hylouis233/epic-site-pages/epic-disease-query

EPIC Disease Query

Use EPIC for exploratory public-health monitoring and event-intelligence research. Treat every result as a source-linked monitoring signal, not a confirmed diagnosis, complete global inventory or official risk alert.

Query Workflow

1. Read data/v1/manifest.json before querying records. Surface source_status, data_as_of, last_successful_ingest_at, staleness_hours and warnings.
2. If status is degraded, stale, or failed, place that warning before the results. Do not describe the build timestamp as the data update time.
3. Query data/v1/records.json and filter only on documented Schema v1 fields.
4. Preserve event_id, the EPIC detail URL, original source, source organization, dates and quality score unless the user explicitly asks for fewer fields.
5. State that EPIC is incomplete public-source monitoring and not clinical advice.

Helper

Run scripts/query.py from this skill directory or pass an explicit path.

python scripts/query.py \
  --disease dengue \
  --continent Asia \
  --from-date 2026-06-01 \
  --sort cases:desc \
  --format markdown \
  --limit 10

Useful options:

• --keyword: search normalized/raw disease, location, source organization and summary;
• --disease: match disease ID, Chinese/English normalized name or raw alias;
• --location, --continent, --source-type: narrow by geography or provenance;
• --from-date, --to-date: filter event_start_date, falling back to published_at;
• --sort FIELD:asc|desc: sort by cases, deaths, dates, quality score or another public scalar field;
• --fields: comma-separated output field selection;
• --format markdown|json|ndjson: output mode;
• --deduplicate: deduplicate by event_id, signal_id or cluster_id;
• --records-file, --manifest-file: deterministic local/offline query inputs;
• --timeout, --retries: network failure handling.

Result Rules

• Keep nulls distinct from zero. “No parsed case count” does not mean zero cases.
• Do not sum values across different period_type values without explaining the incompatibility.
• Prefer disease_id for filtering; display disease_name_zh and retain disease_raw when alias context matters.
• Link to https://hylouis233.github.io/epic-site-pages/events/{event_id}/ for EPIC context and to source for source verification.
• If the schema version is not compatible with major version 1, stop with a clear schema error instead of guessing.
• On timeout, invalid JSON or unavailable files, return a nonzero exit code and an actionable error on stderr.

Default Output

For natural-language answers, lead with the data-status warning, then a concise table with:

event date | disease | location | cases | source organization | quality | EPIC event | original source

End with the active filters, total matches before truncation, data_as_of, and the monitoring/medical limitation.