Enhancements to Semantic Search Functionality in PyFluent

[seanpearsonuk]

While reviewing the documented examples for PyFluent's semantic search feature (docs), I observed a few areas that could be improved:

• Search is limited to the entire API tree (pf.search(...)). It would be helpful to support more targeted searches (e.g., within a specific object or sub-tree).
• Foreign language support appears to rely on translation and string matching. The translated term must still match a meaningful substring within each returned path. This may not be sufficient for truly semantic search and can produce noisy or inconsistent results.
• Wildcard support must be explicitly enabled via a special argument, rather than being inferred from the query string itself. It’s unclear if this should be necessary—there might be value in supporting inferred wildcards or a more intuitive query syntax.

We may want to test-drive an improved design that better aligns with user expectations, especially around flexibility and internationalization.

[seanpearsonuk]

@mkundu1 for additional comments.

[mkundu1]

For the implementation, we should build a search-index instead of each search() call scanning through all paths. We can start by building a cache of name -> ordered_results map, which would cover string-matching based and wildcard based search. For semantic search, we need to additionally cache the synsets, which we are already doing. The foreign language search can probably be implemented based on these 2 caches.

For reference, here is a user-script for a basic string-matching search for settings API with a cache of name -> ordered_results map.

The caches should be part of the package. Based on their size, we may need to employ some strategy for its packaging.

As we incremnetally implement this, we should write unittests based on some test data (not the actual settings API).