v0.0.35

0.0.35 - October 22, 2024

Added

• Support for Anthropic's updated Claude 3.5 Sonnet released today

Changed

• claude-3.5-sonnet now points to version claude-3-5-sonnet-latest

v0.0.34

0.0.34 - September 30, 2024

Caution

This release has breaking changes! Please read the changelog carefully.

Added

• New supported models gemma-2-9b, llama-3.2-1b, and llama-3.2-3b via Groq.

Changed

• In order to be more consistent with l2m2's naming scheme, the following model ids have been updated:
  • llama3-8b → llama-3-8b
  • llama3-70b → llama-3-70b
  • llama3.1-8b → llama-3.1-8b
  • llama3.1-70b → llama-3.1-70b
  • llama3.1-405b → llama-3.1-405b
• This is a breaking change!!! Calls using the old model_ids (llama3-8b, etc.) will fail.

Removed

• Provider octoai has been removed as they have been acquired and are shutting down their cloud platform. This is a breaking change!!! Calls using the octoai provider will fail.
  • All previous OctoAI supported models (mixtral-8x22b, mixtral-8x7b, mistral-7b, llama-3-70b, llama-3.1-8b, llama-3.1-70b, and llama-3.1-405b) are still available via Mistral, Groq, and/or Replicate.

v0.0.33

0.0.33 - September 11, 2024

Changed

• Updated gpt-4o version from gpt-4o-2024-05-13 to gpt-4o-2024-08-06.

v0.0.32

0.0.32 - August 5, 2024

Added

• Mistral provider support via La Plateforme.

• Mistral Large 2 model availibility from Mistral.

• Mistral 7B, Mixtral 8x7B, and Mixtral 8x22B model availibility from Mistral in addition to existing providers.

• 0.0.30 and 0.0.31 are skipped due to a packaging error and a model key typo.

v0.0.29

0.0.29 - August 4, 2024

Caution

This release has breaking changes! Please read the changelog carefully.

Added

• alt_memory and bypass_memory have been added as parameters to call and call_custom in LLMClient and AsyncLLMClient. These parameters allow you to specify alternative memory streams to use for the call, or to bypass memory entirely.

Changed

• Previously, the LLMClient and AsyncLLMClient constructors took memory_type, memory_window_size, and memory_loading_type as arguments. Now, it just takes memory as an argument, while window_size and loading_type can be set on the memory object itself. These changes make the memory API far more consistent and easy to use, especially with the additions of alt_memory and bypass_memory.

Removed

• The MemoryType enum has been removed. This is a breaking change!!! Instances of client = LLMClient(memory_type=MemoryType.CHAT) should be replaced with client = LLMClient(memory=ChatMemory()), and so on.

v0.0.28

[client] add default provider activation

v0.0.27

v0.0.27 (#8)

v0.0.26

[models] add gpt-4o-mini

v0.0.25

[exceptions] add LLMRateLimitError

v0.0.24

[feature] add timeout option to call and call_custom