median quest
Hi, I’m trying to understand the intended memory-wiki workflow.
Setup: memory-core + lossless-claw, memory-wiki in bridge mode, imports/dream reports/events enabled, autoCompile/dashboards on, shared/all search.
What works:
wiki_status shows the vault/bridge as healthy.
Bridge imports source pages from memory.
wiki_search finds source pages and synthesis pages.
wiki_apply synthesis ... works.
wiki compile and wiki lint work.
If I manually create pages under entities/ and concepts/, wiki_status counts them and wiki_search ranks them correctly.
Observed issue:
Before manually creating entity/concept pages, the vault had many sources, some syntheses and reports, but:
0 entities
0 concepts
The docs describe memory-wiki as a compiled knowledge vault with sources/, entities/, concepts/, syntheses/, reports/, structured claims/evidence, dashboards, etc.
But I can’t find a CLI/tool path that actually extracts or generates entity/concept pages from bridge sources.
openclaw wiki apply --help only exposes:
synthesis
metadata
No obvious command like:
extract entities
generate concepts
promote sources to entity/concept
pages
backfill wiki knowledge layer
Manual proof-of-concept:
I manually created a few generic entity/concept pages with pageType, stable id, sourceIds, structured claims, and evidence pointing back to bridge source IDs.
After that, compile/lint/search all worked as expected.
Questions:
- Is automatic entity/concept extraction currently implemented in memory-wiki, or are these pages intended to be created by an agent/curator workflow?
- If implemented, what command/config triggers it?
- If not implemented yet, is the intended workflow currently: bridge import → curated synthesis/entity/concept creation by agent → compile → lint?
Should wiki_apply eventually support entity/concept page creation directly, similar to synthesis?
Trying to align with the documented architecture before building a custom extractor.