Release Notes

This page contains the release history and changelog for params-proto.

Version 3.3.0 (2025-02-04)

✨ Features

Deep Nested Dot Notation for CLI: Support for overriding nested dataclass fields via CLI
bash
```
# Before (not supported):
python train.py train-config --model.hidden-size 512  # error

# Now works:
python train.py train-config --epochs 200 --model.hidden-size 512 --model.num-layers 8
```
- Recursively detects nested dataclass fields in type annotations
- Supports arbitrary nesting depth (e.g., --model.encoder.layers 12)
- Automatically constructs nested dataclass instances

Version 3.2.4 (2025-02-04)

🐛 Bug Fixes

Context Manager Protocol: Fixed regression where __enter__ and __exit__ methods were stripped from @proto and @proto.prefix decorated classes
- Classes implementing context managers now work correctly with with statements
- Also preserves all user-defined protocol methods: __call__, __iter__, __next__, __getitem__, __len__, etc.
Simplified Implementation: Removed method wrapping that caused issues
- Decorated class is now a proper subclass of the original, so methods are inherited naturally
- No more wrapping of methods to return self - use explicit return self if needed
- Standard Python semantics: classmethods receive cls, not bound to instance

🧪 Testing

Added comprehensive tests for context manager and other Python protocol preservation

Version 3.2.2 (2025-02-03)

🐛 Bug Fixes

super() in @proto Methods: Fixed super() calls failing with TypeError: super(type, obj): obj must be an instance or subtype of type
- Methods using super() in @proto or @proto.prefix decorated classes now work correctly
- Root cause: decorated class was recreated as sibling instead of subclass of original
- Fix: decorated class is now a proper subclass, preserving the MRO chain
@proto Inheriting from @proto: Fixed TypeError: duplicate base class ptype when a @proto class inherits from another @proto class
- Nested @proto inheritance now works correctly
- Metaclass merging skips duplicate ptype bases

Version 3.2.1 (2025-02-01)

🐛 Bug Fixes

Positional Arguments in Subcommands: Fixed positional arguments being silently ignored after subcommand name
- Before: myapp add my-env/v1.2.3 → positional arg silently dropped
- After: myapp add my-env/v1.2.3 → positional arg captured by subcommand's required field
- Enables CLI patterns like pip install requests, cargo add serde
- Raises clear error for extra unrecognized positional arguments

📚 Documentation

Fixed outdated notes claiming Optional[str] was not supported (it works)

Version 3.2.0 (2025-02-01)

✨ Features

Unprefixed CLI Subcommand Attributes: Subcommand attributes no longer require prefix by default
- Old: python train.py train-config --config.epochs 200
- New: python train.py train-config --epochs 200
- Prefixed syntax still works for backwards compatibility
@proto.prefix Controls CLI Prefix Requirement: Classes decorated with @proto.prefix require prefixed CLI syntax
- Regular dataclasses: --epochs 200 (unprefixed)
- @proto.prefix classes: --config.epochs 200 (prefixed required)

🐛 Bug Fixes

isinstance() for @proto.prefix Instances: Fixed isinstance(instance, DecoratedClass) returning False
- Instances of @proto.prefix decorated classes now correctly pass isinstance checks
- Enables proper type checking in Union subcommand handlers

🧪 Testing

Added comprehensive nested CLI subcommand tests (test_nested_cli.py)
Added parallel test execution with pytest-xdist
Run tests in parallel: pytest -n auto

Version 3.1.2 (2025-01-27)

📚 Documentation

Claude Code Plugin: Fixed install command format in plugin documentation

Version 3.1.1 (2025-01-25)

🐛 Bug Fixes

EnvVar Descriptor Protocol: EnvVar now works in plain classes without @proto decorator
- Previously EnvVar @ "VAR" | default returned _EnvVar object in plain classes
- Now auto-resolves via __get__ descriptor when accessed as class attribute
- Fixes: AttributeError: '_EnvVar' object has no attribute 'decode'

♻️ API Changes

Removed .get() method: Use class attribute access instead
- Old: EnvVar("PORT", dtype=int).get()
- New: class C: port = EnvVar("PORT", dtype=int) then C.port
- Use invalidate_cache() to force re-read from environment

Version 3.1.0 (2025-01-23)

✨ Features

_dict Property and dict() Support: Get a clean dict of parameter values from proto classes and functions
- Config._dict returns {'lr': 0.001, 'batch_size': 32} (defaults merged with overrides)
- dict(Config) works identically via __iter__ support
- Works for both @proto/@proto.prefix classes and @proto.cli/@proto.prefix functions

📚 Documentation

Added _dict property documentation to API reference
Updated Claude skill references with clean dict examples

Version 3.0.0 (2025-01-12)

✨ Features

Automatic Type Inference for Untyped Attributes: Class attributes without explicit type annotations are now automatically inferred
- name = "hello" is treated as name: str = "hello"
- count = 42 is treated as count: int = 42
- data = None is treated as data: Any = None (generic type)
- Untyped attributes now appear in vars(self) inside __post_init__
- Explicit type annotations are always preserved
EnvVar OR Operation: Try multiple environment variable names in order
- EnvVar @ "PRIMARY" @ "FALLBACK" | default - tries PRIMARY first, then FALLBACK
- Function syntax: EnvVar("PRIMARY", "FALLBACK", default="value")
- Returns first env var that is set, or default if none are set
EnvVar Lazy Loading: Environment variables are cached after first access
- Values cached after first access via descriptor
- invalidate_cache() clears cached value for re-read

🐛 Bug Fixes

Untyped class attributes now correctly appear in vars(self) during __post_init__
Consolidated duplicate _EnvVar class definitions

Version 3.0.0-rc25 (2025-01-08)

✨ Features

New piter @ Operator Syntax: Cleaner, more readable parameter iteration
- Use piter @ {"lr": [0.001, 0.01]} instead of piter({"lr": [0.001, 0.01]})
- Chain with * without repeating piter: piter @ {"lr": [0.001, 0.01]} * {"batch_size": [32, 64]}
- Legacy function-call syntax still supported for backward compatibility

📚 Documentation

Added Operators Quick Reference table to parameter-iteration docs
Added Operator Precedence section with examples
Updated all code examples to use the new @ syntax

Version 3.0.0-rc24 (2025-12-29)

✨ Features

Literal[...] Type Support: Full validation of allowed values
- Restricts parameters to specific set of allowed values
- Works with strings, numbers, and mixed types
- Case-sensitive validation with clear error messages
- Example: Literal["adam", "sgd", "rmsprop"]
Enum Type Conversion: Automatic conversion to enum members
- Case-insensitive CLI matching for user convenience
- Supports auto() and custom value enums
- Member name displayed in help text
- Example: Optimizer.ADAM from --optimizer adam
Path Type Instantiation: Automatic pathlib.Path creation
- Strings converted to Path objects automatically
- Works with relative and absolute paths
- Ready to use with pathlib methods
- Part of general callable type instantiation pattern
dict Type Parsing: Safe parsing using ast.literal_eval
- Supports nested structures and mixed types
- No code execution risk - only evaluates literals
- Works with both single and double quotes
- Example: '{"lr": 0.01, "batch_size": 32}'

🎯 General Callable Type Instantiation

Implemented unified approach for custom types:

Any callable type annotation is automatically instantiated with string value
Path(value) for paths
Custom classes with string constructors
dataclasses and other callables

📚 Documentation

Comprehensive sections for Literal, Enum, Path, dict types
Security note on ast.literal_eval for dict parsing
CLI usage examples for all new types
Updated Type Support Matrix showing all types as ✅ Full

📊 Test Coverage

Added 17 comprehensive tests in test_advanced_types.py:

✅ Literal validation (strings, numbers, mixed)
✅ Enum conversion (auto() and custom values)
✅ Path instantiation (relative and absolute)
✅ dict parsing (simple, nested, mixed types)
✅ Combined usage of multiple advanced types
✅ Help text generation for all types
All existing 48 CLI parsing tests still passing (65 total)

🔄 Type System Complete

Now fully supporting 13 type categories:

Primitive: int, float, str, bool
Special: Optional[T], Union types
Collections: List[T], Tuple[T, ...], dict
Validation: Literal[...], Enum
Paths: pathlib.Path
Custom: Any callable type

Version 3.0.0-rc23 (2025-12-29)

✨ Features

Tuple[T, ...] CLI Parsing: Full support for variable-length and fixed-size tuples
- Variable-length tuples: Tuple[int, ...] collects values into tuple with consistent type
- Fixed-size tuples: Tuple[int, str, float] with mixed types at each position
- CLI collects multiple values: python script.py --sizes 256 512 1024 → sizes=(256, 512, 1024)
- Automatic element/position type conversion with type safety
- Works with defaults and @proto.prefix classes
- Help text shows tuple notation: (INT,...) or (INT,STR,FLOAT)
Implementation:
- Updated _convert_type() in type_utils.py to handle both variable and fixed-size tuples
- Updated CLI parser in cli_parse.py to collect multiple values for Tuple parameters
- Updated _get_type_name() to display tuple signatures in help text
Test Suite: All 9 comprehensive test cases now PASSING in tests/test_v3/test_cli_parsing.py:
- ✅ test_tuple_variable_length_int - Variable-length integer tuples
- ✅ test_tuple_variable_length_float - Variable-length float tuples
- ✅ test_tuple_variable_length_str - Variable-length string tuples
- ✅ test_tuple_fixed_size_mixed - Fixed-size with mixed types (int, str, float)
- ✅ test_tuple_with_defaults - Overriding tuple defaults
- ✅ test_tuple_single_value - Single value wrapped in tuple
- ✅ test_tuple_empty_initialization - Empty tuple defaults
- ✅ test_tuple_with_prefix_class - Tuples in @proto.prefix classes
- ✅ test_tuple_help_strings - Help text generation

📋 Type System Updates

Type Support Matrix: Updated to show Tuple[T, ...] as ✅ Full support
Documentation: Comprehensive guide for tuple usage with examples and CLI patterns
Total Fully Working Types: int, float, str, bool, Optional[T], List[T], Tuple[T, ...], Union types

Version 3.0.0-rc22 (2025-12-29)

📚 Documentation

List[T] CLI Parsing Documentation: Comprehensive guide for using list types
- Updated docs/key_concepts/type-system.md with practical examples
- Added section showing CLI usage patterns: --items a b c → list of values
- Help text generation examples showing [STR], [INT], [FLOAT] notation
- Explanation of how multiple values are collected until next flag
- Type support matrix updated to show List[T] as ✅ Full support

Version 3.0.0-rc21 (2025-12-29)

✨ Features

List[T] CLI Parsing: Full support for List[str], List[int], List[float] and other list types
- CLI collects multiple values: python script.py --items a b c → items=['a', 'b', 'c']
- Automatic element type conversion: python script.py --counts 1 2 3 → counts=[1, 2, 3]
- Works with defaults: --items x y z overrides items: List[str] = ["default"]
- Help text shows element type: --items [STR], --counts [INT], --ratios [FLOAT]
- Support for List parameters in @proto.prefix classes
Implementation:
- Updated _convert_type() in type_utils.py to handle generic List[T] types
- Updated CLI parser in cli_parse.py to collect multiple CLI arguments for List parameters
- Updated _get_type_name() to display [INT], [STR], [FLOAT] in help text
Test Suite: All 9 comprehensive test cases now PASSING in tests/test_v3/test_cli_parsing.py:
- ✅ test_list_str_cli_parsing - Multiple string values
- ✅ test_list_int_cli_parsing - Multiple integers with type conversion
- ✅ test_list_float_cli_parsing - Multiple floats
- ✅ test_list_with_defaults - Overriding list defaults
- ✅ test_list_with_prefix_class - List in @proto.prefix classes
- ✅ test_list_empty_initialization - Empty list defaults
- ✅ test_list_single_vs_multiple_values - Single value wrapped in list
- ✅ test_list_help_strings - Help text generation
- ✅ test_list_str_whitespace_handling - Paths and special characters

📋 Documentation & Known Issues

Type System Documentation: Updated type support matrix to accurately reflect CLI parsing status
- ✅ Fully working: int, float, str, bool, Optional[T], List[T], Union[Class, Class], dataclass unions
- ⚠️ Partial: Literal[...], Enum (help text works, no runtime conversion)
- ❌ Broken: Tuple[T, ...], Path, dict (collection types)
Path Type Issue: Documented that Path type annotation is not converted from strings
- Help text shows correctly, but CLI strings are not wrapped in Path objects
Enhanced Union Types Documentation:
- Added "Why Union Types Matter" section in union_types.md
- Reorganized examples to show CLI usage first, then implementation
- Clarified that Union types are a core feature for multi-way dispatching

Migration Notes:

Workaround for Path parameters: Accept string and convert in function body

Version 3.0.0-rc20 (2025-12-28)

🐛 Bug Fixes

Optional[T] CLI Parsing: Fixed Optional[str], Optional[int], and other Optional[T] types failing to parse correctly in CLI

The issue: Optional[T] types were incorrectly treated as Union subcommands, requiring special syntax instead of working as simple optional parameters.

Before (v3.0.0-rc19):
python
```
from typing import Optional
from params_proto import proto

@proto.cli
def train(checkpoint: Optional[str] = None, learning_rate: float = 0.001):
    print(f"checkpoint={checkpoint}, lr={learning_rate}")
```
❌ This would fail:
bash
```
python train.py --checkpoint model.pt
# error: unrecognized argument: --checkpoint
```
✅ After (v3.0.0-rc20):
bash
```
# Now works correctly with standard syntax
python train.py --checkpoint model.pt
# Output: checkpoint=model.pt, lr=0.001

# Still supports omitting the optional parameter
python train.py
# Output: checkpoint=None, lr=0.001

# Works seamlessly with other parameters
python train.py --checkpoint model.pt --learning-rate 0.01
# Output: checkpoint=model.pt, lr=0.01
```
Improved help output:
```
Before:  --checkpoint VALUE   Path to checkpoint file
After:   --checkpoint STR     Path to checkpoint file
```
Technical fixes:
- Fixed cli_parse.py:_get_union_classes(): Now filters out NoneType from Union type arguments
- Fixed cli_parse.py Union handling: Added detection for Optional[T] patterns (Union with single non-None type) and treats them as regular optional parameters
- Fixed type_utils.py:_get_type_name(): Now recognizes Optional[T] patterns and recursively extracts the correct inner type name for help text
- Help text now shows specific types (STR, INT, FLOAT) instead of generic VALUE for Optional types

Version 3.0.0-rc19 (2025-12-28)

🧪 Testing

Added comprehensive test cases for Optional[str] and Optional[int] CLI parsing
- Documents current limitation where Optional[T] types don't parse correctly with normal --param value syntax
- Tests verify expected behavior once the issue is fixed

📚 Documentation

New: Created dedicated Union Types documentation page (docs/key_concepts/union_types.md)
- Quick reference with 3 common patterns (Union selection, single class, optional parameters)
- Clear distinction between Union[ClassA, ClassB] and Optional[T]
- Detailed examples and syntax variations
- Documents Optional[str] limitation and workaround
Refactored: Streamlined cli_guide.md to reduce verbosity
- Moved verbose Union/Optional explanation to dedicated union_types.md
- Replaced with concise 3-line reference for quick navigation
- Maintains clarity while keeping main guide focused

Version 3.0.0-rc18 (2025-12-26)

🐛 Bug Fixes

EnvVar Instantiation Fix: Fixed @proto.prefix classes with EnvVar fields failing on instantiation with AttributeError: '_EnvVar' object has no attribute '__get__'.

The bug occurred because _EnvVar is callable (has __call__), so it was incorrectly detected as a method during instance creation. The fix uses precise method detection with inspect.isfunction and inspect.ismethod instead of the overly broad callable() check.
python
```
from params_proto import proto, EnvVar

@proto.prefix
class Config:
    host: str = EnvVar @ "HOST" | "localhost"

# Before fix: AttributeError: '_EnvVar' object has no attribute '__get__'
# After fix: works correctly
c = Config()
```

📚 Documentation

Added EnvVar + inheritance documentation and tests
Documented that inherited EnvVar fields are resolved and type-converted correctly

Version 3.0.0-rc17 (2025-12-26)

✨ Features

Inherited Fields: Support inherited fields in @proto classes. Parent class fields are now properly included in vars(), CLI args, and work with EnvVar.

python

class BaseConfig:
    host: str = EnvVar @ "HOST" | "localhost"
    port: int = EnvVar @ "PORT" | 8080

@proto.prefix
class AppConfig(BaseConfig):
    debug: bool = EnvVar @ "DEBUG" | False

Version 3.0.0-rc16 (2025-12-26)

🐛 Bug Fixes

Python 3.10 Support: Use Union[] syntax instead of | operator for type hints to support Python 3.10 (#17).

Version 3.0.0-rc15 (2025-12-19)

🐛 Bug Fixes

EnvVar dtype Conversion: Fixed EnvVar.get() to apply the dtype parameter for type conversion. Previously, the dtype was stored but not used when reading from environment variables, causing values to always be returned as strings.
python
```
import os
from params_proto import EnvVar

os.environ["PORT"] = "9000"

# Before fix: returned '9000' (str)
# After fix: returns 9000 (int)
port = EnvVar("PORT", dtype=int, default=8012).get()
```
All dtypes are now properly applied: int, float, bool, and str.

Version 3.0.0-rc10 (2025-12-17)

🐛 Bug Fixes

Classmethod/Staticmethod Support: Fixed @proto, @proto.cli, and @proto.partial decorators to properly handle @classmethod and @staticmethod descriptors. Previously, decorating methods with @proto would incorrectly include self/cls in CLI parameters or corrupt method calls.

Correct decorator order (proto decorator on the OUTSIDE):
python
```
class Trainer:
    @proto.cli      # proto.cli on OUTSIDE receives the descriptor
    @classmethod
    def train(cls, lr: float = 0.01):
        return cls.run_training(lr)

    @proto.cli
    @staticmethod
    def evaluate(model_path: str):
        return load_and_eval(model_path)
```
The decorators now:
- Detect classmethod/staticmethod descriptors via isinstance()
- Properly unwrap to get the underlying function signature
- Exclude cls parameter for classmethods automatically
- Implement descriptor protocol (__get__) for proper method binding
- Re-wrap results in classmethod()/staticmethod() for proto.partial
VAR_POSITIONAL/VAR_KEYWORD Handling: *args and **kwargs parameters are now properly excluded from CLI parameters by checking inspect.Parameter.kind instead of just parameter names.

Version 3.0.0-rc7 (2025-12-16)

✨ Features

Claude Skill: Added hierarchical Claude skill documentation in skill/ directory. Provides AI assistants with comprehensive params-proto knowledge including:
- Quick reference and cheat sheet
- API documentation for all decorators
- Feature guides (help generation, environment variables, sweeps)
- Common patterns and examples

Version 3.0.0-rc6 (2025-12-16)

🐛 Bug Fixes

Boolean Type Display: Boolean flags now show BOOL type in help text for consistency with other types (INT, STR, FLOAT).
```
--verbose BOOL       Enable verbose output (default: False)
--no-cuda BOOL       Use CUDA acceleration (default: True)
```

Version 3.0.0-rc5 (2025-12-16)

🐛 Bug Fixes

Boolean Flag Help Text: Fixed help text for boolean flags with default=True. Previously, --flag was shown in help even when the flag defaulted to True (making --flag a no-op). Now shows --no-flag for booleans defaulting to True, making it clear how to disable the feature.
python
```
@proto.cli
def train(cuda: bool = True):  # Use CUDA acceleration
    ...
```
Previously: --cuda Use CUDA acceleration (default: True) Now: --no-cuda Use CUDA acceleration (default: True)
ANSI Help Colorization: Fixed regex that incorrectly colored the first word of boolean flag descriptions as a type. Now only uppercase type names (INT, STR, FLOAT, BOOL) and enum choices ({A,B,C}) are colorized as types.

Version 3.0.0-rc4 (2025-12-14)

🐛 Bug Fixes

EnvVar Class Resolution: Fixed EnvVar not resolving at import time for class-based configs. Previously, accessing Config.ip on a @proto.prefix class with ip: str = EnvVar @ "VAR" | "default" would return the EnvVar object instead of the resolved value. Now EnvVar values are resolved at decoration time for both functions and classes.

Version 3.0.0 (Upcoming)

🎉 Major Release: Complete v3 Redesign

params-proto v3 is a complete rewrite focused on simplicity and modern Python type hints.

✨ Key Features

Documentation & Help Generation

Multi-line Documentation: Inline comments and docstring Args are now concatenated on separate lines for better readability

python

@proto.cli
def train(
    batch_size: int = 32,  # Training batch size
):
    """Args:
        batch_size: Controls memory usage and gradient noise
    """

Generates:

--batch-size INT     Training batch size
                     Controls memory usage and gradient noise (default: 32)

Automatic Help Generation: Parse inline comments (#) for parameter documentation
Docstring Args Support: Extract parameter docs from Google/NumPy-style docstrings
Smart Deduplication: Identical inline and docstring descriptions are shown only once
Auto-generated Descriptions: Fallback descriptions generated from parameter names

Environment Variables

EnvVar Support: Read configuration from environment variables with type conversion
Pipe Operator Syntax: Clean default value syntax: EnvVar @ "VAR" | default
Template Expansion: Support for $VAR, ${VAR}, and multiple variable substitution
Type Conversion: Automatic conversion of env var strings to annotated types

CLI Improvements

Function CLI: Decorate functions with @proto.cli for instant CLI programs
prog Parameter: Override script name for predictable help output in tests
Rich Type Support: Optional, List, Literal, Tuple, Enum, Path, Union types
Multi-namespace Configs: Use @proto.prefix for organized, modular configurations

🔄 API Changes

Decorator-based: @proto and @proto.cli instead of class inheritance
Type Hints Required: Full type hint support for IDE integration
Simplified Singleton: @proto.prefix for singleton configs
Function Support: First-class support for function-based configs

📚 Documentation

Complete Rewrite: All documentation updated for v3 API
Comprehensive Examples: Real-world ML training and RL agent examples
Migration Guide: Step-by-step guide for upgrading from v2
Type System Guide: Complete documentation of supported type hints

Version 2.13.2 (2025-08-03)

📚 Documentation

Major: Added comprehensive Sphinx documentation site with Furo theme
Added: Complete API documentation for proto, hyper, and utils modules
Added: Extensive tutorial collection covering:
- Basic usage patterns and CLI integration
- Advanced features including dynamic configs and validation
- Environment variables for flexible deployment
- Nested configurations for complex applications
- Hyperparameter sweeps and experiment management
Added: Read the Docs integration with automatic builds
Updated: Repository moved to geyang/params-proto with main branch as default

🔧 Configuration Management

Added: Documentation build targets to main Makefile (make docs, make preview)
Added: .readthedocs.yaml configuration for automated documentation builds

Version 2.13.0 (2025-01-15)

✨ Features

Added: Environment name checking to ensure all env names are defined in env string
Fixed: Dollar-sign handling in environment variable strings
Improved: Code formatting with ruff configuration
Removed: typing dependency (no longer needed for Python 3.8+)

🧹 Maintenance

Added: Comprehensive ruff configuration for code formatting
Added: PyCharm/IntelliJ IDE configuration files
Updated: Setup.py dependency management
Improved: Test coverage for environment variable parsing

Version 2.12.1 (2024-04-20)

🐛 Bug Fixes

Improved: Error traces for better debugging experience
Enhanced: Exception handling and error reporting

Version 2.12.0 (2023-12-21)

🧹 Maintenance

Removed: textwrap dependency (was causing import issues)
Fixed: README documentation formatting
Updated: Dependency cleanup

Version 2.11.x Series (2023)

The 2.11.x series focused on advanced parameter management and hierarchical configurations.

Version 2.11.16 (2023-09-03)

Added: Tree mode support for both Meta instances and ParamsProto instances
Improved: Nested dictionary handling with _tree attribute
Enhanced: Parameter attribute management

Version 2.11.14 (2023-07-24)

Fixed: __vars__ property for ParamsProto descendants
Improved: Property descriptor handling for dynamic attributes

Version 2.11.13 (2023-07-24)

Enhanced: vars(Args) functionality for better introspection
Added: Dynamic property access improvements

Version 2.11.12 (2023-07-24)

Changed: Args.property now returns property descriptor instead of value
Improved: Property-based parameter access patterns

Version 2.11.11 (2023-07-23)

Removed: Debug print statements from production code
Cleaned: Console output for cleaner user experience

Version 2.11.10 (2023-07-23)

Added: Parameter freezing once namespace is instantiated
Improved: Immutability patterns for configuration safety

Version 2.11.9 (2023-07-22)

Fixed: Function detection in parameter attributes
Improved: Dynamic attribute handling

Version 2.11.8 (2023-07-22)

Changed: Child detection moved to initialization time
Improved: Performance of nested configuration detection

Version 2.11.7 (2023-07-22)

Fixed: __new__ method super() call issues
Improved: Object instantiation patterns

Version 2.11.6 (2023-07-22)

Fixed: ParamsProto set as non-recursive Bear to prevent dict→Bear conversion
Improved: Nested attribute handling

Version 2.11.5 (2023-07-22)

Added: Support for object properties in configurations
Enhanced: Property-based parameter definitions

Version 2.11.4 (2023-07-22)

Improved: Internal attribute management
Enhanced: Performance optimizations

Version 2.11.3 (2023-07-21)

Fixed: Critical __vars__ property bug
Improved: Dictionary representation of configurations

Version 2.11.2 (2023-07-21)

Fixed: Argument parsing edge cases
Improved: CLI argument handling robustness

Version 2.11.1 (2023-07-21)

Added: Children attribute support for nested configurations
Enhanced: Hierarchical parameter management

Version 2.11.0 (2023-07-21)

Major: Added hierarchical hydration support
Added: Nested configuration update mechanisms
Enhanced: Multi-level parameter management

Older Versions

Version 2.12.0

Removed: textwrap dependency
Fixed: Documentation links and formatting

Version 2.11.x Series

Multiple bug fixes and improvements
Enhanced CLI argument parsing
Better error handling and validation

Version 2.10.x Series

Stability improvements
Performance optimizations
Bug fixes in parameter handling

Version 2.9.x Series

Enhanced hyperparameter sweep functionality
Improved nested configuration support
Better environment variable handling

Version 2.8.x Series (Long-term stable)

Core functionality stabilization
Extensive bug fixes and improvements
Enhanced CLI features

Version 2.4.0 - 2.8.0

Major feature additions
API improvements and stabilization
Enhanced documentation

Version 2.0.3 - 2.2.0

Initial stable releases
Core parameter management functionality
Basic CLI integration

Migration Guide

Upgrading to 2.13.x

The 2.13.x series introduces comprehensive documentation but maintains full backward compatibility. No code changes are required.

Key Changes Since 2.0.x

Environment Variables: Enhanced support for environment variable defaults and validation
Nested Configurations: Improved support for hierarchical parameter structures
Hyperparameter Sweeps: Advanced sweep functionality with the Sweep class
Type Safety: Better type hints and validation throughout the codebase
Documentation: Complete rewrite with examples and API reference

Breaking Changes

Version 2.13.0

Removed: typing dependency - ensure your Python version supports built-in type hints
Repository: Moved from episodeyang/params_proto to geyang/params-proto

Earlier Versions

See individual version notes above for specific breaking changes

Contributors

Ge Yang - Primary author and maintainer
Claude Code - Documentation generation and enhancement

Support

Documentation: https://params-proto.readthedocs.io/
Issues: https://github.com/geyang/params-proto/issues
Repository: https://github.com/geyang/params-proto