Support Console Resolver
Workflow
- Read the prompt, every payload file, and
answer_template.jsonbefore calling the API. - Use the task's base URL exactly as provided. Confirm
/healthand/api/catalog, then fetch authoritative support-console records; never decide from payload wording alone. - Preserve the output shape, key names, enums, item ordering, numeric precision, and "return only JSON" requirement from the template.
- For lookup and joins, use references/support_console_rules.md. For faster evidence gathering, run
scripts/fetch_support_console.pywith the task base URL and the IDs from the payload. - Build a small evidence table per item, apply the priority rules, compute summaries from the final decisions, and validate the JSON before answering.
Evidence Fetching
Use the helper when several related records are needed:
python3 scripts/fetch_support_console.py --base-url "$TASK_ENV_BASE_URL" --tickets TCK-1 TCK-2
python3 scripts/fetch_support_console.py --base-url "$TASK_ENV_BASE_URL" --cases CASE-1 CASE-2
python3 scripts/fetch_support_console.py --base-url "$TASK_ENV_BASE_URL" --enterprise-incidents INC-1
python3 scripts/fetch_support_console.py --base-url "$TASK_ENV_BASE_URL" --search "client or record id"
The helper is intentionally read-only. If an endpoint alias fails, inspect /api/catalog and retry with the catalog path.
Answer Discipline
- Keep ticket rows in payload order unless the template explicitly says otherwise.
- Sort case outputs by ascending
case_idwhen requested, even if the payload order differs. - Use empty strings,
false,0.0, orNO_ACTIONonly when the template says the field is not applicable. - Compute all summary counts from the completed decision array. Include zero counts for summary keys present in the template.
- Do not include explanation, markdown, evidence notes, or fields not present in the template.