Skip to content

Latest commit

 

History

History
135 lines (93 loc) · 7.68 KB

CHANGELOG.md

File metadata and controls

135 lines (93 loc) · 7.68 KB
Raw

Changelog

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

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.2.2] - 2024-05-05

Added

  • Added support for llama-3-sonar-small-32k-chat, llama-3-sonar-large-32k-chat, llama-3-sonar-small-32k-online and llama-3-sonar-large-32k-online.

Changed

  • Updated default base models in config.py to use llama-3-sonar-large-32k-chat for chat functionalities and llama-3-sonar-large-32k-online for search functionalities.

Removed

  • Removed references to older models such as sonar-small-chat, sonar-medium-chat, sonar-small-online and sonar-medium-online in the code and documentation to streamline the use of newer llama-3-sonar models.
  • Removed references to codellama-70b-instruct, mistral-7b-instruct and mixtral-8x22b-instruct, as Perplexity Labs no loger supports the models.

[1.2.1] - 2024-04-22

Added

  • Added support for llama-3
  • Added syntax for exiting/quitting the program to the beginning of the conversation in perplexity.py
  • Added goodbye syntax to the end of the search request in perplexity.py

Changed

  • Changed variable names in config.py
  • Set default models to models from config or env in perplexity.py
  • Updated syntax in cli.py

Removed

  • Removed unused timeout in client.py

[1.2.0] - 2024-03-23

Added

  • config.py: Introduced a configuration management system to load environment variables and default settings.
  • client.py: Implemented a new client architecture for making API requests, including streaming support.

Improved

  • loading.py: No changes detected in functionality, the code remains identical.
  • perplexity.py: Enhanced with a modular approach separating chat and search functionalities into distinct classes with improved error handling.
  • cli.py: Major overhaul to CLI interface, incorporating new options and better help documentation, facilitating both chat and search operations with the new client architecture.

Changed

  • Environment variable management has been centralized and now requires specific keys (PERPLEXITY_API_KEY, PERPLEXITY_DEFAULT_CHAT_MODEL, PERPLEXITY_DEFAULT_SEARCH_MODEL, PERPLEXITY_BASE_URL, PERPLEXITY_TIMEOUT), but has defaults already set for ease of use.
  • The base API interaction logic has been encapsulated within the Client class, abstracting the complexities of request handling, including streaming.

Removed

  • pplx.py, pplx_cli.py, base_api.py, pplx_search.py, and pplx_chat.py have been replaced with the new client.py, config.py, perplexity.py, and cli.py files, indicating a structural overhaul.
  • The direct dependency on .env file loading within API wrapper files has been removed, now managed centrally in config.py.

Fixed

  • The handling of API keys and configuration settings has been standardized, fixing inconsistencies in how environment variables were previously managed.

[1.1.2] - 02/25/2024

base_api.py

Improved

  • Enhanced error handling in the post method to provide clearer error messages and gracefully exit the application upon encountering a critical error, improving user experience and debuggability.
  • Added a finally block to ensure that the loading animation is always stopped, even if an error occurs, preventing potential terminal display issues.
  • Modified the loading animation to start only in non-streaming requests to avoid overlap with streaming output, enhancing output readability.
  • Streamlined the API key retrieval process with a more descriptive error message if the API key is not found, aiding users in configuration setup.

Fixed

  • Fixed an issue where the loading animation could potentially continue running or the cursor remained hidden if an exception was thrown during a request.
  • Addressed a potential bug by ensuring the terminal is cleared only on error, preserving user input and previous interactions for reference.

pplx_cli.py

Added

-Introduced a custom argparse formatter CustomFormatter combining ArgumentDefaultsHelpFormatter and RawDescriptionHelpFormatter to improve the help text readability.

  • Implemented detailed command descriptions and examples in the CLI help output, providing immediate guidance to users without external documentation.

Changed

  • Unified the -a, --api_key argument declaration across both chat and search commands to improve code maintainability.
  • Implemented a shared function to add common arguments to both chat and search parsers, reducing code duplication.
  • Updated argument descriptions for enhanced clarity, making it easier for users to understand the purpose and usage of each command.
  • Modified all argument metavariables to an empty string, streamlining the help output by removing uppercase type hints for a cleaner interface.

Removed

  • Eliminated redundant argument declarations, specifically for --api_key in both chat and search subparsers, centralizing its declaration for cleaner code.

[1.1.1] - 02/23/2024

Added

  • Support for Perplexity Labs latest sonar-small-chat, sonar-small-online, sonar-medium-chat, and sonar-medium-online AI models offering improvements in cost-efficiency, speed, and performance.
  • Extended context window support, now accommodating up to 16k tokens for models like mixtral-8x7b-instruct and all Perplexity models.
  • Increased public rate limits across all models to accommodate approximately 2x more requests.

Warning

On March 15, the pplx-70b-chat, pplx-70b-online, llama-2-70b-chat, and codellama-34b-instruct models will no longer be available through the Perplexity API.

[1.1.0] - 02/22/2024

Added

  • loading.py for implementing a loading spinner, enhancing user experience during network requests.
  • base_api.py introducing BaseAPI class for shared API functionality, including request handling and streaming support.
  • pplx_chat.py and pplx_search.py classes, extending BaseAPI to separate concerns for chat and search functionalities.
  • Detailed error handling and environmental variable support for API key configuration, increasing usability and flexibility.
  • A comprehensive command-line interface setup in pplx_cli.py, facilitating the use of both chat and search functionalities through a unified interface.

Changed

  • Modularized the codebase into separate files (pplx_cli.py, loading.py, base_api.py, pplx_chat.py, pplx_search.py), improving code organization and maintainability.
  • Enhanced the command-line interface with more detailed options, including model selection, temperature, top_p, top_k, presence penalty, and frequency penalty settings, allowing for a more customized user experience.
  • Updated the streaming functionality to use a loading spinner, providing real-time feedback during asynchronous operations.
  • Improved API key management by supporting environmental variables and .env files, simplifying configuration.

Removed

  • The single-file script structure, replacing it with a more modular and scalable project architecture.
  • Direct use of requests and json in the CLI functions, moving this logic to the BaseAPI class to reduce redundancy.

Fixed

  • Fixed issue where streaming was not working.
  • Fixed issue where command line parameters were not being set.
  • Fixed argument parsing structure in pplx_cli.py with subparsers for chat and search commands, enabling a more structured and versatile command-line interface.

Security

  • Implemented secure API key handling through environment variables and .env files, reducing the risk of key exposure.

[1.0.1] - 01/30/2024

Added

  • Added support for new model "codellama-70b-instruct" by Meta.

[1.0.0] - 01/10/2024

Added

  • Initial release.