v0.2.0

Release Notes - v0.2.0

Release Date: December 28, 2025

🎉 What's New

Multi-Provider Support

PromptCache now supports three major AI providers for embeddings and semantic verification:

• OpenAI - text-embedding-3-small + gpt-4o-mini
• Mistral AI - mistral-embed + mistral-small-latest
• Claude (Anthropic) - voyage-3 (Voyage AI) + claude-3-haiku-20240307

Dynamic Provider Switching

Switch providers at runtime without restarting the service:

# Get current provider
curl http://localhost:8080/v1/config/provider

# Switch to Mistral
curl -X POST http://localhost:8080/v1/config/provider \
  -H "Content-Type: application/json" \
  -d '{"provider": "mistral"}'

Enhanced Configuration

New environment variables for fine-tuned control:

Provider Selection:

export EMBEDDING_PROVIDER=openai  # Options: openai, mistral, claude

Configurable Thresholds:

export CACHE_HIGH_THRESHOLD=0.70   # Direct cache hit (default: 0.70)
export CACHE_LOW_THRESHOLD=0.30    # Clear miss (default: 0.30)

Gray Zone Control:

export ENABLE_GRAY_ZONE_VERIFIER=true  # Enable/disable LLM verification

Performance & Benchmarking

New tools for testing and optimization:

• scripts/run.sh - Easy project startup with validation
• scripts/benchmark.sh - HTTP endpoint performance testing
• internal/semantic/benchmark_test.go - Micro-benchmarks for core functions
• Enhanced Makefile with make benchmark and make bench-go

Comprehensive Documentation

Full documentation site deployed to GitHub Pages:

• Getting Started Guide
• Complete REST API Reference
• Provider Configuration Guide
• Contributing Guidelines
• All with code examples in Python, cURL, and JavaScript

🔧 API Changes

New Endpoints

GET /v1/config/provider

Get the currently active provider:

{
  "provider": "openai",
  "available_providers": ["openai", "mistral", "claude"]
}

POST /v1/config/provider

Switch providers dynamically:

{
  "provider": "mistral"
}

Existing Endpoints

• POST /v1/chat/completions - Unchanged, fully backward compatible

📊 Performance Improvements

• Thread-safe provider switching with RWMutex
• Optimized cosine similarity calculations
• Efficient vector serialization/deserialization
• Concurrent request handling preserved

Benchmark Results:

BenchmarkCosineSimilarity-10                     3252341    369.8 ns/op      0 B/op    0 allocs/op
BenchmarkFindSimilar-10                           194473   6183 ns/op    3585 B/op   13 allocs/op
BenchmarkSetProvider-10                          1000000   1096 ns/op       0 B/op    0 allocs/op
BenchmarkFindSimilar_MultipleEmbeddings-10         17383  68913 ns/op   35761 B/op  130 allocs/op

🧪 Testing

• 25+ comprehensive unit tests covering all providers
• Provider factory tests with edge cases
• Thread-safety tests for concurrent operations
• Configuration validation tests
• Gray zone behavior tests
• All tests passing with >80% code coverage

📝 Documentation

New Documentation Files

• docs/index.md - Documentation homepage
• docs/getting-started.md - Installation and quick start
• docs/api-reference.md - Complete REST API docs
• docs/configuration.md - Environment variables and tuning
• docs/providers.md - Provider setup and comparison
• CONTRIBUTING.md - Contribution guidelines

Updated Files

• README.md - Enhanced with provider guides and API management
• CHANGELOG.md - Detailed version history
• docker-compose.yml - All provider environment variables

🔄 Migration Guide

From v0.1.x to v0.2.0

No breaking changes! v0.2.0 is fully backward compatible.

Optional Updates

1. Set provider explicitly (defaults to OpenAI):

   export EMBEDDING_PROVIDER=openai

2. Customize thresholds (optional):

   export CACHE_HIGH_THRESHOLD=0.70
   export CACHE_LOW_THRESHOLD=0.30

3. Control gray zone verification (enabled by default):

   export ENABLE_GRAY_ZONE_VERIFIER=true

🐛 Bug Fixes

• Fixed thread-safety in provider access
• Improved error handling for missing API keys
• Enhanced configuration validation with fallback defaults

🔐 Security

• No security vulnerabilities introduced
• All API keys remain secure in environment variables
• No sensitive data logged

📦 Dependencies

• Go 1.21+
• Redis 7.0+
• Provider SDKs:
  • OpenAI Go SDK
  • Mistral Go SDK (unofficial)
  • Anthropic Go SDK

🚀 Deployment

Docker

docker-compose up

Manual

export EMBEDDING_PROVIDER=openai
export OPENAI_API_KEY=your-key
make run

🙏 Acknowledgments

Thanks to all contributors and users who provided feedback!

📚 Resources

• Documentation: https://messkan.github.io/prompt-cache
• Repository: https://github.com/messkan/prompt-cache
• Issues: https://github.com/messkan/prompt-cache/issues

💬 Support

• Open an issue on GitHub
• Check the documentation
• Review the contributing guide

---

Full Changelog: https://github.com/messkan/prompt-cache/blob/main/CHANGELOG.md