Human-readable resolver walkthrough
Try one lookup. Understand every outcome.
This page sends one read-only request to the same public API used by desktop applications and local models. It explains what the response means, what you can safely do next, and what the response never authorizes.
Before you call
Know what the deployed service can do.
reviewed_exact_registry_first
The current deployment cannot return a positive exact resolution.
You may still submit a request and observe explicit abstention. An empty or incomplete public corpus is never permission to guess a ConceptCode.
1 · Send the minimum request
What expression needs a stable identity?
Send the expression itself—not the whole conversation, hidden prompt, credentials, or unrelated private context. Add a language tag only when you know it.
Outcome guide
The next action follows the response—not intuition.
resolved
Reviewed registry evidence identified one published Concept.
Tell the human: A governed identity was found. The ConceptCode below is the registry-backed result, not a guess.
Next: Review the Concept name and definition, then use the code only for the intended semantic reference.
Never: Do not rewrite, shorten, translate, alias, or reconstruct the returned ConceptCode.
unknown_expression
No governed reviewed exact identity is available for this expression.
Tell the human: No reviewed registry identity was found. This is a deliberate unknown result, not a system guess.
Next: Recheck the wording or preserve the expression as unresolved; do not substitute a likely code.
Never: Do not infer from labels, candidates, similarity, embeddings, or local-model intuition.
ambiguous_expression
More than one governed exact identity remains possible.
Tell the human: The expression needs more context before one ConceptCode can be selected safely.
Next: Ask a clarifying question that distinguishes the possible meanings, then submit the clarified expression.
Never: Do not choose a candidate silently or use model confidence as authority.
other_abstention
The resolver intentionally declined to assign semantic identity.
Tell the human: The service did not assign a ConceptCode and reported the reason in data.reason.
Next: Preserve the reason and follow any deployment or evidence guidance without inventing identity.
Never: Do not convert an abstention into a best-effort resolution.
input_or_service_error
No semantic result was produced.
Tell the human: The request needs correction or the service response could not be trusted.
Next: Correct the stated input problem or report the service failure, then retry only when appropriate.
Never: Do not fill the gap with an inferred semantic result.
For machine agents
Use one ordered decision procedure.
A client should evaluate the envelope in this order and fail closed when the response does not match the published contract.
-
Step 1
error is not nullThe request was invalid or the service could not provide a semantic result. Repair or report only the stated error. Assign no ConceptCode. -
Step 2
data.status == "resolved" and data.concept.code is a non-empty stringThe governed registry returned an authoritative published identity. Use the exact returned code and preserve registryVersion. -
Step 3
data.status == "abstained"The resolver intentionally assigned no ConceptCode. Preserve data.reason and follow the matching abstention guidance. Never guess. -
Step 4
anything elseThe response does not satisfy the published contract. Fail closed, assign no ConceptCode, and report a contract error.