Changelog
Track all updates, new features, improvements, and fixes to the LLMYourWay platform. We ship fast and iterate based on your feedback.
Version 0.5.0
November 4, 2025Features Page
New /features page showcasing platform capabilities with real-time stats
- Smart Discovery with 9,076+ models
- Performance Benchmarks (LMSYS, OpenLLM v2, MTEB)
- Pricing Intelligence across providers
- Side-by-Side Comparison tool
- Scenario Builder for custom use cases
- Trending & Analytics dashboard
Baseline Model Comparison
Compare open-source models against commercial market leaders
- OpenAI (GPT-4o, GPT-4 Turbo, GPT-3.5 Turbo)
- Anthropic (Claude 3.5 Sonnet, Claude 3 Opus, Claude 3 Haiku)
- Google (Gemini 1.5 Pro, Gemini 1.5 Flash)
- Pricing comparison per 1M tokens
- Context window and capability comparisons
Stats API Enhancement
Enhanced /api/stats with comprehensive coverage metrics
- Benchmark coverage percentage (52%)
- Description coverage (99.4%)
- Context window coverage (61.5%)
- 5-minute caching for performance
- Real-time author and model counts
Version 0.4.0
November 3, 2025Trending Page Loading
Fixed infinite loading state with improved error handling
- Added .finally() to guarantee loading state updates
- Proper error state management
- Page now returns HTTP 200 consistently
Compare Page Model Limit
Removed hardcoded MOCK_MODELS fallback
- Increased API limit from 60 to 100 models
- Now loads 100+ real models from database
- Removed 6-model limitation
Category Filtering
Fixed category search returning 0 results
- Pass category as API parameter instead of text search
- All 6 categories now work correctly
- CODE_MODEL, EMBEDDING_MODEL, etc. properly filtered
Task Selection Checkboxes
Rewrote task filtering logic
- Now uses tags and category fields correctly
- Chat, Coding, RAG, Vision filters functional
- Removed non-existent capabilities field check
Benchmark Ingestion Boost
Massive benchmark data coverage improvement
- Coverage increased from 15% → 52%
- 31,728 total benchmarks across models
- LMSYS Arena: 266 models ingested
- OpenLLM v2: 4,200 benchmark records
- 4,758 of 9,076 models now have benchmarks
Author Stats Aggregation
Fixed author download counts showing as 0
- Aggregated 2,245 authors with real download stats
- Top author: sentence-transformers (192M downloads)
- Qwen: 85M downloads across models
- Real community impact metrics
Version 0.3.5
November 2, 2025Global Command Bar
Cmd+K search from any page
- Instant model search across 9,076 models
- Keyboard-first navigation
- Recent searches and suggestions
Server-Side Models Directory
/models page now searches all models
- Live API integration
- Pagination support
- Provider filtering
- Real-time search results
Trending Page
See momentum leaders with real data
- Download trends
- Community engagement metrics
- Emerging model detection
Scenario Builder Task Alignment
Updated to use actual OpenLLM v2 benchmarks
- OpenLLM Avg, BBH, GPQA, IFEval
- Math Lvl 5, MMLU-Pro, MUSR, Arena Elo
- Replaced outdated mtbench, mmlu, gsm8k tasks
HuggingFace Authentication
Added free HF token to fix rate limiting
- 16x rate limit increase (60 → 1,000 req/hour)
- Access to gated models
- 99.9% enrichment success rate
- HTTP 429 errors eliminated
Data Cleaning Infrastructure
Comprehensive cleaning pipeline for model data
- HTML/markdown/boilerplate removal from descriptions
- User-friendly display names
- 100+ tag normalization mappings
- cleanDescription and cleanTags fields added
Version 0.3.0
October 19, 2025Benchmark Leaderboards
New /benchmarks/leaderboard page with 7 task categories
- General tasks (MMLU, ARC, HellaSwag)
- Math benchmarks (GSM8K, MATH, AIME)
- Chat performance (MT-Bench, Arena ELO)
- Embeddings (MTEB)
- Long context benchmarks
Enhanced Model Metadata
- Architecture details (model type, layers, attention heads)
- Tokenizer information (class, max length)
- Context window specifications
- README summaries on detail pages
Model Detail Pages
Enriched with HuggingFace data
- Context windows displayed
- Better data attribution
- Source links to official documentation
Initial Database Stats
- 222 models tracked
- 119 authors/organizations
- 221 models synced in last 24 hours
- Benchmark data from 7 major sources
Version 0.2.0
October 18, 2025Automated Data Pipeline
- Rolling HuggingFace ingestion (10-minute batches)
- Daily benchmark collection from 7 sources
- Model enrichment with config/tokenizer details
Production Deployment Automation
- Git push triggers auto-deployment
- Environment preservation across deployments
- Automated schema migrations
Infrastructure
- Database schema with typed enrichment fields
- Hybrid storage strategy (typed + JSON)
- 60-second API response caching
- Deployment workflow safety
Deployment Issues
- Git authentication (switched to SSH)
- Docker deployment permission errors
- Environment file preservation
- Database schema synchronization
Version 0.1.0
October 17, 2025Homepage
- Hero section with value proposition
- Live stats (models, authors, updates)
- Featured models section
- Trending models
- Newsletter signup
Model Pages
- Browse all models with filtering
- Model detail pages with specifications
- Category grouping
- Framework filtering
Benchmarks Overview
- Explanation of common benchmarks
- Methodology documentation
Infrastructure
- Next.js 14 App Router
- Prisma + PostgreSQL
- Docker Compose deployment
- Digital Ocean hosting
- Automatic HTTPS with Caddy
Want to see something specific? Share your feedback at GitHub