feat: implement weighted search for title and description

[Kayd-06]

Closes #83.

Description

This PR updates the search functionality to query both the title and description fields with a weighted ranking system, rather than just the generic icontains lookup. This significantly improves model discoverability by factoring in metadata relevance.

Changes made

• PostgreSQL Full-Text Search: Utilized Django's PostgreSQL features (SearchVector, SearchQuery, and SearchRank).
• Weighted Rankings:
  • title matches are assigned weight A.
  • description matches are assigned weight B.
• Filtering & Ordering: Applied a minimum rank threshold (rank__gte=0.01) to exclude completely irrelevant results, and sorted the results by -rank to ensure the most contextually relevant models appear first.
• Frontend & API: Updated both mainapp/views.py (frontend search) and mainapp/api.py (API endpoints search_full and search_title) to maintain consistency across the entire app.
• Fixed a minor precedence bug where sorting by -rank in the views.py search view mistakenly applied to unrelated querysets (like category).

Testing

• Ensure PostgreSQL is set up for the developer environment (as the project already expects django.contrib.postgres).
• Ran existing tests in test_search.py and test_search_full.py, which all pass successfully.

[Kayd-06]

Hi @AyushDharDubey, I've opened a PR to address this issue.

I implemented PostgreSQL's full-text search with a weighted ranking system (A for title, B for description) to improve model discoverability and adjusted the API/views to utilize it.

[AyushDharDubey]

Looks mostly good, could you please include tests to support the implementation.

[Kayd-06]

@AyushDharDubey I've pushed the requested changes:

• Renamed API Endpoint: Changed /api/search/title to /api/search/model across the codebase and docs.
• Fixed Query Filter Loop: Refactored views.py from elif to if statements. This fixes an UnboundLocalError bug and ensures a search query correctly applies on top of any tag/category filters.
• Added Typo Tolerance: Added a migration for PostgreSQL's pg_trgm extension and implemented fuzzy matching using TrigramSimilarity on both titles and descriptions.
• Updated Tests: Adjusted the test suite to expect and accurately verify the new fuzzy trigram matching behavior.

Everything is tested and passing locally.

[lonvia]

I haven't tried it but looks good to me in general. We might want to adjust the similarity threshold a bit upwards later. 0.1 is quite small. But we can try with real data first to see how it comes out.

[Kayd-06]

@lonvia I originally kept the variable when refactoring the filtering precedence, but you're absolutely right that it's no longer necessary. I've updated views.py to remove the unused variable and assign directly to filtered_models.

[Kayd-06]

Hi @AyushDharDubey, thanks for the approval! Just a quick bump to see if you or @lonvia could approve the GitHub actions workflow to run so this can merge. No rush!

[AyushDharDubey]

Hello, we were actually expecting for you to address Sarah's review, which may have been overlooked.

[Kayd-06]

@AyushDharDubey I have done the changes. Is it good to go?

[AyushDharDubey]

Hello Kunal, this partially addresses the previous review.

However, we still need to restore the title parameter for api/search/title/<title>. As Sarah indicated earlier, Changing or removing this would result in a breaking change for the existing API users.

The updated search_full seems fine.

[Kayd-06]

@AyushDharDubey Sorry for the partial address last time! I've now restored the /api/search/title/<title> endpoints and renamed search_model back to search_title, ensuring it only performs the weighted search on the title field to maintain backward compatibility. Is it good to go now?