medium10 pts · first answer countsragvector-databases

Insert fails with a dimension error

After switching to a new embedding API, inserts fail with "expected 768 dimensions, got 1536". What is the right fix?

A.Build a new 1,536-dimension index and re-embed everythingB.Cut each new vector to its first 768 numbers and insert itC.Pad the old vectors with 768 zeros so they matchD.Average pairs of numbers in each new vector to get 768

Discussion 0

Join free to post in the discussion.

Insert fails with a dimension error

Discussion 0

Related questions