You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, index settings & mappings are only initialized if selecting from a preset use-case, and quick-configure fields are defined. But, if starting from scratch (selecting "custom" template), if ML ingest processors are configured, or document fields with special embeddings are defined, nothing is automatically updated, and it is in the default/empty state, with no references or helper UI for the users.
For example, if configuring an ML processor to generate embeddings, it will always require index.knn: true setting to be set, and a mapping of type knn_vector with the correct embedding length is required, for proper knn search to work later on.
A few ideas to improve the overall experience, of which are not mutually exclusive:
offer presets, similar to the query building modal, of index settings / mappings (e.g., a knn index should have index.knn: true setting in some preset)
add logic to try to generate valid index mappings & settings as users fill out or update the form (selecting an ML processor, updating the sample source data documents, etc.)
The text was updated successfully, but these errors were encountered:
ohltyler
changed the title
[ENHANCEMENT] Auto-populate index settings & mappings based on configurations
[ENHANCEMENT] Improve index settings & mappings configurations
Dec 26, 2024
Currently, index settings & mappings are only initialized if selecting from a preset use-case, and quick-configure fields are defined. But, if starting from scratch (selecting "custom" template), if ML ingest processors are configured, or document fields with special embeddings are defined, nothing is automatically updated, and it is in the default/empty state, with no references or helper UI for the users.
For example, if configuring an ML processor to generate embeddings, it will always require
index.knn: true
setting to be set, and a mapping of typeknn_vector
with the correct embedding length is required, for proper knn search to work later on.A few ideas to improve the overall experience, of which are not mutually exclusive:
knn
index should haveindex.knn: true
setting in some preset)The text was updated successfully, but these errors were encountered: