Prepare a deduplicated feed digest

What this template does

Normalize HTTPS links, remove common tracking parameters, retain the first title and keep a valid publication date when one is supplied. It does not fetch a feed or summarize an article.

Prepare these inputs

Parsed entries with title, url and optional YYYY-MM-DD published date. Keep feed retrieval and any permission checks in your approved upstream system.

Field map: title, url, published

From input to reviewable output.

Use the synthetic example or edit its JSON structure. This demo sends no input to an AI service or our server. Keep real personal data in your approved environment. Up to 1,000 records and 100 KB of JSON.

Ready. Run the sample to inspect the result.

Use the starter in n8n

1. Download the JSON and import it using n8n’s import-from-file option. Keep it unpublished.
2. Inspect the three nodes: Manual start → Sample input → Reviewable output. Run it manually with the synthetic example.
3. Compare records, rejected and the counts with the demo here. The output is one item containing an envelope; split the records array if a later node expects individual items.
4. Replace Sample input with an approved source only after mapping its data to the same single-item JSON structure shown above. Supply an explicit as_of where required.
5. Add a separate review or approval step before any external write. Test credentials and permissions in your own environment, then decide whether a schedule is appropriate.

The download has no live connector, schedule, credential or external request. Connecting a mailbox, CRM, spreadsheet or notification destination is a separate setup and test.

SHA-256: 8a9c9e268bf89374011f0b1dec8005158c4846e78343b3ddd120a4a2137b2a80

Limits and review points

A later duplicate can contain a corrected headline. Inspect source freshness before publishing a digest. An accepted URL is not a security or uptime assessment. This uses a conservative ASCII HTTPS parser, with optional valid port and percent-encoded UTF-8 query values. IPv6 and raw Unicode URLs are rejected. Dot segments in paths are not resolved. This is string processing, not a full URL validator.

Rejected rows are returned with their one-based input position and a reason. Review them before using the result; an empty output can be a valid outcome. Unknown or malformed top-level inputs produce an error instead of a guessed result. Keep an untouched input export for reconciliation.

Run costs and data handling

The browser step has no service fee, API call or account requirement. It uses no browser storage for your input. Ordinary website access logs still apply. Downloaded results are saved only when you request them. Your own n8n hosting, cloud plan and any later connected service may have charges; no paid AI model is required by this starter.