docs: small updates to astra vectorize docs #2497
Conversation
dosubot
bot
added
size:M
| name="provider_api_key", | ||
| display_name="Provider API Key", | ||
| info="An alternative to the Astra Authentication that passes an API key for the provider with each request to Astra DB. This may be used when Vectorize is configured for the collection, but no corresponding provider secret is stored within Astra's key management system.", | ||
| ), |
There was a problem hiding this comment.
I made this appear by default. The reasoning is that
a) This is the quickest path to testing out vectorize on a new collection
b) If the collection is already configured for vectorize, but the key was not stored in astra, this is (by my guess) the most common pattern for authentication.
There was a problem hiding this comment.
I think the most common pattern is to store keys on astra, that's why I've set api_key_name in the front page.
If you go through the documentation the steps are to import the key in astra and refer to it when creating the collection
| info="The name of the embeddings provider API key stored on Astra. If set, it will override the 'ProviderKey' in the authentication parameters.", | ||
| advanced=True, |
There was a problem hiding this comment.
Is there a common case where the providerKey is set differently than the default?
There was a problem hiding this comment.
what do you mean? there's no default value. The value is defined by the user
There was a problem hiding this comment.
I believe it does default just to providerKey, like so:
CollectionVectorServiceOptions(
...
authentication={
"providerKey": os.environ["OPENAI_API_KEY"],
}
)
But going through the UI flow, I see that you are correct in your other comment that the user-defined key name is required during creation. So I will undo these changes, and have Provider API Key Name shown by default, and Provider API Key in Advanced.
| @@ -116,6 +116,7 @@ class AstraVectorStoreComponent(LCVectorStoreComponent): | |||
| name="embedding", | |||
| display_name="Embedding or Astra Vectorize", | |||
| input_types=["Embeddings", "dict"], | |||
| info="Allows either an embedding model or an Astra Vectorize configuration. If Astra Vectorize is already configured for the collection, this field is not required.", | |||
There was a problem hiding this comment.
did you try what happens if you don't set it ?
I believe the AstraDBVectorStore requires one of those.
There was a problem hiding this comment.
Ugh, yeah. I guess that's something to figure out. It's required by langchain-astradb to have one or the other at the moment.
There was a problem hiding this comment.
Perhaps we could loosen that constraint iff the setup mode is SetupMode.OFF?
github-actions
bot
added
documentation
small updates to vectorize docs Co-authored-by: Gabriel Luiz Freitas Almeida <[email protected]>
small updates to vectorize docs Co-authored-by: Gabriel Luiz Freitas Almeida <[email protected]> (cherry picked from commit a933139)
No description provided.