search

deepset-ai / haystack

:mag: AI orchestration framework to build customizable, production-ready LLM applications. Connect components (models, vector DBs, file converters) to pipelines or agents that can interact with your data. With advanced retrieval methods, it's best suited for building RAG, question answering, semantic search or conversational agent chatbots.
https://haystack.deepset.ai
Apache License 2.0
16.94k stars 1.85k forks source link

docs: Pipeline.inputs() #8271

Closed TuanaCelik closed 1 week ago

TuanaCelik commented 1 month ago

Once a pipeline is created, it's difficult for users to know how they should run the pipeline. We have quite a useful utility function for pipelines which is .inputs() which lists all the expected/required inputs for the components.

We should use this in our docs heavily imo. This function is hardly visible, and we don't really provide any help on how a pipeline should be run other than trial and errors. Once a pipeline fails, the error us quite handy. And if a user knows to ues .show() that is also handy. But I think we should add this everywhere as something that a user colud make use of while creating/running pipelines.

  1. In the pipelines doc
  2. In all component docs where we have example pipelines..
aantti commented 1 month ago

I'd love that. Specifically, btw, struggled recently with OpenAITextEmbedder trying to figure out the correct parameter ('text').

dfokina commented 2 weeks ago

I have added the following paragraph to the docs that should be well-searchable for pipeline inputs or component inputs: https://docs.haystack.deepset.ai/docs/creating-pipelines#pipeline-inputs