Skip to content

Changelog

All notable changes to Prompt Amplifier are documented here.

[0.2.2] - 2024-12-12

🚀 New Features

Caching Layer

  • MemoryCache: Fast in-memory LRU cache with TTL support
  • DiskCache: Persistent disk-based cache for cross-session caching
  • CacheConfig: Configurable TTL, max size, and granular cache controls
  • Automatic caching for search() and expand() operations
  • Cache statistics with hit rate tracking
  • Per-operation cache bypass with use_cache=False

✨ Improvements

  • Added cache, cache_config, and enable_cache parameters to PromptForge
  • Added get_cache_stats() method for cache performance monitoring
  • Added clear_cache() method for cache management
  • Added cache_enabled property
  • Updated __repr__ to show cache status

📚 Documentation

  • Added comprehensive Caching guide
  • Updated quickstart with API key requirements
  • Added troubleshooting section
  • Added Google Colab ready examples

[0.2.1] - 2024-12-11

✨ Improvements

  • Input Validation: Clear errors for empty prompts, texts, queries
  • Structured Logging: Debug visibility with logging module
  • Better Error Messages: Actionable suggestions in error text
  • Fixed Generators: Separated Anthropic and Google into dedicated files

🐛 Bug Fixes

  • Fixed AnthropicGenerator import from wrong module
  • Fixed GoogleGenerator import from wrong module
  • Fixed TF-IDF error message for insufficient documents

[0.2.0] - 2024-12-09

🚀 New Features

Web Loaders

  • WebLoader: Fetch and parse web pages with BeautifulSoup
  • YouTubeLoader: Extract video transcripts using youtube-transcript-api
  • SitemapLoader: Crawl entire websites via sitemap.xml
  • RSSLoader: Load content from RSS/Atom feeds

New Embedders

  • CohereEmbedder: Cohere embedding API integration
  • CohereRerankEmbedder: Cohere reranking for improved retrieval
  • VoyageEmbedder: Voyage AI embeddings for specialized domains
  • JinaEmbedder: Jina AI embeddings with multilingual support
  • MistralEmbedder: Mistral AI embeddings (EU-based)

New Generators

  • OllamaGenerator: Run LLMs locally with no API costs
  • MistralGenerator: Mistral AI API for European compliance
  • TogetherGenerator: Access 50+ open-source models via Together AI

Evaluation Module

  • calculate_expansion_quality(): Measure prompt quality (structure, specificity, completeness, readability)
  • calculate_retrieval_metrics(): Standard IR metrics (precision, recall, MRR, NDCG)
  • calculate_diversity_score(): Measure result diversity
  • calculate_coherence_score(): Measure prompt-context coherence
  • compare_embedders(): Benchmark multiple embedders
  • benchmark_generators(): Benchmark multiple LLMs
  • EvaluationSuite: Run comprehensive test suites

CLI Tool

  • prompt-amplifier expand: Expand prompts from command line
  • prompt-amplifier search: Search documents
  • prompt-amplifier compare-embedders: Benchmark embedders
  • prompt-amplifier evaluate: Run evaluation suite
  • prompt-amplifier version: Show version

📦 Dependencies

New optional dependency groups: - [loaders]: beautifulsoup4, youtube-transcript-api, feedparser - [embeddings-cohere]: cohere - [embeddings-voyage]: voyageai - [embeddings-jina]: jina-embeddings - [embeddings-mistral]: mistralai - [generators-ollama]: ollama - [generators-mistral]: mistralai - [generators-together]: together

📚 Documentation

  • Added Web Loaders guide
  • Added Evaluation Metrics guide
  • Added Additional Embedders guide
  • Added Additional Generators guide
  • Added CLI documentation
  • Updated API reference

[0.1.5] - 2024-12-08

🐛 Bug Fixes

  • Fixed CI/CD pipeline for all Python versions (3.9-3.12)
  • Fixed type hints compatibility for Python 3.9/3.10
  • Fixed Black formatting issues
  • Optimized GitHub Actions disk space usage

📦 Dependencies

  • Updated numpy>=1.24.0 for Python 3.12 compatibility
  • Updated scikit-learn>=1.3.0 for Python 3.12 compatibility

[0.1.4] - 2024-12-08

✨ Improvements

  • Added from __future__ import annotations for Python 3.9/3.10 compatibility
  • Fixed Ruff linting configuration

[0.1.3] - 2024-12-08

🔧 Changes

  • Renamed package from promptforge to prompt-amplifier
  • Updated all import paths
  • Updated documentation

[0.1.2] - 2024-12-07

✨ Initial Release

Core features: - PromptForge Engine: Main orchestration class - Document Loaders: TXT, CSV, JSON, DOCX, Excel, PDF - Chunkers: RecursiveChunker with configurable size/overlap - Embedders: TF-IDF, BM25, Sentence Transformers, OpenAI, Google - Vector Stores: Memory, ChromaDB, FAISS - Retrievers: Vector, Hybrid - Generators: OpenAI, Anthropic, Google

Roadmap

v0.3.0 (Planned)

  • More vector store integrations (Pinecone, Qdrant, Weaviate)
  • Streaming support for generators
  • Async API support
  • Custom prompt templates
  • Multi-modal support (images)

v1.0.0 (Future)

  • Stable API
  • Production-ready performance
  • Enterprise features
  • Comprehensive benchmarks

Contributing

See Contributing Guide for how to submit changes.

Report bugs and feature requests on GitHub Issues.