Auto-tag new NetX assets with AI and review in Slack

On a cron every hour during business hours (Monday to Friday, 9am to 6pm in my local timezone), find newly imported NetX assets that are missing tags or descriptions, use AI vision to draft metadata for each one, and queue the suggestions in a Slack channel for a librarian to review.

On the first run, call NetX List Attribute Definitions to discover which custom attribute fields actually exist on this instance (for example Keywords, Description, AltText, Title). Cache the schema and only write to fields that are defined. If a typical field name is missing, skip writing that piece of metadata rather than erroring out.

Each run, call NetX Search Assets by Query to find assets created since the last successful run that have empty or sparse attributes (no keywords, no description, no alt text). Persist the timestamp of the most recent asset processed so the next run picks up where this one left off. Skip assets that already have rich metadata. Skip assets whose file type is not an image or a PDF, since vision analysis cannot meaningfully handle other formats.

For each remaining asset, call NetX Download Asset File to grab the image (or the first-page proxy for a PDF). Run vision-based analysis on the file and produce: a short title (under 60 characters), alt text (one sentence, accessibility-focused), a one-sentence description, and 5 to 10 keyword tags. Keep the language plain and avoid speculation about brand or people identities you cannot confirm from the image alone.

Write the suggestions back to the asset using NetX Set Asset Attributes, mapping each piece of generated metadata to the matching custom field discovered earlier. Treat these as draft values for human review, not final publication.

Then post a message to a designated Slack review channel using Slack Bot Send a Message. Each message should include: the asset thumbnail (use the NetX proxy URL so it renders inline), the suggested title, the suggested keyword tags, the one-sentence description, the alt text, and a direct link back to the asset in NetX so the librarian can confirm or correct in one click. Group related assets into a thread when more than five new assets show up in a single run so the channel stays scannable.

Make the Slack channel, the business-hours window, the sparse-metadata definition, and the maximum number of assets processed per run configurable. Default to processing at most 25 assets per run so the queue stays manageable. Log every asset processed, every skip reason, and every Slack post so the team can audit what the AI suggested.

Additional information

What does this prompt do?

Runs every hour during business hours and finds assets newly imported into NetX since the last check.
Looks at each image or PDF and drafts a short title, alt text, a one-sentence description, and 5 to 10 keyword tags.
Saves the suggestions as draft values on the asset so a librarian can confirm or correct them inside NetX.
Posts a Slack message to your review channel with the thumbnail, the suggested tags, and a link back to the asset for one-click sign-off.

What do I need to use this?

A NetX login with permission to read assets and write custom attributes.
A Slack workspace and a dedicated channel for the asset review queue.
A short list of which NetX custom fields you want filled in (for example Keywords, Description, Alt Text).

How can I customize it?

Change how often it runs, or limit it to specific business hours or weekdays.
Narrow the search to certain NetX folders, file types, or import sources.
Pick which Slack channel receives the review queue, or route different asset types to different channels.
Adjust which metadata fields get filled in and how many keyword tags the AI suggests.

Frequently asked questions

Will this overwrite metadata my librarian already added?

No. It only targets assets with empty or very sparse metadata, and it writes the AI suggestions as drafts so a human still has to confirm them inside NetX before they go live.

Does it work for videos or other file types?

It focuses on still images and PDFs because those are what the AI vision model can meaningfully analyze. Videos, audio, and unsupported file types are skipped.

How do reviewers approve a suggestion?

They open the asset link from the Slack message, confirm or edit the draft values in NetX, and save. Nothing is auto-published without a human in the loop.

What if my NetX instance uses different custom fields?

On the first run it reads your custom field schema and only writes to fields that actually exist in your library, so it adapts to whatever Keywords, Description, or Alt Text fields you already use.

Can I send different asset types to different reviewers?

Yes. You can route by folder, file type, or import source so product photography goes to one Slack channel and marketing creative goes to another.