Knowledge Updates
Resolved threads, corrections, and approved drafts flowing back into the knowledge base — versioned, attributed, and reviewable.
A knowledge base that only grows through manual writing is a knowledge base that quietly ages. Updates from real conversations keep it current.
What it solves
Keeps knowledge sources current at the rate at which the underlying reality changes. Reduces the staleness that quietly poisons retrieval.
How we build it
Resolved thread or correction proposes a draft; channel or topic owner reviews; approved drafts land in the knowledge source with attribution and version. Retrieval picks up the change at the next refresh; the eval set re-runs against the updated corpus.
- Draft from real interaction
- Owner review and edit
- Versioned update with attribution
- Eval re-run on corpus change
What changes when it is in place
Knowledge freshness becomes a metric. Retrieval quality stays high as the underlying world changes.