V6 Documentation Guide — Egbe Companions

This guide is the curated front door to V6, Egbe: The Agentic Companion Universe. V6 centers autonomous agents called Ori who live persistent lives in Orun while the player acts as steward rather than puppeteer. The docs join that emotional product promise to biography and memory, deterministic life simulation, cognition and grounding, embodiment, communication, story, cross-game incarnation, welfare, cost, safety, and operations.

V6 differs from V2–V5: its feature document is an originating design source, and the backlog is derived from it. After derivation, the backlog and current implementation/evidence—not the ambition of the originating document—govern completion status.

Documentation map#

mindmap root((V6 documentation)) Agent experience Orun persistent world Ori identity biography and memory Stewardship communication and fate Commons creation and live service Agent runtime Moirai deterministic life simulation Cognition grounding and action VAC voice and text Clio story and Foundry legacy Embodiment and continuity UE client and world server Pixel streaming and web fallback Aye cross-game incarnation Ori portability and schema evolution Governance and delivery Welfare safety and provenance Evals cost and budgets Performance security and testing Backlog release and evidence

Start with the question#

Question	Canonical starting point	What it covers
What does V6 promise the player and the Ori?	Feature hub, feature index, and overview/design posture	Stewardship, autonomy, persistent lives, world, communication, legacy, creation, and launch posture.
How does an Ori remain the same individual?	Ori feature, biography service, and schema/DR docs	Identity, biography, memory, consent, versioning, compaction, recovery, and portability.
How does an agent think and act?	Agent minds, Moirai kernel, and cognition stack	Deterministic simulation, model tiering, perception, planning, grounding, commitment, and replay.
How does the player communicate without taking control?	Stewardship, VAC feature, and VAC architecture	Intent, confirmation, refusal, voice/text, autonomy, disclosure, and recovery.
How are endings, stories, and cross-game continuity governed?	Fate/legacy feature, Clio, Foundry, and Aye bridge	Narrative derivation, terminal transitions, legacy, consented incarnation, and provenance.
What is actually complete?	V6 backlog, product review, current code/tests, flagship evidence, and release records	Evidence-backed status by agent/runtime/world/client/governance surface.
Which systems are composed?	Dependency registry and determinism/data/integration	V1/V3/shared substrate reuse, ownership, data flow, and fallback boundaries.
Where are the visuals?	V6 diagram atlas and the global diagram gallery	Curated cognition, biography, world, lifecycle, ownership, and release models.

Truth and welfare model#

stateDiagram-v2 [*] --> Designed Designed --> Simulated: deterministic state transition exists Simulated --> CognitivelyIntegrated: perception memory planning and action compose CognitivelyIntegrated --> Embodied: supported client or fallback presents the Ori Embodied --> WelfareVerified: autonomy safety continuity and quality evidence passes WelfareVerified --> ReleaseReady: cost scale operations recovery and rollout pass Simulated --> FixtureOnly: scripted or seeded behavior stands in for autonomy CognitivelyIntegrated --> TierLimited: behavior depends on named fidelity or model tier WelfareVerified --> Regressed: drift harm continuity or quality evidence fails Regressed --> Simulated: repair returns below release readiness

Anthropomorphic presentation raises the documentation bar. Do not infer a persistent life from a chat transcript, autonomy from generated prose, memory from a context window, consent from a default, welfare from a tone rubric, or cross-game identity from a shared avatar identifier. Claims require durable state, governed transitions, disclosure, user controls, replayable evidence, and explicit model/client tiers.

Recommended reading journey#

journey title From V6 companion promise to supported life cycle section Orient Choose world identity cognition or stewardship path: 5: Product, Design Pair experience with architecture and welfare boundary: 5: Architecture, Governance section Trace Follow perception memory simulation decision and action: 5: Engineering Inspect embodiment continuity consent and provenance: 5: Engineering, Governance section Verify Reproduce autonomy continuity refusal and recovery: 5: QA Validate welfare drift cost scale and accessibility: 5: QA, SRE section Release Prove recovery rollout and rollback evidence: 5: Release Update docs diagrams evals and runbooks: 5: Owners

Product, narrative, and welfare design#

Start with the feature index. Read the experience page alongside governance/safety, cost/evals, and the owning architecture. Specify what the player can influence, what the Ori can refuse, how uncertainty and generation are disclosed, and how endings or migration preserve dignity, history, consent, and reversibility where promised.

Engineering and agent systems#

Start with the architecture index. Trace durable Ori identity through Moirai ticks, cognition, grounding, tools/actions, world authority, VAC, Clio/Foundry, embodiment, Aye incarnation, stores/events, and tests. Preserve deterministic state transitions around probabilistic model calls and record the model/prompt/tool/provenance envelope needed for replay.

QA, accessibility, safety, and operations#

Exercise offline progression, clock drift, duplicate ticks, memory edit/delete/export, refused guidance, model/provider loss, tier fallback, voice/text alternatives, abusive stewardship, ending/legacy, migration, compaction/restore, cost exhaustion, and incident rollback. Evaluation sets are release gates only when current, representative, independently interpreted, and tied to supported runtime paths.

Page-set organization#

Feature index — 16 focused topics covering world/Ori, cognition, stewardship/communication/fate, creation/live service, trust, accessibility, cost, and rollout.
Architecture index — 16 focused topics covering biography, Moirai/cognition, world/edge/client, VAC/story/lifecycle/Aye, integration, welfare, and launch.
V6 diagram atlas — cognition, identity, lifecycle, composition, ownership, stewardship, and release models.
docs/, evals/, flagship-evidence/, testing/, performance/, security/, and release/ — focused evidence and operational details.

Documentation quality contract#

Every canonical feature and architecture topic stays substantial, indexed exactly once, visually explicit, and honest about deterministic state, probabilistic generation, fixture behavior, tier limits, consent, welfare, and release evidence. Behavior changes update contracts, schemas/migrations, evals/tests, backlog/evidence, runbooks, and diagrams together. Generated Docs Center integrity, freshness, search, gallery, and desktop/mobile checks pass.