Use case
Voice dictation for developers
OSSFlow pastes dictated text into the Android field you focused: an issue, a doc, a commit message, or a chat box. It is not a code-aware editor and it does not understand a repository. Treat it as a way to speak prose, then fix identifiers by hand.
Where it helps
Issue bodies, pull request notes, README sentences, and Slack or email replies are ordinary text fields. Focus one and use the bubble. The prompt guide is the same idea inside an assistant's input.
Symbols, file paths, and camel-case names are where speech models slip. Put the ones you repeat into the personal dictionary. Cleanup must not be asked to "write the patch". Its job is to return the sentence you meant.
What it will not do
- It will not open Android Studio or a terminal.
- It will not read the repo or suggest a diff.
- It will not work offline. Audio goes to the provider you set.
- It will not replace a keyboard. The keyboard stays up under the bubble.
Keys
Grok is the default. A Groq or OpenAI key uses the OpenAI-compatible mode, documented in the API key guide. The app is free, bring your own key. There is no account.
Questions
- Can I dictate a commit message on my phone?
- Yes, if the git client or the hosting app has a text field. OSSFlow fills that field. You still review the message before you commit.
- Will it spell function names?
- Only as well as the model and your dictionary allow. Check them. Do not expect a speech model to preserve punctuation-heavy code.
- Is the source available?
- Yes. OSSFlow is GPL-3.0 only. The public source is linked from the site footer. The speech model is whichever API you configure, and that model is not part of the GPL code.
- Does it run prompts I dictate?
- No. It pastes them. Submit the prompt in the assistant app yourself.