search

Kitt-AI / snowboy

Future versions with model training module will be maintained through a forked version here: https://github.com/seasalt-ai/snowboy
Other
3.07k stars 998 forks source link

Missing API Documentation/Closed Source Problem #382

Open NicoHood opened 6 years ago

NicoHood commented 6 years ago

I am trying to understand all the options and parameters of the python wrapper. However at some point it seems to call a closed source .so file which I cannot either debug nor read to understand. There are no docs except setting up a basic blink for a raspberry pi, which I am not interested in. Can you please provide the source of this API or either document it please? Otherwise I have to blindly guess which may result in unexpected behaviour.

chenguoguo commented 6 years ago

For APIs, the header file is actually self explaining: https://github.com/Kitt-AI/snowboy/blob/master/include/snowboy-detect.h

There aren't a lot of APIs that have to be modified by common users. The sensitivity, and the recently added frontend option, and that's pretty much it. You can follow examples under examples.

NicoHood commented 6 years ago

I also found the include file, which gives a bit more details. It looks like this is written with doxygen, maybe you can publish it along with the other docs?

It would be nice if you can clarify what frontend means: https://github.com/Kitt-AI/snowboy/issues/380

One of the biggest advantages you can get from opening the source is that people can check and patch your code. For example I would simply add this feature https://github.com/Kitt-AI/snowboy/issues/384 myself, but I am unable to do, as its part of the closed source library. It could safe you lots of time and more people can review new code :)

chenguoguo commented 6 years ago

Added the explanation for frontend audio processing.

I didn't want to expose those details because I felt that was too much for a normal developer. And for hard-core developers, they would still be able to find the explanations in the header file.

NicoHood commented 6 years ago

It is much better now. Thanks for clarifying.

I still dont like the idea of closing the source. I mean you could - at your option - still close the source of the website and the model generation. But the source of the recognizing could be open. Maybe someone then feels free to port it to windows.

And for the commercial aspect people still need to use your service via the website and pay if they need better models. It would help you to possibly get more community support and bugfixes which results in a better software with more users. :)

chenguoguo commented 6 years ago

Don't take it wrong, I'm pro-opensourcing, but there are a lot of other issues that we have to consider.

Perhaps one day I'll build purely open source speech recognition tools for developers (not for researchers).