Process email text and images

Use General Document Parsing for EML or MSG files, inspect body text and available embedded or external images, then pass results to chunking, extraction, or indexing.

Prepare mail and attachments

  1. Select a small representative sample and confirm that the messages open correctly.

  2. Upload them to a Catalog volume in the current workspace and retain readable originals.

  3. Save attachments separately when their contents are required. Confirm that they open, then upload and parse them by format. Extract archives first. Have an authorized person decrypt protected attachments or obtain readable copies from their owner.

Successful body parsing does not establish that every attachment was processed. Compare the attachment inventory and inspect each required item separately.

Material

Processing

EML or MSG text and images

Use General Document Parsing below.

Separately saved PDF, Word, or other attachments

Select them through General Document Parsing.

Archives requiring structured header and recipient tables

Use Email archive parsing and check paths and output tables.

Configure and run

  1. Add General Document Parsing in the workflow editor.

  2. Select the uploaded messages or bind sources from an upstream volume-reading node.

  3. Configure parsing and image options as needed. Prepare an available model for image understanding.

  4. Test with a small sample and inspect node output and errors.

  5. Compare text, images, and sources before connecting chunking, extraction, or knowledge-base indexing.

Check names, text, tables, and images against the originals. File counts and successful node status do not replace content inspection.

Check external images

HTML email may reference external HTTP(S) images. Parsing attempts to retrieve accessible public raster images. Loopback, private, link-local, and other nonpublic destinations are refused. Downloads have a 20 MiB limit, redirect and timeout limits, and actual image-content validation.

A failed download may leave the original URL in the result. Body text can remain readable while the image was not saved. Inspect actual preview and image results; a retained link does not establish a successful download.

Web pages, cloud-storage download pages, and resources requiring browser sign-in are different from directly accessible images. Obtain authorized file copies and upload them separately when needed. Do not attach passwords or tokens to URLs.

Troubleshoot

Symptom

Check and action

Garbled or incomplete text

Inspect the original in an email client and check encoding and file-specific errors.

Image remains a URL

Check public access, raster content, size, redirects, and network errors.

Body succeeds but attachment content is absent

Export required attachments separately and inspect their parsing results.

Attachment requires a password

Obtain an authorized readable copy and upload it separately.

Processing completes but retrieval fails

Check segmentation, indexing, and enabled state in Manage knowledge-base data.

Preserve the message name, run ID, and exact error. Verify corrected inputs with the same small sample first.

Last updated on