Get Stats
Get aggregate statistics for a retriever's learned fusion system.
Path parameters
Response
Successful Response
Total number of unique users with learned weights.
Unique users the system is actively learning for (alias of total_learners — backs the Studio Overview 'Active Learners' card).
Average number of interactions per user.
Maximum divergence of any feature weight from uniform.
Count of users at each context level (personal/demographic/global).
Fraction of learners that have reached a personalized (non-global) context — enough interactions for personal weights (0-1).
Whether learned fusion is currently enabled for this retriever.
Effective percentage of traffic receiving learned fusion weights (a live override takes precedence over the configured value).
Whether a live rollout override is active (vs the config value).
The retriever's configured rollout_pct (ignoring any live override).
Whether learned fusion is running in shadow mode (logging only).
Which store served the weights: 'signals' (the namespace's _signals collection) or 'clickhouse' (analytics fallback).
Set when the response contains fallback defaults due to an internal error.
Set when learned fusion is NOT configured for this retriever — the payload is a well-formed empty state and this explains how to enable learning (BACKE-2525).