Wiki Arcana is the ecosystem's knowledge orchestration layer. It is designed to store knowledge as a graph plus vectors, classify it into a hierarchy of knowledge spaces, and enforce a Jedi-Archive-style access model (public, archivist, council, hidden holocron) through Auth Arcana OIDC — never local logins.
It does not replace Scrutator or Long-Term Memory: embeddings are to be minted by Scrutator's BGE-M3 engine, search is to be delegated, and memory stays with LTM; Wiki Arcana is designed to bind them together behind a single API and an MCP server for agents. Phase 1 ships the research corpus, the storage abstraction and access registry, and the public skeleton — knowledge content and links come next. Open source, MIT license.
- Status
- In development (Phase 1)
- Autonomy (AAL)
- L1 → L4
- Site
- arcanada.wiki
- Repositories
- Two public repos, self-hosted CI
01
Graph plus vectors
The design stores knowledge as a graph plus vectors, classified into a hierarchy of knowledge spaces; activating the graph and vector engine is a roadmap step.
02
Jedi-Archive access model
The design enforces public, archivist, council and hidden holocron levels through Auth Arcana OIDC — no local logins; this access model is planned.
03
Orchestrates, does not duplicate
By design, Scrutator mints embeddings and serves search while LTM keeps episodic memory, and Wiki Arcana binds them behind one API and an MCP server for agents; the delegation is planned.
Capabilities
Working now
- Storage abstraction with engine-pure ports — the graph+vector engine is chosen by consilium, not locked in Phase 1
Planned
- Graph + vector storage layer with a knowledge-space classifier (a hierarchy of spaces)
- Jedi-Archive-style RBAC — public / archivist / council / hidden holocron — via Auth Arcana OIDC (deny-wins, no local auth)
- Delegates embeddings and search to Scrutator (BGE-M3); episodic memory stays in LTM — it orchestrates, it does not duplicate
Status not stated
- Single API + MCP server (Streamable HTTP) as the agent entry point
- Ecosystem-native — identity via Auth Arcana, retrieval via Scrutator, memory via LTM
Current autonomy level
Weakest link
Phase 1 — the research corpus, two public repos with self-hosted CI, the live arcanada.wiki landing, the relational access/space registry, and the API/MCP skeleton are in place. Knowledge content, graph/vector activation, and cross-links are deliberately deferred to later tasks, and the API is not yet publicly exposed.
Roadmap to L4
- Phase 1 → Phase 2 — activate the graph + vector engine (per consilium) behind the storage ports and benchmark it before any content build.
- Phase 2 → Phase 3 — ingest knowledge, wire Scrutator-delegated search + LTM memory, and populate knowledge-space cross-links.
- Phase 3 → Phase 4 — full RBAC enforcement across the API and MCP, per-user sharing and delegation, agent-facing retrieval at scale.
Next step
Review the current build and its development plan
BUILDING means active development, not a promise of public availability. Use the current capabilities and roadmap above as the reference.
Review the path to the target level