Skip to content

CHANGELOG.md

Latest commit

 

History

History
145 lines (101 loc) · 5.72 KB

CHANGELOG.md

File metadata and controls

145 lines (101 loc) · 5.72 KB

Changelog

All notable changes to this project will be documented in this file.

[0.2.10] - 2025-11-18

Changed

  • Make version requirement more flexible for platformdirs dependency, from >=4.4 to >=4.

[0.2.9] - 2025-11-13

Added

  • Fix an issue where telemetry did not have access to files in ~/home directory

[0.2.8] - 2025-11-11

Added

  • Anonymously track install_id and install_date

[0.2.7] - 2025-11-06

Added

  • Anonymously track model_path and model_version.

[0.2.6] - 2025-11-06

Changed

  • Fix an issue because of which telemetry events were not sent.

[0.2.5] - 2025-11-04

Added

  • Track additional and anonymous runtime properties like local numpy and pandas versions.

[0.2.4] - 2025-10-28

Changed

  • Fix an issue due to which telemetry was not recorded for TabPFN-TS using the API client

[0.2.3] - 2025-10-16

Changed

  • Telemetry: Surpress any telemetry logs, e.g. when internet connection not available.

[0.2.2] - 2025-09-23

Added

  • Ping Event: Add frequency to ping events, allowing to send weekly and monthly pings.

[0.2.1] - 2025-09-23

Added

  • License: Define open source licensing, Apache 2.0.

[0.2.0] - 2025-09-19

Added

  • Toggle telemeetry: Control telemetry toggle using a remote configuration file.
  • Extensions: Allow setting the context for TabPFN extensions.

[0.1.9] - 2025-09-10

Added

  • Hierarchical Extension Context: Implemented context-aware extension tracking using Python's contextvars
  • Extension Decorator: Added set_extension() decorator for automatic extension context management
  • Context Preservation: Extension context is preserved across nested function calls without overriding higher-level contexts

Enhanced

  • Telemetry Events: Events now automatically include the current extension name when available
  • Context Management: Improved extension context handling with proper cleanup and inheritance

[0.1.7] - 2025-09-09

Enhanced

  • Interactive Prompts: Improved user input validation for telemetry consent prompts
  • Input Validation: Now requires explicit "y" or "n" input instead of accepting empty strings
  • Retry Logic: Added retry mechanism with up to 3 attempts for invalid input
  • User Experience: More specific error messages and clearer input requirements

Changed

  • Prompt Behavior: Identity prompt now enforces explicit user choice rather than defaulting on empty input
  • Error Handling: Enhanced input parsing with customizable retry messages
  • Type Safety: Updated type annotations to support optional outcomes for invalid input

[0.1.6] - 2025-09-08

Added

  • Interactive Telemetry Module: New telemetry.interactive package with optional dependencies
  • Runtime Detection: Automatic detection of IPython, Jupyter, and TTY environments
  • User Prompts: Interactive consent prompts for telemetry opt-in
  • State Management: File-based state storage for user preferences
  • Model Tracking: Decorators for automatic model call tracking
  • Optional Dependencies: Install with pip install tabpfn_common_utils[telemetry-interactive] for interactive features

Changed

  • Package Structure: Refactored telemetry into core/ and interactive/ modules
  • API Separation: Core telemetry always available, interactive features behind [telemetry-interactive] extra
  • Import Strategy: Graceful fallback when interactive dependencies are not installed

[0.1.4] - 2025-09-08

Changed

  • Dependency Compatibility: Updated numpy and pandas requirements to be compatible with scipy 1.11.1 and tabpfn ecosystem
  • NumPy Version: Downgraded numpy requirement from >=2,<3 to >=1.21.6,<1.28.0 for scipy compatibility
  • Pandas Version: Updated pandas requirement to >=1.4.0,<3 for broader compatibility

[0.1.2] - 2025-09-05

Changed

  • NumPy Compatibility: Updated scikit-learn dependency to match tabpfn requirements
  • Development Dependencies: Added pytest as an explicit dev dependency for consistent testing across environments

[0.1.1] - 2025-09-05

Added

  • Privacy-First Telemetry System: New GDPR-compliant telemetry system using PostHog for anonymous, aggregated usage data collection
  • Telemetry Events: Implemented specific events (fit_called, dataset, predict_called, PingEvent) for TabPFN usage tracking
  • ProductTelemetry Class: Singleton service for capturing and pushing events with opt-out via TABPFN_DISABLE_TELEMETRY environment variable
  • Python 3.13 Support: Added Python 3.13 classifier and CI testing
  • Comprehensive Documentation: Complete README overhaul with installation guides, quick start examples, and privacy compliance details

Changed

  • Package Metadata: Updated pyproject.toml with authors, maintainers, keywords, and extended Python version support
  • Dependencies: Added posthog~=6.7 as runtime dependency
  • Type Hints: Enhanced type safety across utility modules with explicit typing imports
  • Version Management: Dynamic package version retrieval using importlib.metadata

Fixed

  • Type Annotations: Corrected return type annotations in get_example_dataset() function
  • DataFrame Compatibility: Improved pandas DataFrame column initialization for better type safety
  • Code Quality: Enhanced type hinting and removed outdated modules

Removed

  • load_test.py: Removed outdated module (moved to API repository)
  • Outdated Configuration: Cleaned up redundant pyright and ruff configuration sections

[0.1.0] - Initial Release

Added

  • Core utility functions for TabPFN
  • Regression prediction result handling
  • Data processing and serialization utilities
  • Basic project structure and testing framework