Search
Index Lane search is live and provider-backed.
That means the backend does four things:
- fan out the query to connected providers
- use provider-native search behavior where available
- normalize the results into one shared shape
- apply lightweight ranking on top
Result shape
Every result preserves provenance so an agent can reason about where it came from.
Each hit includes:
- provider
- item id
- item type
- title
- snippet
- timestamp when available
- account label when available
- metadata
Ranking
The current ranking is intentionally lightweight.
It favors:
- title match
- body or snippet match
- recency
- native provider rank when the provider already returned ordered results
Compact mode
Compact mode is the default because it is the correct behavior for agent workflows.
You should not fetch the full body for every candidate up front.
Full mode
Use --full when you explicitly want larger excerpts inline.
Even then, prefer get for the single result you actually want to expand.