docs/integrations/retrievers/ragatouille/

[giscus[bot]]

docs/integrations/retrievers/ragatouille/

RAGatouille makes it as simple as can be to use ColBERT!

https://python.langchain.com/docs/integrations/retrievers/ragatouille/

[xupeka]

At the moment there is no easy way to change the directory where it stores indexes. I had to do below to set my own path

colbert = RAGPretrainedModel.from_pretrained("colbert-ir/colbertv2.0")
colbert.model.config.index_path=f"{project_name}/colbert_db/indexes/my_index"

colbert.model.run_config.overwrite=True
colbert.model.run_config.root=f"{project_name}"
colbert.model.run_config.experiment = ""

colbert.index(
    collection=collection,
    document_ids=document_ids,
    document_metadatas=document_metadatas,
    index_name="my_index",
    overwrite_index=True,
    document_splitter_fn=None,
    preprocessing_fn=None,
)

colbert_path=f"{project_name}/colbert_db/indexes/my_index"
colbert_database = RAGPretrainedModel.from_index(colbert_path)

ret=colbert_database.as_langchain_retriever(k=3)
ret.invoke("Ask you query")