Is there a way to write the graph to Neo4j instead of SQLite (or a supported export path)?

[haiderRizvii]

Hi, and thanks for the project.

Context

• Version: codebase-memory-mcp 0.10.8, Windows 11
• We index a large multi-repo tree (about 1,450 repos, roughly 26 GB after sparse checkout) as a single project with cli index_repository --repo-path --name .
• We would like to use the resulting graph in Neo4j (some graph database) so we can run native Cypher, use graph tooling, and share one graph server across the team.

What I looked at

• codebase-memory-mcp --help shows no storage or output backend option.
• In the source, the indexer builds the graph in RAM (graph_buffer.h) and then writes it with cbm_gbuf_dump_to_sqlite, cbm_gbuf_flush_to_store and cbm_gbuf_merge_into_store. The query tools read from the same SQLite store.
• A search for neo4j, bolt and graphml finds nothing in the code, only a comment in tests/test_cypher.c.

Questions

1. Is a Neo4j (or other graph DB) backend supported, planned, or something you would accept as a contribution? If so, is there a storage abstraction I should target instead of cbm_gbuf_dump_to_sqlite?
2. If not, is there a recommended way to export the graph (nodes, edges and properties) to Neo4j? Rodes and edges tables from the .db and load them over Bolt. Is that safe, i.e. is the SQLite schemaconsidered stable across versions, or should I go through the CLI/MCP tools instead?
3. For the SQLite approach: are there columns I should treat specially, such as the properties JSONe key, and the generated columns on edges? Are there edge types that are derived and better skipped, such as SIMILAR_TO and SEMANTICALLY_RELATED?
4. Is there a hook or event at the end of indexing that I could use to trigger a sync, for example etect_changes output?

Thanks!

[DeusData]

Hey, thanks for the question. Yes this is in scope but currently in the planning phase. As this will be quite an architectural shift from the current baseline, it will take likely a couple weeks until this is done. If you like we can build this together, feel free to reach out in discord for further discussions. To give a small headsup: We plan generally an integration layer for different parts of the project, one which is also in this case the storage option