Skip to contents

Changelog

Source: NEWS.md

Development Version 0.2.5

Major Features

Better embedding functions with improved output and error handling and new documentation. New article on using embeddings with tidyllm. Support for embedding models on azure with azure_openai_embedding()

Breaking Changes

  • The output format of embed() and the related API-specific functions was changed from a matrix to a tibble with an input column and a list column containing one embedding vector and one input per row.

Version 0.2.4

Refinements of the new interface

One disadvantage of the first iteration of the new interface was that all arguements that needed to be passed to provider-specific functions, were going through the provider function. This feels, unintuitive, because users expect common arguments (e.g., .model, .temperature) to be set directly in main verbs like chat() or send_batch().Moreover, provider functions don’t expose arguments for autocomplete, making it harder for users to explore options. Therefore, the main API verbs now directly accept common arguements, and check them against the available arguements for each API.

Bug-fixes

  • New error message for not setting a provider in main verbs
  • Missing export of main verbs fixed
  • Wrong documentation fixed

Version 0.2.3

Major Interface Overhaul

tidyllm has introduced a verb-based interface overhaul to provide a more intuitive and flexible user experience. Previously, provider-specific functions like claude(), openai(), and others were directly used for chat-based workflows. Now, these functions primarily serve as provider configuration for some general verbs like chat().

Key Changes:

Each verb and provider combination routes the interaction to provider-specific functions like openai_chat() or claude_chat() that do the work in the background. These functions can also be called directly as an alternative more verbose and provider-specific interface.

Old Usage:

llm_message("Hello World") |>
  openai(.model = "gpt-4o")

New Usage:

# Recommended Verb-Based Approach
llm_message("Hello World") |>
  chat(openai(.model = "gpt-4o"))
  
# Or even configuring a provider outside
my_ollama <- ollama(.model = "llama3.2-vision:90B",
       .ollama_server = "https://ollama.example-server.de",
       .temperature = 0)

llm_message("Hello World") |>
  chat(my_ollama)

# Alternative Approach is to use more verbose specific functions:
llm_message("Hello World") |>
  openai_chat(.model = "gpt-4o")
  • Backward Compatibility:
    • The old functions (openai(), claude(), etc.) still work if you directly supply an LLMMessage as arguement, but issue deprecation warnings when used directly for chat.
    • Users are encouraged to transition to the new interface for future-proof workflows.

Version 0.2.2

Major Features

  • Added functions to work with the Google Gemini API, with the new gemini() main API-function
  • Support for the file upload workflows for Gemini:
#Upload a file for use with gemini
upload_info <- gemini_upload_file("example.mp3")

#Make the file available during a Gemini API call
llm_message("Summarize this speech") |>
  gemini(.fileid = upload_info$name)
  
#Delte the file from the Google servers
gemini_delete_file(upload_info$name)
  • Brings video and audio support to tidyllm
  • Google Gemini is the second API to fully support tidyllm_schema()
  • gemini()-requests allow for a wide range of file types that can be used for context in messages
  • Supported document formats for gemini() file workflows:
    • PDF: application/pdf
    • TXT: text/plain
    • HTML: text/html
    • CSS: text/css
    • Markdown: text/md
    • CSV: text/csv
    • XML: text/xml
    • RTF: text/rtf
  • Supported code formats for gemini() file workflows:
    • JavaScript: application/x-javascript, text/javascript
    • Python: application/x-python, text/x-python
  • Supported image formats for gemini() file workflows:
    • PNG: image/png
    • JPEG: image/jpeg
    • WEBP: image/webp
    • HEIC: image/heic
    • HEIF: image/heif
  • Supported video formats for gemini() file workflows:
    • MP4: video/mp4
    • MPEG: video/mpeg
    • MOV: video/mov
    • AVI: video/avi
    • FLV: video/x-flv
    • MPG: video/mpg
    • WEBM: video/webm
    • WMV: video/wmv
    • 3GPP: video/3gpp
  • Supported audio formats for gemini() file workflows:
    • WAV: audio/wav
    • MP3: audio/mp3
    • AIFF: audio/aiff
    • AAC: audio/aac
    • OGG Vorbis: audio/ogg
    • FLAC: audio/flac

Version 0.2.1

Major Features:

  • Added get_metadata() function to retrieve and format metadata from LLMMessage objects.
  • Enhanced the print method for LLMMessage to support printing metadata, controlled via the new tidyllm_print_metadata option or a new .meta-arguement for the print method.
conversation <- llm_message("Write a short poem about software development") |>
  claude()
  
#Get metdata on token usage and model as tibble  
get_metadata(conversation)

#or print it with the message
print(conversation,.meta=TRUE)

#Or allways print it
options(tidyllm_print_metadata=TRUE)

Bug-fixes:

  • Fixed a bug in send_openai_batch() caused by a missing .json-arguement not being passed for messages without schema

Version 0.2.0

CRAN release: 2024-11-07

New CRAN release. Largest changes compared to 0.1.0:

Major Features:

  • Batch Request Support: Added support for batch requests with both Anthropic and OpenAI APIs, enabling large-scale request handling.
  • Schema Support: Improved structured outputs in JSON mode with advanced .json_schema handling in openai(), enhancing support for well-defined JSON responses.
  • Azure OpenAI Integration: Introduced azure_openai() function for accessing the Azure OpenAI service, with full support for rate-limiting and batch operations tailored to Azure’s API structure.
  • Embedding Model Support: Added embedding generation functions for the OpenAI, Ollama, and Mistral APIs, supporting message content and media embedding.
  • Mistral API Integration: New mistral() function provides full support for Mistral models hosted in the EU, including rate-limiting and streaming capabilities.
  • PDF Batch Processing: Introduced the pdf_page_batch() function, which processes PDFs page by page, allowing users to define page-specific prompts for detailed analysis.
  • Support for OpenAI-compatible APIs: Introduced a .compatible argument (and flexible url and path) in openai() to allow compatibility with third-party OpenAI-compatible APIs.

Improvements:

  • API Format Refactoring: Complete refactor of to_api_format() to reduce code duplication, simplify API format generation, and improve maintainability.
  • Improved Error Handling: Enhanced input validation and error messaging for all API-functions functions, making troubleshooting easier.
  • Rate-Limiting Enhancements: Updated rate limiting to use httr2::req_retry() in addition to the rate-limit tracking functions in tidyllm, using 429 headers to wait for rate limit resets.
  • Expanded Testing: Added comprehensive tests for API functions using httptest2

Breaking Changes:

  • Redesigned Reply Functions: get_reply() was split into get_reply() for text outputs and get_reply_data() for structured outputs, improving type stability compared to an earlier function that had different outputs based on a .json-arguement.
  • Deprecation of chatgpt(): The chatgpt() function has been deprecated in favor of openai() for feature alignment and improved consistency.

Minor Updates and Bug Fixes:

  • Expanded PDF Support in llm_message(): Allows extraction of specific page ranges from PDFs, improving flexibility in document handling.
  • New ollama_download_model() function to download models from the Ollama API
  • All sequential chat API functions now support streaming

Version 0.1.11

Major Features

  • Support for both the Anthropic and the OpenAI batch request API added
  • New .compatible-arguement in openai() to allow working with compatible third party APIs

Improvements

  • Complete refactor of to_api_format(): API format generation now has much less code duplication and is more maintainable.

Version 0.1.10

Breaking Changes

Improvements

  • Rate limiting updated to use httr2::req_retry(): Rate limiting now uses the right 429 headers where they come.

Version 0.1.9

Major Features

  • Enhanced Input Validation: All API functions now have improved input validation, ensuring better alignment with API documentation

  • Improved error handling More human-readable error messages for failed requests from the API

  • Advanced JSON Mode in openai(): The openai() function now supports advanced .json_schemas, allowing structured output in JSON mode for more precise responses.

  • Reasoning Models Support: Support for O1 reasoning models has been added, with better handling of system prompts in the openai() function.

  • Streaming callback functions refactored: Given that the streaming callback format for Open AI, Mistral and Groq is nearly identical the three now rely on the same callback function.

Breaking Changes

  • chatgpt() Deprecated: The chatgpt() function has been deprecated in favor of openai(). Users should migrate to openai() to take advantage of the new features and enhancements.

Improvements

  • Better Error Handling: The openai(), ollama(), and claude() functions now return more informative error messages when API calls fail, helping with debugging and troubleshooting.

Version 0.1.8

Major Features

  • Embedding Models Support: Embedding model support for three APIs:
    • Embedding functions process message histories and combine text from message content and media attachments for embedding models.
    • ollama_embedding() to generate embeddings using the Ollama API.
    • openai_embedding() to generate embeddings using the OpenAI API.
    • mistral_embedding() to generate embeddings using the Mistral API.

Improvements

  • PDF Page Support in llm_message(): The llm_message() function now supports specifying a range of pages in a PDF by passing a list with filename, start_page, and end_page. This allows users to extract and process specific pages of a PDF.

Version 0.1.7

Major Features

  • PDF Page Batch Processing: Introduced the pdf_page_batch() function, which processes PDF files page by page, extracting text and converting each page into an image, allowing for a general prompt or page-specific prompts. The function generates a list of LLMMessage objects that can be sent to an API and work with the batch-API functions in tidyllm.

Version 0.1.6

Major Features

  • Support for the Mistral API: New mistral() function to use Mistral Models on Le Platforme on servers hosted in the EU, with rate-limiting and streaming support.

Version 0.1.5

Major Features

  • Message Retrieval Functions: Added functions to retrieve single messages from conversations:

Improvements

  • Easier Troubleshooting in API-function: All API functions now support the .dry_run argument, allowing users to generate an httr2-request for easier debugging and inspection.
  • API Function Tests: Implemented httptest2-based tests with mock responses for all API functions, covering both basic functionality and rate-limiting.

Version 0.1.4

Major Features

  • New Ollama functions:
    • Model Download: Introduced the ollama_download_model() function to download models from the Ollama API. It supports a streaming mode that provides live progress bar updates on the download progress.

Improvements

Version 0.1.3

Major Features

  • The groq() function now supports images.
  • More complete streaming support across API-functions.

Breaking Changes

  • Groq Models: System prompts are no longer sent for Groq models, since many models on Groq do not support them and all multimodal models on Groq disallow them.

Version 0.1.2

Improvements

  • New unit tests for llm_message().
  • Improvements in streaming functions.

Version 0.1.1

Major Features

  • JSON Mode: JSON mode is now more widely supported across all API functions, allowing for structured outputs when APIs support them. The .json argument is now passed only to API functions, specifying how the API should respond, and it is not needed anymore in last_reply().

  • Improved last_reply() Behavior: The behavior of the last_reply() function has changed. It now automatically handles JSON replies by parsing them into structured data and falling back to raw text in case of errors. You can still force raw text replies even for JSON output using the .raw argument.

Breaking Changes

  • last_reply(): The .json argument is no longer used, and JSON replies are automatically parsed. Use .raw to force raw text replies.