Expand description
Publishing Mjolnir’s own checkpointed sessions into the user’s SessionWiki index, and the daemon-side job that keeps that index current.
SessionWiki keeps one searchable index of AI coding sessions across every
tool a user runs. Mjolnir links it as a library and registers
MjolnirAdapter beside SessionWiki’s built-in adapters, so a Mjolnir
session is searchable next to a Claude Code or Codex one. The adapter is a
“shared store” adapter: checkpoints are not one-file-per-session in a shape
SessionWiki can parse, so the indexer enumerates sessions by key and asks
this adapter to parse the ones whose checkpoint changed.
Structs§
- Archived
Session - What a restore needs from the index: the transcript as a snapshot the compaction pipeline accepts, plus the title and project of the session it came from.
- Mjolnir
Adapter - Mjolnir’s sessions, as SessionWiki sees them.
- Wiki
Indexer - The daemon’s SessionWiki sync job.
Constants§
- DEFAULT_
WIKI_ LIMIT - The page size a caller that names none gets.
- MAX_
WIKI_ LIMIT - The largest page a caller may ask a wiki query for.
- SYNC_
STALE_ AFTER - How stale the index may be before a query triggers a background sync.
Functions§
- archived_
session - Load one indexed session for restore, or
Nonewhen the id names none. - brief
- The briefing for one indexed session, or
Nonewhen the id names none. - index_
state - What a surface should say about this index right now.
- indexed_
with_ messages - Which of
session_idsthe index holds under this instance’s own key, with at least one message and not already archived. - query_
rows - One page of the index, newest first or best match first.
- sessions_
ready_ to_ archive - The stopped sessions that
archive_after_days = older_than_dayshas caught, children before their parents. - sync_
is_ stale - Whether a query should trigger a bounded background sync before it answers.