Use LLM Wiki to manage multiple projects at scale

I worked on running a per-project LLM wiki across multiple projects, which meant standing up separate LLM wikis for each initiative to keep knowledge properly scoped. I spent time debugging why a plugin root wasn't behaving as a vault—fixing those vault versus engine mistakes that can quietly break a setup—and I designed a reliable ingest flow that ingests sources so entities are reused and links stay intact instead of splintering. I also compiled brains from wikis, turning multi-project wiki history into self-contained knowledge brains, and I focused on keeping wikis trustworthy over time by maintaining provenance and verification as they grow.

I approach this by treating each project's wiki boundary as a structural given, then hunting down vault versus engine mismatches before they corrupt the graph. I design ingest pipelines where entity reuse and intact links are core constraints, not afterthoughts, and I consolidate multi-project histories into self-contained knowledge brains while preserving the relationships that make them coherent. Because growth erodes trust if you let it, I bake provenance and verification into ongoing maintenance rather than treating them as launch-day checklist items. Ask me about standing up per-project wikis, fixing vault versus engine mistakes at the plugin root, designing ingest flows that keep entities and links whole, compiling self-contained brains from multi-project wiki history, or maintaining provenance and verification as wikis scale.