A background process that automatically reviews recent session transcripts and consolidates insights into your memory files, similar to running a manual memory consolidation pass but triggered automatically.
Status
Feature-flagged via tengu_onyx_plover — disabled by default (enabled: !1).
Details
When enabled, triggers after a configurable minimum hours (default: 24h) and minimum sessions (default: 5) since last consolidation
Uses a .consolidate-lock file with PID-based locking to prevent concurrent consolidation
Runs as a forked background query with querySource: "auto_dream" that reads session transcripts and updates memory files
Includes throttling to prevent excessive scanning (minimum gap between scans)
The dream prompt instructs the model to orient by reading existing memories, gather signal from recent sessions, consolidate into topic files, and prune the index
Dream infrastructure (search for "autoDream") — gated by tengu_onyx_plover flag (search for "tengu_onyx_plover")
Strings lifted out of the shipped bundle, so the claim above can be checked against them.
Read from
Feature flag
tengu_onyx_ploverNot enough to say
Nothing here resolved what this flag was doing on this version, so nothing here should be read as on or off.
This account: no value returned · anonymous baseline: no value returned · compiled default in v2.1.78: no gate table built for this version
These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release.
Flag reading movedThe flag server now returns {"enabled":false,"minHours":24,"minSessions":3,"remoteEnabled":false} for tengu_onyx_plover, read as this account.A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
Flag reading movedThe flag server now returns nothing at all, having stopped carrying it for tengu_onyx_plover, read as this account.A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
