Imports and exports
Choose the interface by source format and transaction boundary. File paths are resolved by the process executing the query, usually the Bolt server, not your laptop.
| Input or task | Interface | Commit boundary |
|---|---|---|
| Small CSV | LOAD CSV | One statement |
| Large CSV | CALL IN TRANSACTIONS | Independent batches; earlier batches survive a later failure |
| Application row data | UNWIND over Bolt | One batch per statement |
| APOC JSON/JSONL | apoc.import.json | One atomic statement; input is materialised |
| Cypher script | gdb.cypher.import | Each statement commits; failures are counted and skipped |
Neo4j 4.x/5.x .dump | gdb.neo4j.import or Explorer Import | One atomic statement; empty target required |
| Native GDB snapshot | gdb.restore | Replaces the selected database |
| Logical JSON/Cypher export | apoc.export.* | Text output is materialised; file effects do not roll back |
Use gdb.restore for native GDB snapshots and gdb.neo4j.import for supported
Neo4j record-store dumps. Unsupported store formats and property encodings need
a logical export. Validate schema and endpoint counts before
a live migration.
Articles
-
Native Neo4j 4.x/5.x dump imports: supported standard/aligned stores, empty targets and atomic rollback.
Related articles
Backup, restore and transfer · Transactions and concurrency · Server configuration