phialsbasement/donotbeaviber — explained in plain English
Analysis updated 2026-05-18
Convert a single described logic step into working code using your own Claude Code login.
Split an overly broad request into separate single-step sentences to translate one at a time.
Ask a conceptual question about two code approaches without letting the AI choose for you.
Review a local log of every sentence to code pair you have accepted.
| phialsbasement/donotbeaviber | 0xallam/my-recipe | 0xhassaan/nn-from-scratch | |
|---|---|---|---|
| Stars | 0 | — | 0 |
| Language | Python | Python | Python |
| Last pushed | — | 2022-11-22 | — |
| Maintenance | — | Dormant | — |
| Setup difficulty | easy | moderate | moderate |
| Complexity | 2/5 | 2/5 | 4/5 |
| Audience | developer | general | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires Claude Code already installed and logged in, plus Python 3.10 or later on your PATH.
Don't Be a Viber is a VS Code extension that turns one plain English sentence into at most one small unit of code, using your own existing Claude Code installation to do the translation. You still do the actual thinking, such as breaking a problem into steps, choosing control flow, and picking data structures, while the AI only converts your already decided logic into working syntax. Ask for more than one logical step at once and the tool refuses, splitting your request into separate single step sentences you can work through one at a time. The author frames this as an alternative to vibe coding, where a person hands off the thinking itself to an AI tool. Every code suggestion still requires you to press a key to trigger it and manually accept or discard the result as ghost text in your editor. Every accepted pair of sentence and code is saved to a local log file you can review later, though this log never leaves your machine. To use it, you need Claude Code already installed and logged in, along with Python 3.10 or later to run a small local backend, though no extra Python packages are required. You write a sentence describing one step, press a keyboard shortcut, and the generated code appears below it as a suggestion you can accept or reject. A separate shortcut lets you ask conceptual questions about code, such as which of two functions better fits your intent, though it refuses to make broader design decisions for you. Settings let you choose the underlying Claude model and effort level, control whether your sentence stays as a comment after accepting code, cap how many lines a generation may produce, and set how long an idle session stays warm. Internally, a local daemon keeps a warm Claude Code session open per project, using a locked system prompt to decide what counts as one logical unit versus a broader request. That session can only read your files for context and cannot edit files, run commands, or reach the network on its own. The project is organized into a backend folder for the daemon, an extension folder for the VS Code integration, and a tools folder for packaging. It includes test scripts for checking the boundary between accepted and rejected requests, and the README asks contributors to report cases where that boundary judges a request incorrectly.
A VS Code extension that turns one sentence into at most one small code unit using your own Claude Code login, refusing bigger requests.
Mainly Python. The stack also includes Python, VS Code Extension, Claude Code.
No license information is stated in the README.
Setup difficulty is rated easy, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.