfest 2026 event.
Named SlothSort, the project kicked off on Wednesday, October 21. It connects to Telegram via MTProto to gather context from accessible chats, then prompts a local Gemma 3 model to propose suitable folders. The operative word here is "propose." Before any changes happen in Telegram, users can review every assignment, rename folders, or leave chats unassigned.
Stock photo for illustration only, not from the actual event
The technical architecture separates concerns cleanly. Gemma 3 4B running through Ollama serves as the default classification engine, Telethon handles MTProto access, and FastAPI powers the review interface. This division is vital: the AI suggests structures but cannot call the Telegram API or create folders independently. The application validates chat IDs and waits for human confirmation.
Choosing not to let the AI sort everything fully automatically highlights a mature design philosophy in modern AI development. Chat titles can be ambiguous, and recent messages can easily misrepresent a conversation. Positioning the AI as a preliminary filter that handles repetitive heavy lifting—while keeping the human firmly in control of the final decision—prevents costly categorization errors.
Building around the Slow and Steady challenge prompt provided a useful constraint. Instead of making the convenient false promise of fully automated sorting, the creator prioritized control. The review screen allows users to inspect metadata and descriptions, optionally pulling in recent message context, while safety guards prevent unintended deletions via the undo feature.
Ultimately, the project demonstrated that saving effort does not require removing people from the decision loop. SlothSort absorbs the repetitive first pass of organization, leaving the final authority over the Telegram account right where it belongs.
Source: Dev.to
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment