How can i add my Faiss document store to the haystack Rest API???

[ali-issa9]

Hello, I am trying to create a Haystack rest API using this tutorial. And they were using ElasticSearch as their document store. i have already downloaded
the project as they mentioned using docker. But I don't know how can I instance of FAISS with haystack rest API.

I created an instance of FAISS using these two lines of code :

haystack.document_stores import FAISSDocumentStore
document_store = FAISSDocumentStore(faiss_index_factory_str="Flat",similarity='cosine')

after writing the data inside FAISS, I saved them using this code :

document_store.save("Demo")

does anyone know how we can use this saved instance with haystack rest API?

thank you.

[ZanSara]

Hey @ali-issa9, the Haystack REST API use Pipeline YAML files to load a pipeline. Have you tried to modify your Pipeline YAML to load your existing FAISSDocumentStore? I am referring to this file: https://github.com/deepset-ai/haystack/blob/main/rest_api/rest_api/pipeline/pipelines.haystack-pipeline.yml If yes, make sure it's being loaded properly by the REST API by using volumes.

[ali-issa9]

Hello @Suhaib441 do you have any idea how we can use BioGPT as a retriever?

[Suhaib441]

Hello @ali-issa9
i haven't tried BioGPT yet but im using other retrieval model i think you can use BioGPT in same way

[Suhaib441]

Hello @ZanSara the error caused by the indentation of some parameter, So the pipeline load_from_yaml works fine but i got empty answer list. it looks like the pipeline can not find my Faiss index. Do you know how force the path of an index in Yaml file ?

{'answers': [],
'documents': [],
'root_node': 'Query',}
Thank You

[ali-issa9]

hello @Suhaib441 i tried to load it and set it as a retriever but it seems that the model should be loaded in a different way. let me know if you find a way to use it. thank you

[Suhaib441]

try to load and run a simple pipeline in a different script or notebook, i have the same bug when im using a YAML file
"Failed loading pipeline component 'Retriever'. See the stacktrace above for more information"

but it works in jupyter notebook like here https://haystack.deepset.ai/tutorials/06_better_retrieval_via_embedding_retrieval