It can't go that far. Postgres is limited by dictionaries, stemming/processing, query semantics (like fuzzy searching), and the biggest issue of all being a lack of modern relevance algorithms. It's good for limited scenarios where you just need more than a SQL LIKE statement, and chaining some functions together can get you decent results [1] without adding another datastore.
However search tech is pretty mature with Lucene at the core and there are many better options [2] from in-process libraries to simple standalone servers to full distributed systems like Elastic. There are also other databases (relational like MemSQL, or documentstores like MongoDB/RavenDB) that are adding search as native querying functions with most of the abilities of ES. If search is a core or complex part of your application (like patterns in raw image data or similarities in audio waveforms) then that's where ES will excel.
However search tech is pretty mature with Lucene at the core and there are many better options [2] from in-process libraries to simple standalone servers to full distributed systems like Elastic. There are also other databases (relational like MemSQL, or documentstores like MongoDB/RavenDB) that are adding search as native querying functions with most of the abilities of ES. If search is a core or complex part of your application (like patterns in raw image data or similarities in audio waveforms) then that's where ES will excel.
1. https://stackoverflow.com/questions/46122175/fulltext-search...
2. https://gist.github.com/manigandham/58320ddb24fed654b57b4ba2...