If you use Java then Lucene has Okapi BM25 (https://lucene.apache.org/)

Wumpus also has Okapi BM25 (http://www.wumpus-search.org/)

Programming a search engine is not that difficult and you can tweak the parameters of your algorithm with less effort. If you decide this route I suggest one of my main references for my implementation which is:

Information Retrieval: Implementing and Evaluating Search Engines by Stefan Büttcher, Charles L. A. Clarke, Gordon V. Cormack

Even if you do not code it yourself it is an invaluable source

Check this link:

https://dl.acm.org/citation.cfm?id=1646237

Yes, as far as I know, Lucene includes BM25 implementation. However, I have implemented this myself to avoid various boosts that may have introduced by Lucerne for their implementation.

Thanks all for your suggestions

please, any suggestion on how to search for Arabic text? most of the search engine implementations use English text

BM25 is language agnostic. Lucene should be able to process UTF-8 characters. The part that you will most likely have to code is the tokenizer, stemmer, list of stopwords, etc. While I think you could find these online you will have to provide them in a format that Lucene understands.