ValueError
pinecone.exceptions.PineconeException: ValueError: vector dimension mismatch
Stack trace
ValueError: Vector dimension mismatch: expected 1536, got 512
File "site-packages/pinecone/client.py", line 234, in upsert
raise ValueError(f"Vector dimension mismatch: expected {self.dimension}, got {len(vector)}") Why it happens
Pinecone indexes are created with a fixed vector dimension. When inserting vectors, the length of the vector must exactly match this dimension. If the vector length differs, Pinecone raises a dimension mismatch error to prevent corrupting the index.
Detection
Check vector length before insertion by asserting len(vector) == index.dimension or catch ValueError from the upsert call and log the vector size and index dimension.
Causes & fixes
The vector generated by your embedding model has a different dimension than the Pinecone index was created with.
Verify the embedding model output dimension matches the Pinecone index dimension; recreate the index with the correct dimension if needed.
You are mixing vectors from different embedding models with different output sizes in the same Pinecone index.
Ensure all vectors inserted into a Pinecone index come from the same embedding model with consistent output dimension.
The Pinecone index was created with a default or incorrect dimension setting not matching your vectors.
Explicitly specify the correct dimension when creating the Pinecone index to match your embedding vectors.
Code: broken vs fixed
import os
from pinecone import Pinecone
pinecone = Pinecone(api_key=os.environ['PINECONE_API_KEY'])
index = pinecone.Index('my-index')
vector = [0.1] * 512 # Vector dimension 512
# This line raises ValueError due to dimension mismatch
index.upsert(vectors=[('vec1', vector)])import os
from pinecone import Pinecone
pinecone = Pinecone(api_key=os.environ['PINECONE_API_KEY'])
index = pinecone.Index('my-index')
vector = [0.1] * 1536 # Fixed vector dimension to match index
# Now upsert works without dimension mismatch error
index.upsert(vectors=[('vec1', vector)])
print('Vector inserted successfully')Workaround
Catch the ValueError on upsert, log the vector length and index dimension, then skip or fix the vector before retrying insertion.
Prevention
Always create Pinecone indexes with the exact dimension of your embedding model output and validate vector lengths before insertion to avoid mismatches.