Ingests every non-ledgered buffer file, resolves model-call run ids, and (re)applies the consumer views. See the README.
Only --sessions is a spool: each session file that ingests is moved
into <dir>/.ingested/. --gateway and --evals files are left where
they are — other tools read those directories — and re-runs skip them
through the ingest ledger.
The spool's write protocol: a producer writes each file once, under a
dot-prefixed temporary name in the same directory, and renames it to its
final *.jsonl name when it is complete. Only those published names are
read; a temporary file is neither ingested nor moved.
--db accepts a local path (dev/test) or motherduck:<db_name>
(deployment; token from MOTHERDUCK_TOKEN).