llmflows

llmflows copied to clipboard

### The problem you are trying to solve: I would like to use Weaviate just like I am able to easily use Pinecone in flows. ### Suggested new feature or...

stoyan-stoyanov

### The problem you are trying to solve: I would like to use Chroma just like I am able to easily use Pinecone in flows. ### Suggested new feature or...

stoyan-stoyanov

### The problem you are trying to solve: I want the assistants I build to be able to call functions based on the contents in a conversation. OpenAI's has released...

stoyan-stoyanov

style: use type hints for input arguments

1

https://github.com/stoyan-stoyanov/llmflows/blob/6ce79a8cc3b06e799d2b6eb6f6e8e78354f68dcd/llmflows/llms/openai.py#L55 It would be much appreciated if you used type hints for the input arguments in your methods - with this suggestion extended to the whole code base.

inspiralpatterns

https://github.com/stoyan-stoyanov/llmflows/blob/6ce79a8cc3b06e799d2b6eb6f6e8e78354f68dcd/llmflows/llms/openai_embeddings.py#L50-L52 I am not a fan of functions that can return either one or the other type, however, I assume that given a `List[VectorDoc]`, `generate` will always return a `List[VectorDoc]`....

inspiralpatterns

https://github.com/stoyan-stoyanov/llmflows/blob/6ce79a8cc3b06e799d2b6eb6f6e8e78354f68dcd/llmflows/llms/llm_utils.py#L19

vickychudinov

https://github.com/stoyan-stoyanov/llmflows/blob/6ce79a8cc3b06e799d2b6eb6f6e8e78354f68dcd/llmflows/flows/chat_flowstep.py#L96 I find using `Any` everywhere generally frowned upon: you want to use type hints to leverage the power of a static type checker such as e.g. `mypy` _and_ for...

inspiralpatterns

FunctionalFlowStep

2

Why not use **kwargs here? https://github.com/stoyan-stoyanov/llmflows/blob/3c4ceb7964dd85d0ab59f7865aa37337201454fa/llmflows/flows/functional_flowstep.py#L42 Then you can name keyword args to the function (i.e., `generate`), and the remaining will be the arguments for `flowstep_fn`, so this will go...

OlaRonning

refactor: why not just using containers?

4

https://github.com/stoyan-stoyanov/llmflows/blob/6ce79a8cc3b06e799d2b6eb6f6e8e78354f68dcd/llmflows/llms/openai_embeddings.py#L66-L68 If the whole point of having `VectorDoc` vs `List[VectorDoc]` is to put the first into a list and for the generated output to be accessed at index 0, why...

inspiralpatterns