Hello,

While working with Kaldi for speech recognition, we found relatively few serving solutions, most of which haven’t worked out nicely for us so we have been building a simple server here.

Although, the program should not be considered stable, is missing many features we want (like supporting on-the-fly HLCG creation for providing speech contexts like in gspeech) and we are going relatively slow on it, it can be used without unreasonable effort if you have regular trained chain models.

Feel free to suggest changes, improvements, PRs etc.