Release Material: Product Version (Application Teams)

Problem

Application teams need memory that survives real traffic and supports policy-driven behavior, not a demo-only vector lookup that is hard to operate.

Architecture Principles

1. Audit-first: every memory write is trackable and recoverable.
2. Derived async: background derivations do not block user-facing write flows.
3. Memory -> Policy: memory signals directly influence planner/tool decisions and can learn from feedback.

Evidence

1. Stable API contract for write and recall_text.
2. SDK availability in TypeScript and Python for fast integration.
3. Operational runbooks, health gates, and consistency checks for release confidence.
4. Published container artifact for standard deployment path.

Boundaries

1. Aionis is a memory kernel, not a full workflow orchestrator.
2. Retrieval quality still depends on prompt strategy and application-side eval loops.
3. Some enterprise governance workflows require staged rollout with platform teams.

Next Step

1. Start with one production flow and lock KPI targets.
2. Enable policy/rule feedback loop for one high-value decision point.
3. Expand scope only after latency, quality, and ops metrics are stable for two release cycles.