A1 Interface

A1 is the diplomatic channel between the two RICs — it runs from the non-RT RIC down to the near-RT RIC and carries guidance rather than direct control. Three kinds of thing flow over it: policies (declarative goals like "keep this slice's latency under X for these users"), enrichment information (outside context the RAN can't see on its own, such as a traffic or mobility prediction), and ML model management. The near-RT RIC takes that guidance and turns it into concrete fast-loop actions through its xApps.

The mental model that helps: A1 is about intent and steering, not commands. The non-RT RIC doesn't reach into the RAN and flip a parameter — it hands the near-RT RIC a policy and trusts the local xApps to honour it within the near-real-time loop. Contrast that with E2, which does carry direct control down to the RAN nodes. Keeping A1 (policy, slow, RIC-to-RIC) and E2 (control, fast, RIC-to-RAN) straight is one of the more common O-RAN interview stumbles.