Skip to content

Latest commit

 

History

History
203 lines (154 loc) · 8.39 KB

CHANGELOG.md

File metadata and controls

203 lines (154 loc) · 8.39 KB
Raw

Changelog

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

v2.1.4

  • Fixes

    • CUDA compatibility

  • Internal

    • Added a GitHub Action to push automatically on PyPI on a release

  • Doc

    • Added delivery method to the contributing file

v2.1.3

  • Features

    • Added additionnal_headers in Model to add specific values in the header of the request send to the API.
    • ask can now be used without specify the user parameter name.

  • Fixes

    • Fixed the bug with the f-string in python 3.10 or below.
    • Removed simple_api_call.
    • Hosta now support a nested pydantic type like List[PydanticType].

  • Doc

    • Added two example to work with Microsoft Azure and Ollama.
    • Added documentation for the emulate functions.
    • Remplace all links in the README by absolute link

v2.1.2

  • Doc

    • Fix the example for the custom Model class with a Llama Model

  • Fixes

    • Remove the PIL dependency

v2.1.1

  • Enhancement

    • Changed examples syntax in the system prompt for better performances.

  • Doc

    • Added an example for a custom Model class with an Llama model.

v2.0-beta2

  • Added
    • Introduced BaseEncoder as an Abstract Base Class (ABC) for all specific encoders, providing a standardized interface.
    • Added BoolEncoder to handle encoding of boolean values.
    • Introduced PredictConfig data class to encapsulate parameters for the predict function.
    • Initial implementation of the HostaCache class in core/cache.py with generic type support.
    • Initial implementation of the ModelCachedData and HostaModelCache classes in predict/cache.py.
  • Changed
    • Refactored IntEncoder and FloatEncoder to inherit from BaseEncoder and implement the encode method.
    • Updated HostaEncoder to use a dictionary (self.encoders) for mapping data types to their corresponding encoders. This allows for easier extension and maintenance.
    • Improved type handling in HostaEncoder.encode method for better extensibility and readability.
    • Refactored predict function to accept a single PredictConfig object instead of multiple parameters. This change improves readability and maintainability of the function.
  • Fixed
    • Enhanced exception handling in FloatEncoder to provide more informative error messages.
    • Removed unnecessary constructors from encoder classes, streamlining the code.

v2.0-beta1

  • Features

    • Added max_tokens args for emulate
    • Added use_locals/self_as_ctx args for emulate for clarity and modularity
    • thought is now used to add chain of thoughts in a emulated function
    • PromptManager is now available for users to change all meta-prompt.

  • Changes

    • thought function become thinkof
    • suggest is removed
    • creativity become temperature and diversity become top_p
    • There's no more hostacache for emulated functions

OpenHosta v1.2.1 - 10/14/24

  • Fixes
    • OpenHosta now can't be used with Python 3.13 due to the incompatibility with pytorch

v1.2.0 - 10/14/2024

  • New Features

    • predict Function
      The predict function is now available, allowing you to create internal models—currently supporting linear regression—based on user-provided training data. This simplifies model generation without relying on external APIs. Key functionalities include:

      • .retrain: Retrain models with specified parameters.
      • .continue_train: Continue training with existing model weights.
      • .emulate: Run predictions through an LLM or create a model directly using internal linear regression based on training data.

    • TrainingSet Management
      Manage training datasets effortlessly with new tools:

      • .visualize: Inspect current data visually.
      • .add: Add new examples.

  • Enhancements

    • Expanded Dataset Support:
      load_training_example (previously load_examples) supports JSON, JSONL, and CSV formats for easier integration.

    • Verbose Mode in predict:
      Track detailed model training and define target losses with get_loss.

v1.1.1 10/07/24

  • Features

    • Added _last_request attributs to the Model object

  • Optimization

    • Reduce the emulate's user prompt to only the function call string and move the rest in the system prompt.
    • Added markdow header in emulate prompt and enhance strucure
    • Remove unecessary setence like confidence level and transition

  • Internal

    • Refacto the prompt building function in emulate
    • Remove the validate function in LLM call response handler

  • Fixes

    • Added code block syntax for the function definition.
    • Re-added diagramm attributs but decrepated
    • Added explicitly a neutral response in the emulate prompt (None)

v1.1-rc4 09/27/24

  • Feature

    • Added suggest function. Works the same as the __suggest__ attributs but in a function

  • Doc

    • Many inconsistencies and errors corrected.

  • Internal

    • Added Github workflows for linting, formating and testing when pushing to dev and main

  • Fixes

    • suggest attribute diagramm is now diagram

v1.1-rc3 09/26/23

  • Fixes

    • emulate now works when emulated function is called inside another one
    • emulate works now with lru_cache decorator from functools module

  • Internal

    • Added custom Exception classes for request and frame errors.
    • Added a loop to find the frame in _extend_scope
    • Added a Makefile for cleaning and packaging and tests

v1.1 09/13/2024

  • Fixes

    • the emulate function is now decorator-resistant.
    • The function config.set_default_model works now w/ config.set_default_apiKey
    • thought function now accept multiple arguments
    • suggest and analytics call now th LLM with Model class (ai_call)
    • emulate works now in a nested function.
    • Added Flask compatibility

  • Enhancement

    • the thought function guess automatically his return type with LLM
    • suggest return now his data returned by LLM in a dict
    • emulate works now for class methods
    • emulate now integrate locals variable of emulated function to the LLM prompt.
    • _hostacache_ are now available for function infos storage
    • Added support for the typing module: You can now use specific return types from the typing module, including [List, Dict, Tuple, Set, FrozenSet, Deque, Iterable, Sequence, Mapping, Union, Optional, Literal].
    • emulate now includes a verification output that checks and converts the output to the specified type if necessary. Supported types include Pydantic models, all types from the typing module mentioned above, as well as built-in types such as dict, int, float, str, list, set, frozenset, tuple, and bool.
      • The complex type is not supported. If you need to use complex numbers, please pass them as a tuple and manually convert them after processing.
    • The return type predicted by thought is now attached as an attribute to the object created: _return_type.
    • Added a _last_request attribute to an emulated function giving access to the prompt sent to the LLM.

  • Features

    • example function that can add some example for a specified hosta-injected function (inside or outside it) and add it to the cache
    • save_examples function that can save in a JSONL file all he example of an hosta-injected function
    • load_examples function that can load an example file an a cache for an hosta-injected function
    • set_prompt in PromptManager enables to change automatically a prompt un "prompt.json"

  • Optimization

    • emulate prompt changed: confidence level removed: (~20% speed gained) (emulate: v1.1)
    • All prompt enhanced: added code block, separator and reference to the name of the processed function (emulate: v1.2, enhance v1.2, thought: v1.1, estimate: v1.1)

  • Internal

    • Functional tests have been added for each of the library's features

  • Doc

    • Documentation now integrates Google Cobal link

v1.0 08/29/2024:

  • Features
    • Function emulate to emulate a function by LLM.
    • Pydantic typing compatibility with emulate.
    • Function thought replicating the behavior of a lambda.
    • Configuration system allowing all LLM with class inheritance.
    • __suggest__ attribute giving advice to improve a prompt.
  • Fixes
  • Changes
  • Optimizations
  • Internal