Run Rule Test

Request body for POST /custom-rules/test.

context selects the prompt mode for LLM rules:

• isolated (default) — the prompt carries only the draft rule; good for iterating on criteria wording.
• production — the prompt carries the org's full active rule set with the draft substituted for its saved version (rule_id), exactly like a real scan; the reply is filtered to the draft's tags. Only valid for rule_type="llm".

Response from POST /custom-rules/test.

llm_failed is True when the LLM provider could not complete the evaluation (API error, invalid response). The UI surfaces this as a warning instead of showing a misleading "No match".

llm_call_id / llm_prompt_url / llm_response_url are populated only when an LLM-based rule actually ran and the worker persisted the call's telemetry. The UI renders collapsible prompt and response sections from the two same-origin URLs (served by /api/internal/llm/calls/{id}/...); non-LLM rule types and runs where telemetry didn't land leave them empty/None so the sections stay hidden.