Workflow

Asynchronous agent research with public SQL

A Scry workflow for agents that inspect schema, run bounded SQL, preserve records, and return evidence with sources attached.

Answer shape

Scry lets an agent turn a broad research question into an inspectable sequence: schema, bounded source-native query, result record, and follow-up query.

Steps

  • 1Call /v1/scry/context to learn the current public endpoints.
  • 2Call /v1/scry/schema before writing SQL.
  • 3Run a narrow query with LIMIT and keep the source fields.
  • 4Widen only after checking provenance, freshness, and gaps.
  • 5Return SQL, rows, and record links so the answer is recoverable.

SQL

Treat this as a starting query. Inspect the live schema before widening it.

SELECT hn_id, title, original_author, original_timestamp, uri
FROM hackernews.items
WHERE title != ''
ORDER BY hn_id DESC
LIMIT 20;

Schema

POST /v1/scry/query The synchronous SQL-over-HTTPS endpoint for Scry: send bounded read-only SQL (ClickHouse dialect) as text and receive one JSON document or a typed HTTP error within the advertised execution ceiling.hackernews.items The enabled source-native Hacker News relation: public IDs, authors, timestamps, scores, payload text, and links.

Sources

Hacker News Every Hacker News story and comment, queryable with read-only SQL — source-native records, timestamps, authors, scores, and a live tail.