feat: Embedded Graph Engine for academic researcher network analysis #1
Description
Summary
Add an embedded graph database engine to che-zotero-mcp for academic researcher network analysis with in-memory adjacency lists, custom binary persistence, and graph algorithms.
Architecture
- Property graph with 4 node types (Researcher, Paper, Institution, Journal) and 6 edge types (AUTHORED, CO_AUTHOR, PUBLISHED_IN, AFFILIATED_WITH, CITES, ADVISOR_OF)
- In-memory index-free adjacency (nodes hold direct pointers to edges)
- Binary persistence to
~/.che-zotero-mcp/graph.binwith custom format (magic "CHEG", string table, linked lists) - Simplified Cypher query support (MATCH/WHERE/RETURN)
New MCP Tools (13)
| Tool | Description |
|---|---|
graph_stats |
Graph statistics |
graph_add_node |
Add node (Researcher/Paper/Institution/Journal) |
graph_add_edge |
Add edge with type |
graph_remove_node |
Remove node and connected edges |
graph_remove_edge |
Remove edge |
graph_save |
Persist graph to disk |
graph_neighbors |
Find neighbor nodes |
graph_shortest_path |
BFS shortest path |
graph_co_author_stats |
Co-authorship statistics |
graph_citation_network |
Citation tree traversal |
graph_community |
Community detection |
graph_query |
Simplified Cypher queries |
graph_import_from_zotero |
Import from Zotero collections |
Implementation Plan
See docs/plans/2026-03-08-embedded-graph-engine-plan.md for detailed task-by-task TDD plan (12 tasks).
Tech Stack
Swift 5.9, XCTest, MCP swift-sdk