Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[BUG]: Weaviate: Indexes seem to be queries with underscores removed? #145

Open
bbrala opened this issue Jul 6, 2024 · 10 comments
Open
Labels
bug Something isn't working core-team

Comments

@bbrala
Copy link

bbrala commented Jul 6, 2024

How are you running VectorAdmin?

Docker (local)

What happened?

Error when trying to sync.

Are there known steps to reproduce?

Connected to weaviate with existing named classes that do not use camelcase, then you get following messages consistently when trying to sync.

vector-admin | ClientError: Cannot query field "Bp10153C51068DocumentChunksV1" on type "GetObjectsObj". Did you mean "Bp101_53c51068_documentChunks__v1", "Bp131_13b35c33_documentChunks", "Bp160_15dede47_documentChunks", "Bp100_e19743d9_documentChunks__v1", or "Bp48_8ae7d6f8_documentChunks__v1"?: {"response":{"errors":[{"locations":[{"column":6,"line":1}],"message":"Cannot query field \"Bp10153C51068DocumentChunksV1\" on type \"GetObjectsObj\". Did you mean \"Bp101_53c51068_documentChunks__v1\", \"Bp131_13b35c33_documentChunks\", \"Bp160_15dede47_documentChunks\", \"Bp100_e19743d9_documentChunks__v1\", or \"Bp48_8ae7d6f8_documentChunks__v1\"?","path":null}],"status":200,"headers":{}},"request":{"query":"{Get{Bp10153C51068DocumentChunksV1(limit:1){_additional { vector }}}}"}}

And

vector-admin | Weaviate:namespaceCountWithClient Cannot query field "Bp111F0880665DocumentChunksV1" on type "AggregateObjectsObj". Did you mean "Bp126_0dcea379_documentChunks", "Bp138_15a0c090_documentChunks__v1", "Bp58_ee2680db_documentChunks__v1", "Bp111_f0880665_documentChunks__v1", or "Bp125_414b683f_documentChunks"?: {"response":{"errors":[{"locations":[{"column":12,"line":1}],"message":"Cannot query field \"Bp111F0880665DocumentChunksV1\" on type \"AggregateObjectsObj\". Did you mean \"Bp126_0dcea379_documentChunks\", \"Bp138_15a0c090_documentChunks__v1\", \"Bp58_ee2680db_documentChunks__v1\", \"Bp111_f0880665_documentChunks__v1\", or \"Bp125_414b683f_documentChunks\"?","path":null}],"status":200,"headers":{}},"request":{"query":"{Aggregate{Bp111F0880665DocumentChunksV1{meta { count }}}}"}}

And more of those.

@bbrala bbrala added the possible bug unconfirmed bug report label Jul 6, 2024
@bbrala
Copy link
Author

bbrala commented Jul 6, 2024

After changing all this.camelCase things to seem to sync.

Very ugly hack:
embeddings-fix.patch

@timothycarambat
Copy link
Member

Was this namespace created by AnythingLLM or was it already existing? We shouldn't have issues with referring to spaces created via AnythingLLM since in theory, the rules of namespace should be consistent from creation/update/read/delete

@bbrala
Copy link
Author

bbrala commented Jul 8, 2024

No it was an existing one. Was under the inpression it would allow some interfacing to our existing weaviate database.

@timothycarambat
Copy link
Member

Ah, so technically speaking, if the workspace and the namespace in Weaviate have the same slug value - it would work when it returns vector search results, but they would never show in citations and the embeddings created out of AnythingLLM need to have a text metadata property to even be injected. This kind of use-case is not explicitly supported because of the litany of possible issues from someone embedding their data in any way they deem fit and it not being compatible with the way AnythingLLM expects and manages vectors and their associated metadata.

@timothycarambat timothycarambat closed this as not planned Won't fix, can't repro, duplicate, stale Jul 8, 2024
@bbrala
Copy link
Author

bbrala commented Jul 9, 2024

Ok so vector admin is technically AnythingLVM-admin?

Unfortunate. Thanks.

@timothycarambat
Copy link
Member

No, they are two totally separate repos. If it was just for anythingLLM it would be in this repo

@bbrala
Copy link
Author

bbrala commented Jul 10, 2024

Ok, but you do not aim to support existing databases then. Having fixed rules on how things are named seems kinda counter intuitive.

I could contribute a setting of some sort to enable/disable the forces camelcase if you like.

@timothycarambat
Copy link
Member

I just realized we have been talking on the vector-admin repo which explains all the confusion. This whole time I thought this was AnythingLLM's repo where I get 99% of issues/notifications.

This is a bug for VA, lol. Yes underscores should not matter in this context

@timothycarambat timothycarambat added bug Something isn't working core-team and removed possible bug unconfirmed bug report labels Jul 10, 2024
@bbrala
Copy link
Author

bbrala commented Jul 10, 2024

Hehe, thanks :)

@HassanAITBRIK
Copy link

Any update about this issue ?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working core-team
Projects
None yet
Development

No branches or pull requests

3 participants