Apache Tika uses language profiles from Optimaize Language Detector, which are based on statistical n-gram models. If Farsi isn't recognised, it typically means that:

1. The text sample is too short or ambiguous, or

2. The language profile for Farsi (fa) is missing

You could try;

1. use a longer or more diverse sample in Farsi

2. make sure you are using the OptimizeLangDetector, and

3. ensure you're using a recent Tika version