Hi all,
Here is a short update about NLPCraft project.

Right now, we have paused active development for some time.
The main reason is fast changes in the AI/LLM space. Now, we are
trying to understand where NLPCraft can fit and how it can still be
useful.

The original idea of NLPCraft was to create and provide simple API for
using deterministic intents — partly similar to tools/functions that
we now see in LLMs (like OpenAI function). When we started, we
compared NLPCraft with Amazon Alexa and Google Dialogflow. (They were
not real competitors, but it helps to show what kind of features we
were working on.)
Today, LLM functions are much easier to configure and use. But they
still have the same problem — they are not deterministic.

Recently, we saw growing interest in this area, driven by the constant
development of agent frameworks and a wave of new startups working on
AI agents. This can also be seen in technical news, like the
appearance and strong interest in the MCP protocol, and the recent
release of the Agent2Agent protocol from Google.

Because of that, we are thinking how NLPCraft can be useful now. We
believe our main strengths are: deterministic behavior, lightweight
architecture, local-first usage. So, we plan to continue the project
after we make a clear decision on the direction.

If you have ideas, suggestions, or experience in this area — please
share. Any feedback is welcome.

Best regards,
Sergey Kamov

Reply via email to