Workplace tools
Eight connector types read records from tools your team already works in. They
run on dlt's verified sources, which ship inside the server, so they are always
available; ENABLE_DLT is not needed for them.
For published help centres, sites and course pages, see Help centres, sites and trackers.
These sources have not yet been verified against live vendor accounts. In the automated tests only Jira is run, against canned vendor responses. Run a first sync on a small scope and check the run before relying on one.
Read this first
- Whole workspace, no per-person sharing. A source indexes what its credentials can read and makes it available to everyone who can use the project. Tickets, deals and channel messages often contain things not meant for everyone, so use credentials limited to what all of your users may see, and choose channels and record types with care.
- Only knowledge-like records are indexed. For example, Jira issues are indexed; Jira users and workflows are not.
- Most syncs read the whole source. Unchanged records are skipped, and a record that is no longer there is deleted from the index. Slack is the exception: after the first sync it re-reads the last ten days of each channel (which picks up edits and new replies) and reads everything once a week; messages deleted at the source are removed at that weekly read.
- A capped read deletes nothing. When GitHub reaches
max_items, the run is treated as partial, so older issues stay in the index. - Deleted Zendesk tickets are removed from the index.
- They are created like any other connector: see Connect a source. Field names below are keys of
credentialsandsettings.
What each one needs and indexes
| Type | Credentials | Settings | Indexes |
|---|---|---|---|
dlt:slack | access_token (a bot token) | selected_channels (required; the bot must be a member of each), replies (default false) | Messages from the chosen public channels, and thread replies when replies is on. Joins, leaves and bot messages are left out. A channel the bot cannot see is reported on the run as not_configured |
dlt:jira | email, api_token | subdomain (the part before .atlassian.net) | Issues of a Jira Cloud site |
dlt:hubspot | api_key (a private app token) | resources (default companies, deals, tickets; also products, quotes, contacts) | The chosen record types |
dlt:zendesk | email, token | subdomain | Support tickets |
dlt:github | access_token | owner, name, max_items (default 500, most recent first) | Issues and pull requests of one repository, with their comments |
dlt:freshdesk | api_secret_key | domain (the part before .freshdesk.com) | Support tickets |
dlt:pipedrive | pipedrive_api_key | resources (default deals, organizations, products) | The chosen record types |
dlt:google_sheets | credentials (a service account JSON key the spreadsheet is shared with) | spreadsheet_url_or_id, range_names (sheets or ranges, required) | Rows of the chosen ranges |
GET /api/v1/connectors/types returns the same fields for your deployment.
Which one to use
| You want | Use |
|---|---|
| Zendesk help centre articles | zendesk_help_center |
| Zendesk tickets | dlt:zendesk |
| GitHub issues only, without comments | github_issues |
| GitHub issues and pull requests with comments | dlt:github |
| Slack as a place where the assistant answers | The Slack channel, which is separate from this source |
Generic pipelines
dlt:sql_database, dlt:rest_api and dlt:filesystem (S3, GCS, Azure) are a
different group. They are off unless the deployment sets ENABLE_DLT=true.
They remove deleted records only when every record they read carries an id.
Not included
Asana and Salesforce are not included; their dlt sources need extra dependencies that the server does not ship.