Own project, open-source Telegram client
Enloq
Captured from the real phone-width web interface using authored synthetic conversations and prepared outputs.
A Telegram client that softens hostile messages while keeping every fact, suggests complete replies, and turns group chats into briefings linked to the original messages.
Enloq comes from en (縁, human connection) and loqui (Latin, "to speak").
In short
- Ask this group answers a member's question from that group's own messages, with retrieval-augmented generation (RAG). Its citations are checked, and it says when the messages don't settle a question.
- Published under AGPL-3.0.
- Runs as my personal deployment, shared with a few approved members.
What I built.
I built the Telegram integration, Python service, web and mobile clients, softening of hostile messages, reply suggestions, voice transcription, group briefings, Ask this group and spending controls.
The engineering decision.
Ask a decision model before calling a writer: Perplexity's Decider answers named questions with probabilities and writes no text, and decides whether a block is softened and whether suggestions are prepared. Let a chosen suggestion change only the editable draft, keep Send separate and the originals available, and make uncertain delivery and model spending visible.
The result.
The public repository includes the implementation, design decisions, tests, an isolated preview with authored data, and the decision model's measured latency and cost.
Evidence and scope
Source and verification scope
- Python / FastAPI
- React / React Native
- Telethon / SQLite
The screenshots demonstrate the interface; they do not measure model quality or establish live Telegram delivery.
The system uses a VPS service, Telegram integration and external model routes. The repository documents this personal deployment; it does not provide a supported self-hosting installation.
Backend and web fixture checks, native-device acceptance, live-provider behavior and deployment are separate evidence. Publishing the source does not establish an app-store release or broad user adoption.