Canon / VaERL
Canon is the structured story memory that Arc helps authors build and review. VaERL is the resolution layer that keeps this memory connected to entities, aliases, facts, relationships, and evidence.
VaERL stands for Vault-aware Entity Resolution Layer.
Canon
Canon is not just a list of facts. In Arc, canon should be:
- structured enough to query;
- connected to source evidence;
- editable by the author;
- aware of uncertainty;
- safe to use as context for AI-assisted revision.
A canon claim might be simple:
Mara is the captain of the river guard.
But for Arc, the useful version includes more structure:
- who Mara is;
- whether “captain” is title, role, or nickname;
- where the claim appears;
- whether the claim is current in the timeline;
- whether the author has accepted it.
VaERL
VaERL is the layer that helps stabilize narrative memory over time. It is responsible for questions such as:
- Are “Mara,” “Captain Mara,” and “the captain” the same entity?
- Is this “Ash Gate” a place, an event, or an organization?
- Does this relationship have evidence?
- Is this fact accepted canon or still under review?
- Has a later chapter changed the meaning of an earlier claim?
Why entity resolution matters
Long-form fiction often repeats, hides, renames, and recontextualizes information. A character can have several names. A city can become a faction. A prophecy can be both an object and an event in the story’s structure.
If those references are not resolved carefully, AI context becomes unstable. The model may merge different people, split the same person into duplicates, or invent continuity that the manuscript does not support.
Canon is reviewable, not automatic
Arc can propose structure. The author decides what becomes stable canon.
That distinction is important. The goal is not to let automation overwrite the story bible. The goal is to make the story bible, graph, evidence, and review decisions work together.