Skip to content

CHANGELOG.md

Latest commit

 

History

History
84 lines (72 loc) · 4.34 KB

CHANGELOG.md

File metadata and controls

84 lines (72 loc) · 4.34 KB

Changelog

[0.1.2] - 2025-04-27

Fixed

  • Fixed metadata filtering issue in memory retrieval:
    • Corrected the _matches_metadata_filters method to properly evaluate metadata fields
    • Improved handling of importance score filtering with proper normalization
    • Added debug logging for metadata filter evaluation process
    • Fixed validation tests to use correct metadata fields in test memories
    • Enhanced metadata comparison logic to handle missing fields gracefully
  • Fixed ImportanceStrategy implementation:
    • Resolved issue with importance value parsing from different memory formats
    • Fixed importance comparison logic for non-normalized importance values
    • Corrected sorting behavior to properly prioritize memories with higher importance
    • Added proper handling for memory objects with 'id' field vs 'memory_id' field
    • Fixed query handling for different query formats (direct value vs. dictionary)

Improved

  • Enhanced debug output in validation scripts for better troubleshooting
  • Improved error handling for missing metadata during filtering operations

[0.1.1] - 2025-04-25

Added

  • Added checksum functionality to memory metadata:
    • New checksum field in MemoryMetadata for data integrity verification
    • Added utility functions in memory/utils/checksums.py for generating and validating checksums
    • Implemented automatic checksum generation when storing memories
    • Added checksum validation when retrieving memories
    • Added integrity verification flags in memory metadata

Improved

  • Enhanced data integrity through checksum verification across all memory tiers
  • Added support for different hashing algorithms in checksum generation
  • Implemented comprehensive test suite for checksum functionality

[0.1.0] - 2025-04-05

Refactoring

  • Extracted duplicate code into common utilities in utils.py:
    • cosine_similarity: Moved from vector_store.py for vector comparison
    • flatten_dict: Moved from autoencoder.py for dictionary flattening
    • object_to_text: Moved from text_embeddings.py for object text conversion
    • filter_dict_keys: Moved from compression.py for dictionary filtering

Updated

  • Modified text_embeddings.py to use the common object_to_text function
  • Updated vector_store.py to use the shared cosine_similarity function
  • Revised autoencoder.py to use the centralized flatten_dict function
  • Changed compression.py to use the common filter_dict_keys function
  • Enhanced object_to_text function for better handling of empty values and nested structures
  • Refactored embeddings module by removing direct TextEmbeddingEngine import

Added

  • Added comprehensive module documentation in __init__.py for LLM context
  • Added detailed documentation file docs/Embeddings.md explaining the Embeddings module components and usage
  • Exposed utility functions in __all__ list for easy importing
  • Added comprehensive unit tests for the embeddings module:
    • Tests for AutoEncoder and dimensionality reduction capabilities
    • Tests for NumericExtractor functionality
    • Tests for vector compression operations
    • Tests for text embedding operations
    • Tests for vector store functionality
  • Added vector compression utilities:
    • quantize_vector and dequantize_vector for bit-depth reduction
    • compress_vector_rp and decompress_vector_rp for random projection compression
    • CompressionConfig for managing compression settings
  • Added MockRedis implementation with pipeline and pubsub support for testing
  • Enhanced MockRedis implementation for development and testing:
    • Added configuration options in RedisSTMConfig and RedisIMConfig
    • Created RedisFactory to streamline Redis client instantiation
    • Added comprehensive integration tests for Redis stores with MockRedis
    • Updated README with MockRedis usage instructions

Improved

  • Eliminated code duplication across the embeddings module
  • Enhanced maintainability by centralizing common functionality
  • Improved consistency in utility functions across the codebase
  • Enhanced RedisFactory to support MockRedis with custom ResilientRedisClient for better testing
  • Improved MockRedis module documentation with usage examples and implementation details

Fixed

  • Updated random number generation in vector compression to use local RandomState instead of global seed to prevent interference with other code