Developer guides / Interfaces
Questions in. Evidence alongside the answer.
An interface sketch for applications and agents. Field names below illustrate the intended shape and are not a released API contract.
Coming soon
Sample content for this design. This resource is coming soon.
Request & response
Request
Select the world and name the question.
Illustrative preview
{
"dataset": "movies",
"question": "Who directed Lantern Road?"
}Response
Keep answer content, records and provenance together.
Illustrative preview
{
"answer": "In this fictional sample, Mei directs Lantern Road.",
"records": [
{ "id": "F-021", "type": "Film", "title": "Lantern Road" },
{ "id": "P-014", "type": "Person", "name": "Mei" }
],
"sources": [
{ "record": "CR-008", "relationship": "directed_by" }
]
}SDK
- Configure
- Deployment address and authentication guide pending
- Ask
- Dataset selection and question text
- Inspect
- Answer, matching records and source references
MCP
- Tool purpose
- Ask a question against an ontology
- Inputs
- Dataset selection and question text
- Connection
- Server and authorization details pending
Design the unsuccessful response, too.
Illustrative error shapes for an application. Names and fields are design examples, not a published API contract.
Source unavailable
Illustrative preview
{
"status": "source_unavailable",
"answer": null,
"message": "The selected source could not be reached.",
"retryable": true
}No matching records
Illustrative preview
{
"status": "no_matches",
"answer": "No records match the selected scope.",
"records": [],
"sources": []
}Clarification needed
Illustrative preview
{
"status": "needs_clarification",
"question": "Which release of this film do you mean?",
"choices": ["F-021", "F-034"]
}Coming soon
Film titles, people and identifiers in this workbench are invented. No request is sent.