Schema
Every table the reader role can reach, clustered by schema, with foreign keys as edges. Cached server side and refreshed on a TTL, so the timestamp in the corner is the one to trust.
At a glance
Live count(*) on the tables that say whether data is still flowing. Never planner estimates: most tables on this instance have never been analyzed and would read as empty.
Extraction
Claims and relationships pulled out of the corpus by the extraction runs. This is where the volume is, and it comes before the curated layer for that reason.
Corpus
The source documents and the chunks they were split into for embedding. Select a document to filter its chunks.
Commons
The curated claim layer: canonical claims, the edges between them, and the audit trail behind each change. Small by design, 83 rows against extraction's 36,618.
Agents
Leo, Rio and Theseus. Grant coverage across the three databases is uneven, so a surface that is not readable is labelled as such rather than shown as an empty table.
Governance
Admission receipts, and the telos inquiries and passes that gate what is allowed into the commons.