Find a file
Repository files (latest commit first)
Filename Latest commit message Latest commit date
2026-06-09 19:41:24 -04:00
docs/superpowers/specs init 2026-06-09 19:41:24 -04:00
.gitignore init 2026-06-09 19:41:24 -04:00
LICENSE init 2026-06-09 19:41:24 -04:00
README.md init 2026-06-09 19:41:24 -04:00

Wazz That

Wazz That is an experimental Rust project for a local WhatsApp room translator on macOS. The intended workflow is a quiet assistant that watches a configured WhatsApp chatroom, waits until someone asks for translation, and then uses a local Qwen GGUF model to reply to exact messages with natural, context-aware translations.

The goal is not a machine-like sentence translator. The goal is a helpful bilingual participant that preserves tone, intent, names, and room context while staying concise.

Status

This repository is in the design and early implementation stage.

Current contents:

  • rayan/: local no-std-oriented Rust model/runtime source tree.
  • docs/superpowers/specs/2026-06-09-whatsapp-qwen-room-translator-design.md: approved design for the WhatsApp Qwen room translator.

The WhatsApp watcher, GGUF runtime wrapper, and quoted-reply sender still need to be implemented.

Intended Behavior

  1. Start the program locally on a Mac.
  2. Authenticate to WhatsApp Web with QR or pair-code login.
  3. Watch only configured chatrooms.
  4. Stay silent until someone says a trigger phrase such as translate, qwen translate, or translate please.
  5. Translate new messages while translation mode is active.
  6. Reply directly to the exact source message using WhatsApp quoted replies.
  7. Stop when someone says a stop phrase such as stop translate, or when the room times out from inactivity.

Example reply:

Qwen for Maria:
He means he will be there in about ten minutes, but traffic may make him late.

Planned Architecture

  • whatsapp_watcher: WhatsApp Web connection, login, room event stream, and message normalization.
  • room_state: trigger detection, stop detection, active/inactive state, duplicate suppression, and timeout handling.
  • context_buffer: small in-memory recent-message window per room.
  • qwen_translator: local GGUF model wrapper and prompt builder.
  • reply_sender: quoted replies back to exact source messages.
  • config: model path, watched room allowlist, target language, triggers, stop phrases, context size, token limits, and dry-run mode.

The first model target is Qwen/Qwen2.5-0.5B-Instruct-GGUF. A newer candidate is Qwen/Qwen3-0.6B-GGUF.

Safety, Privacy, And Platform Notice

This project is for educational research and local experimentation only.

It is not affiliated with, endorsed by, or sponsored by WhatsApp, Meta, Qwen, Alibaba Cloud, Hugging Face, or the authors of whatsapp-rust.

The planned WhatsApp integration depends on an unofficial WhatsApp Web client. Using unofficial clients may violate WhatsApp or Meta terms and may lead to account restrictions, suspension, data loss, or other consequences. Use only with accounts and rooms where you have permission from the participants.

The first implementation should run locally by default and should not send room content to a cloud model API.

Do not use this project to spam, scrape, surveil, impersonate, harass, bypass platform controls, or process private communications without consent.

License

This repository is distributed under a custom educational-use, no-liability license. See LICENSE.

This is not an open-source license in the OSI sense because it restricts use to educational and research purposes.