Source Gateway
Turn raw inputs into usable context
A vertical information waterfall. Slack messages, CRM fields, document snippets, database rows, and tool outputs fall as light bands. A horizontal filter plane turns raw above into structured Facts, Entities, Dates, Permissions, and Open Questions below.
The information waterfall
Raw above the plane. Structured below.
RAW INPUTSSTRUCTURED · Facts · Entities · Dates · Permissions · Open Qs
Slack · refund thread
CRM · plan field
Doc · policy.pdf
DB · invoices
Tool · billing.get
Unknown blob
Needs Mapping
missing time / source
missing time / source
Fact · double charge
Entity · Acme Corp
Date · 2026-07-12
Permission · support+
Open · refund amount?
Drag the gold filter plane. Bands missing time, source, or type drift into Needs Mapping. Click a band for the raw→field breakdown and what may proceed to the next agent.
Band detail
Raw
—
Structured
—
Forward to agents
—
Pass The Right Context To Every AI Step
Context that arrives intact — not more context, the necessary context.
RelayLum sits between agents, models, tools, sessions, and memory. It turns raw inputs into Context Packets optimized for the next hop — compact, reliable, and fully traceable.