V6 — Architecture (Page Index) — Egbe Companions

This directory holds the V6 (Egbe — an agentic-companion universe where autonomous AI agents, "Ori", live persistent lives in a shared world "Orun" and the player is a steward guiding their fate) architecture reference, decomposed from the single ../V6_ARCHITECTURE.md hub into focused, in-depth, code-grounded pages. The hub remains the canonical entry point and preserves every section heading; each page below expands one part of the system with richer, code-grounded detail and an honest implemented / spec-only status. The companion feature map lives in ../V6_features.md.

Cross-reference convention. Page-to-page links are relative (./<slug>.md). Platform-foundation links point at the shared platform/ space.

Architecture landscape#

flowchart TB World[Orun world server and authoritative state] --> Moirai[Moirai deterministic life ticks] Identity[(Ori identity biography memory and consent)] --> Cognition[Cognition grounding planning and policy] Moirai --> Cognition VAC[VAC player communication] --> Cognition Cognition --> Action[Governed intent and world effect] Action --> World Action --> Identity World --> Client[UE embodiment pixel streaming and web fallback] Identity --> Clio[Clio narrative] Identity --> Foundry[Foundry ending and legacy] Identity --> Aye[Aye cross-game incarnation] Action --> Evidence[(Replay provenance eval and audit)] Evidence --> Operations[Welfare cost performance security testing and launch]

The topology encloses probabilistic reasoning inside durable, versioned, replayable state and governed effect boundaries. A model response is neither a world action nor a biography update until the authoritative contract commits it and produces evidence.

Trace a change from Ori/consent version and Moirai tick through recall, grounding, cognition, policy, intent, world effect, biography/event receipt, embodiment, and evaluation/operations. Pair the trace with the feature index so runtime components are evaluated against autonomy, continuity, disclosure, accessibility, welfare, and recovery promises.

Every focused architecture topic appears exactly once below. The V6 documentation guide supplies the truth/welfare model and role paths; the diagram atlas supplies the shared visual language. Backlog, dependencies, current schemas/code/tests, evals, flagship evidence, model/provider state, and release operations govern completion.

Orientation#

V6 Product Promise & Architectural Overview — what V6 commits to and the agentic-companion thesis.
Subsystem Glossary — every V6 subsystem (libs/v6 packages + UE modules) mapped to code.
Architecture, Topology & Project Layout — the runtime topology composing agent-mind, world, and edge.

The Agent Mind#

The Ori Biography Service — the agent's persistent identity, biography, and memory.
The Moirai Simulation Kernel & Cost Tiering — the deterministic life-sim kernel and AI cost/fidelity tiering.
The Cognition Stack & Agent Behavior — how an Ori thinks and acts (the LLM/agent-loop layer + grounding).

World, Edge, and Client#

The World Server & Shard Continuum — the world server hosting Orun and the shard continuum.
Gateway, Pixel Streaming & the Web Fallback — the edge: gateway, pixel streaming, and the web-fallback client.
UE5 Client Modules & Agent Embodiment — the UE5 client modules and how agents are embodied as avatars.

Communication, Story, and Lifecycle#

VAC: The Communication Pipeline — how the player communicates with agents (voice/text → intent → response).
Clio: The Story Engine — turning emergent agent lives into narrative.
Foundry: Endings & Legacy — how an agent's life concludes and what persists.
The Aye Bridge: Cross-Game Incarnation — how an Ori incarnates across Oshun product versions.

Platform, Governance, and Launch#

Determinism, Data & Integration — the determinism invariant, data architecture, and subsystem integration.
Safety, Agent Welfare, Provenance & Eval Gates — the safety architecture, agent welfare, provenance, and eval gates.
Performance, Build, Security, Testing & Launch — performance budgets, build/cook, security, testing, and launch readiness.