What happens after you commit

The workspace runs background jobs automatically — you don't trigger any of this.

Commit
  ↓  Parsers        extract plain text from PDF, Office, HTML …
  ↓  NLP pipelines  tokenise · detect language · extract entities · generate embeddings
  ↓  Indexers       populate text and vector search indexes
  ↓
Searchable

What you can tune (once the data is in):

• Which fields are indexed for text search
• Which fields get vector embeddings and with what chunking
• Which entity types the NLP extractor links back to graph nodes (e.g. product names → Product nodes)

These are workspace-level settings — covered in the NLP and Search presentations.

Before going to production:

• Running the connector twice produces identical node and edge counts
• Cursor advances only after a successful commit
• Every access-controlled node has RestrictAccess* set
• An end-user test account sees the data it should — and only that data
• Deletes from source eventually propagate to the graph
• Credentials come from environment variables, not source code

What to read next: