Pushing your own documents

If a system has no connector, you can push documents to LensHub over HTTP. A short script is usually enough.

Set it up

Connectors → Add connector, type Custom push. Name it after the system it represents.
Note the connector's ID from its page.
Create an API key with write access.

Push documents

curl -X POST https://lenshub.example.com/api/ingest/SOURCE_ID/documents \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "documents": [
      {
        "external_ref": "runbook-042",
        "name": "Restarting the batch processor",
        "description": "What to do when the nightly job wedges",
        "body": "# Restarting the batch processor\n\nFull text here...",
        "tags": ["runbook", "operations"]
      }
    ]
  }'

The fields

Field	Required	Notes
`external_ref`	yes	Your stable identifier. Pushing the same ref again updates rather than duplicates
`name`	yes	Title, as it appears in results
`body`	yes	The content. Markdown renders nicely
`description`	no	One line of context
`tags`	no	Strings
`archived`	no	`true` removes it from search

Updating and removing

Updating: push the same external_ref with new content. A new version is written only if the content actually changed, so re-pushing everything nightly is cheap and safe.

Removing: push the same external_ref with "archived": true. It stops appearing in search and remains recoverable.

Limits

Up to 100 documents and 5 MB per request. Split larger batches.

The request returns immediately and processing continues in the background — documents become searchable shortly after.

A pattern that works

Run a script on a schedule that pushes everything, using stable external_refs. Unchanged documents cost nothing, changed ones update, and you never have to track what you sent last time.