Shorter model name params #840
Conversation
|
What do you think to release new versions following Semantic Versioning https://semver.org? If the release introduces breaking change, then bumping major version should help to catch potential problems earlier. |
|
@qarol you're probably right - a breaking change should be a MAJOR version change according to server. @andreibondarev I'll leave that decision to you though... |
CHANGELOG.md
Outdated
| - Deprecate Langchain::LLM::GooglePalm | ||
| - Allow setting response_object: {} parameter when initializing supported Langchain::LLM::* classes | ||
| - Allow setting response_object: {} parameter when initializing supported Langchain::LLM::\* classes |
There was a problem hiding this comment.
Why do we need the slash?
There was a problem hiding this comment.
Weird - I'm not sure why that's there. I think Prettier added it or something. Removing ...
lib/langchain/llm/aws_bedrock.rb
Outdated
| chat_completion_model_name: "anthropic.claude-v2", | ||
| completion_model_name: "anthropic.claude-v2", | ||
| chat_model: "anthropic.claude-v2", | ||
| completion_model: "anthropic.claude-v2", | ||
| embeddings_model_name: "amazon.titan-embed-text-v1", |
There was a problem hiding this comment.
I'm thinking maybe we rename this as well, to embedding_model?
|
@bborn Quick thought -- maybe since the 3 methods are called: def embed
def complete
def chatwe should call the defaults: What do you think? |
|
@andreibondarev I think that makes sense. Updated. My only concern is that |
@qarol I love that! Given that we're pre-1.0.0, do you suggest that we keep incrementing minor versions, 0.18.0, 0.19.0, 0.20.0, etc.? |
|
@bborn Great PR! Thank you! |
@andreibondarev question is, what things blocks from releasing 1.0.0 ;) For pre-1.0.0 it's acceptable to bump just minor version no matter if it's a breaking change or new feature. For bugfixes increasing patch number should be enough IMO |
Fixes #820:
chat_completion_model_nametochat_modelcompletion_model_nametocompletion_modelembeddings_model_nametoembedding_modelThis is a breaking change, and I updated the changelog and bumped the version to reflect that.