hemanthaddagada/repo-discovery

MAJOR FIX: Always use semantic search + fixed pagination for semantic results

Fix stuck loading - add request cancellation and 30s timeout

SAFE SEMANTIC: Background load + memory mapping - no startup delay

SIMPLE VERSION: No semantic, no signal tricks - just working code

FULL FEATURES: Semantic search with timeout protection, graceful fallback

MINIMAL VERSION: Remove semantic search temporarily to fix deployment

COMPLETE FIX: Simplified search_api, static files AFTER all API routes

Fix stuck loading - preload semantic, better errors, 150ms debounce

ALWAYS use semantic search - max discoverability, 50 results, 0.15 threshold

Make semantic search more aggressive - trigger for 3+ word queries

Critical fix: Properly position catch-all route at END of file

Fix: Move catch-all route to END - was blocking /search API

Fix: Add debounced search and SPA fallback route

Fix pagination bug - remove double offset application

Fix Tailwind CSS compilation - downgrade to v3 for compatibility

Add beautiful modern UI with gradients and improved styling

Trigger rebuild with dist files

Fix: Use pre-built React dist to avoid Docker timeout

Add React UI with shadcn/ui - compact table and grid views

COMPLETE SYSTEM: Original scoring (stars+recency+activity) + semantic search + all filters

Revert to clean working search without complex semantic logic

Fix typo: FastAPI import

Restore semantic search and hybrid ranking with embeddings

Implement full search and filtering functionality

DEFINITIVE FIX: Add complete relevance_explanation with all required subfields

Use WORKING local search_api.py - proven to work locally

Add default values for numeric fields to prevent toFixed errors

Replace with simple API that includes relevance_explanation

Complete simple API with relevance_explanation

Add relevance_explanation field to API response

ULTRA SIMPLE search - just list documents, no complex logic

Fix app.py indexing check - force reindex if index is empty

FINAL FIX: Use '*' wildcard for empty queries - guaranteed to return all results

CRITICAL: Disable semantic on empty query to show all 10858 repos

Add root route to serve index.html

Fix API_URL to use relative path for HF Spaces

Use EXACT working local files - search_api.py and index.html

DEFINITIVE FIX: Empty query search without sort, then sort in Python - GUARANTEED results

CRITICAL: Fix results not displaying - added null check for results array

FINAL FIX: Simplified search to use index.search() for all queries - guaranteed to work

Fix search lag (increase debounce), white screen (error handling), and loading issues

CRITICAL FIX: Use get_documents for empty queries to browse all repos

Add console logging to debug results display issue

CRITICAL FIX: Simplified pagination to use Meilisearch directly - no more 'No results'

Re-enable auto-search to show all repos on page load

Remove auto-search on page load to prevent stuck loading state

Fix semantic search: disable on empty queries and preserve full result count

Fix UnboundLocalError in semantic debug logging

Fix API URL to work on HF Spaces

Add FileResponse import