Model Version Consistency
A constraint principle in vector databases or embedding retrieval systems requiring that all embedding vectors in an index be generated by the same model version, including tokenizer, weights, and normalization steps. Embeddings from different model versions have the same dimensions but occupy different geometric spaces, so mixing them renders cosine similarity meaningless and causes silent degradation of retrieval quality.
Timeline (last 90 days)
The hybrid inference model has three key design points: first, the routing strategy (deciding whether a request goes on-device or to the cloud based on input length, task type, and network status); second, the fallback scheme (when the network is unavailable or the cloud times out, the on-device model gives an acceptable suboptimal result rather than directly reporting an error); third, consistency (on-device and cloud model versions need to be updated in coordination).