When starting project? I would never use NoSQL right now, because in most cases it makes no sense and limits your design choices. You can fit document DB (like MongoDB) by using JSON(B), you can have KV store by having table in form of {key, value}, you can have graph DB by using some CTEs and 2 tables. Additionally Postgres + extensions provide you with more functionalities:

• message queue via LISTEN/NOTIFY
• TSDB via Timescale or other plugins
• partial support for columnar DB, but there are also other projects that use the Postgres wire protocol for columnar DBs (ClickHouse)
• full-text search via tsvectors
• GIS via PostGIS

So almost everything I need I have built-in or within my grasp. So I can reduce amount of the moving parts and add more dependencies only when I will know for sure that I will need them. Until then I can have everything in single place.

I have been an SQL guy for over 15 years. But not only that, I have yet to find a situation where Postgres have fallen short. Also, if you are fairly new to the Phoenix eco system, all guides pretty much assumes that you are using postgres.

Why you want to use NoSQL DB at all? Is there any technical reason? Assuming by your question - no, then use PostgreSQL.

I love nosql, I do not love sql, do you have any idea about my question

I cant tell you what to use, except Postgres, but make sure that whatever you use, there is a fairly maintained ecto adapter.