How I use a custom component in internal project?

[lohit8846]

Hello,

I would like to explore using haystack-ai and hayhooks in an internal project within my company. However, I noticed that the documentation is geared around this concept of "Integrations" which is not applicable for internal custom components that are developed in house and not released publicly

Are there any limitation to developing custom components? My concern is that not all features listed in the documentation will be supported. For example, how does pipeline serialization work when having custom components? I saw in the source code that this type is validated in some registry, but it's not clear how a custom component is supported there

I see that types in the YAML are defined like below. So what would a custom component defined in an internal package look like? Also does hayhooks still work given that I'm able to install an internal package or is there any special importing within the python code required. Any input on this is appreciated. Thank you

type: haystack_integrations.document_stores.qdrant.document_store.QdrantDocumentStore

[vblagoje]

Hey @lohit8846 the whole point of Haystack 2 is to allow developers to easily plug in their own components in order to build NLP/AI apps. For custom component serialization all you need to do is to implement to_dict and from_dict contract. We are now adding callbacks to serialization to further customize your serialization if needed. Let us know if you have any additional questions.