Description
Hello everyone,
I’m encountering a persistent error when trying to create objects in my Weaviate instance, and I’m hoping someone in the community might have some insights.
The Error:
Every time I attempt a POST request to /v1/objects for my MunicipioPdf class, the Weaviate logs show the following error:
{"action":"requests_total","api":"rest","class_name":"MunicipioPDF","error":"update vector: unmarshal error response body: Not Found","level":"error","msg":"unexpected error","query_type":"objects","time":"..."}
This error occurs immediately after the POST request is received by Weaviate.
[
Suspicious Content] Our Architecture and Language:
Backend Language: Java (using Spring Boot and the spring-webflux WebClient for interacting with Weaviate).
Weaviate Setup: Running Weaviate in a Docker container (version 1.26.0, as per the logs). The text2vec-huggingface module also appears to be running within the same Weaviate container (based on the logs).
Vectorizer Model: We are using the sentence-transformers/all-mpnet-base-v2 model specified in our Weaviate schema for the MunicipioPdf class.
Our Weaviate Schema for the MunicipioPdf Class:
JSON
{
“classes”: [
{
“class”: “MunicipioPdf”,
“description”: “PDF chunks from Municipio documents”,
“vectorizer”: “text2vec-huggingface”,
“vectorizeClassName”: false,
“moduleConfig”: {
“text2vec-huggingface”: {
“model”: “sentence-transformers/all-mpnet-base-v2”
}
},
“properties”: [
{
“name”: “filename”,
“dataType”: [“text”],
“description”: “Name of the PDF file”,
“tokenization”: “word”,
“indexInverted”: true,
“indexFilterable”: true,
“indexSearchable”: true,
“vectorizePropertyName”: false,
“skipVectorization”: true
},
{
“name”: “content”,
“dataType”: [“text”],
“description”: “Text content chunk extracted from the PDF”,
“tokenization”: “word”,
“indexInverted”: true,
“indexFilterable”: false,
“indexSearchable”: true,
“vectorizePropertyName”: false,
“skipVectorization”: false
},
{
“name”: “chunkIndex”,
“dataType”: [“int”],
“description”: “Index of the chunk within the PDF”,
“indexInverted”: true,
“indexFilterable”: true,
“indexSearchable”: false,
“vectorizePropertyName”: false,
“skipVectorization”: true
}
]
}
]
}
DOCKER
sudo docker run -d
-p 8181:8080
-e AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED=true
-e PERSISTENCE_DATA_PATH=“/var/lib/weaviate”
-e ENABLE_MODULES=text2vec-huggingface
-e DEFAULT_VECTORIZER_MODULE=text2vec-huggingface
-e HUGGINGFACE_APIKEY=XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
-e HF_HOME=“/var/lib/weaviate/hf_cache”
-e TEXT2VEC_HUGGINGFACE_MODEL=“sentence-transformers/all-mpnet-base-v2”
-e TEXT2VEC_HUGGINGFACE_ENDPOINT=“https://api-inference.huggingface.co/pipeline/feature-extraction/sentence-transformers/all-mpnet-base-v2”
-e QUERY_DEFAULTS_LIMIT=25
-e LOG_LEVEL=debug
–name weaviate-huggingface-4
semitechnologies/weaviate:1.26.0
Things We’ve Tried:
We’ve confirmed that our Java application can successfully connect to Weaviate for reading objects.
Even a simple curl POST request to create a basic MunicipioPdf object with just filename and content fails with the same error in the Weaviate logs.
We’ve reviewed the Weaviate container logs (weaviate-huggingface-4), and during startup, the text2vec-huggingface module appears to initialize correctly without any immediate errors related to permissions, API keys, or network issues.
Our Question:
Has anyone encountered this “update vector: unmarshal error response body: Not Found” error specifically during object creation with the text2vec-huggingface module? What could be causing Weaviate to fail during the vector update with a “Not Found” error, even for basic object creation? Any suggestions on where to look next for debugging this issue would be greatly appreciated!
Thank you in advance for your help!