Status: Connectors are not implemented in the current Spectron binary. There are no
/connectorsREST routes and no background sync workers. This page describes the intended direction; API shapes may change before release.
Intended behaviour
A connector would be configuration that tells Spectron:
Where to fetch data (warehouse, SaaS workspace, repository)
What to fetch (SQL query, folder, page set)
When to run (cron or webhook)
Which scope owns the ingested content
Successful syncs would land as document records (or structured fact batches for warehouse records), then flow through the same parse → chunk → extract → embed pipeline as manual uploads.
Planned source types
Data warehouses (structured records)
| Source | Intended use |
|---|---|
| Snowflake | SQL result sets → fact batch or tabular documents |
| Databricks | Same pattern for lakehouse tables |
SaaS document sources
| Source | Intended use |
|---|---|
| Google Drive | Docs and files → POST /documents |
| Notion | Pages and databases |
| GitHub | Repo files and wikis |
| Confluence | Spaces and pages |
Until connectors ship, ingest via:
spectron documents upload/POST /api/v1/{ctx}/documentsspectron remember --from-filefor JSONL transcriptsCI jobs that push bytes or facts on a schedule
What replaced “knowledge nodes”
When connectors ship, warehouse and SaaS sync will feed the unified graph (entities, attributes, relations with source.kind = "document").
Per-connector pages
The Notion, GitHub, Google Drive, Confluence, and warehouse pages in this section are design companions – they describe intended sync behaviour, not live API documentation. Keep them if you are planning integrations; skip them if you only need manual upload today.