Internal Server Error with jina-embeddings-v4 When Late Chunking is Enabled
I am encountering a persistent "Internal server error during encoding" when using the jina-embeddings-v4 model through the API and on the official playground.
This error appears to be directly related to the late chunking feature. The API call fails consistently when late chunking is enabled. However, the API returns a successful response when I switch to using the multi-vector feature instead. This suggests a potential server-side issue with the late chunking implementation.
For your reference, a recent Request ID for a failed request is [RID: 340a0431-5e48-4f11-b268-43f44ee997a2].
The error message is this:
{
"detail": "[RID: 340a0431-5e48-4f11-b268-43f44ee997a2] Internal server error during encoding"
}
Could you please investigate this issue with the late chunking functionality? Thank you!
Thank you for pointing this out.
We identified a bug in our API that occurred when handling late chunking, which was causing an unexpected error in some cases. This has now been fixed.
I have probably the same problem as a user in this conversation "[RID: dc2242e21d9b719a304288b6a48a56c2] Internal server error during encoding."
This error appears to be directly related to the late chunking feature when using the jina-embeddings-v4 model through the HTTP request.
Could you please investigate this issue with the late chunking functionality? Thank you!
I have probably the same problem as a user in this conversation "[RID: dc2242e21d9b719a304288b6a48a56c2] Internal server error during encoding."
This error appears to be directly related to the late chunking feature when using the jina-embeddings-v4 model through the HTTP request.Could you please investigate this issue with the late chunking functionality? Thank you!
I experience the same issue.