[Improvement] Fix performance concerns — memory, N+1 queries, unbounded caches

[d-oit]

Description

Several performance concerns could impact users with large knowledge bases:

Memory Issues

Location	Issue
App.tsx refreshData	Calls getAllEntities() + getAllLinks() — loads everything into memory with no pagination
Editor.tsx	repository.getAllEntities().then(setAllEntities) for mention menu — no virtualization
search.ts initSearch	Bulk-loads ALL entities and claims into Orama index on init — no incremental hydration
search.ts oramaIdMap	In-memory Map grows unbounded — no eviction
@huggingface/transformers	~80MB model download for embeddings, loaded eagerly

N+1 Queries

Location	Issue
ExportPanel.tsx	Sequential getClaimsByEntityId + getNotesByEntityId per entity — classic N+1

Unnecessary Re-renders

Location	Issue
GraphView.tsx	clearEdges() on every update, then re-adds all edges — should diff edges

Missing Debounce

Location	Issue
Chat.tsx handleSend	No debounce — rapid submissions could flood search

Recommended Fix

1. Pagination: Add cursor-based pagination to getAllEntities() and getAllLinks()
2. Virtualized mention menu: Use @tanstack/react-virtual for the entity mention dropdown
3. Incremental search init: Load search index progressively, not all at once
4. Cache eviction: Add LRU eviction to oramaIdMap with max size
5. Batch queries: Add getAllClaimsWithNotes() to repository to avoid N+1
6. Edge diffing: Diff graph edges instead of clearing and re-adding
7. Lazy embeddings: Load @huggingface/transformers only when semantic search is enabled
8. Debounce chat: Add 300ms debounce to Chat submit

Acceptance Criteria

• refreshData supports pagination
• Mention menu uses virtualization
• Export uses batch queries instead of N+1
• Graph edges are diffed, not cleared
• oramaIdMap has max size with eviction
• Chat has debounce on submit
• @huggingface/transformers loaded lazily

[github-actions]

👋 This looks like a duplicate of #197.

The canonical issue has broader scope and/or more tasks — please subscribe there for updates.

⏳ This issue will be auto-closed in 3 days unless you react with 👎 to dispute.

Detected via title similarity analysis