which th best RASA NLU or SNIPS NLU?

I would like to know the difference between Snips NLU and Rasa NLU ? which is simpler and powerful ?

From an NLU standpoint, according to the benchmark published by Snips CTO, snips_nlu and rasa_nlu (with Spacy under the hood) have very good and almost identical performance.

As of today Rasa is probably a better choice if you care about any of the following items:

• readiness to use: Rasa core provides a drop-in replacement for common cloud solutions. It takes minutes to spin you own server on localhost, using either a model trained via rasa or downloaded from the service you want to replace.
• language variety: Snips currently supports six languages, whereas Rasa nine. However some languages are only present in Snips, so it really boils down to your current need.

If these items don't make any difference to you, you're in a great spot: my suggestion is to prepare your dataset with Chatito and test by yourself which of the two libraries learns your specific data best.