Creating and Managing Data Streams

By constantly pulling, scoring, and filtering content from real sources, Data Streams inform your workflows with live context, verified quotations, and up-to-date insights from thousands of articles across the web.

Think of them as your brand-aligned research engine, automating the current-events curation your editors are doing manually. Except faster. Smarter. And always on.

This isn’t just AI. This is AI that reads the news, checks the sources, and gets your niche.

Step-by-Step: Creating a Data Streams

1. Start From the Dashboard
Go to Data Streams in the main navigation and click + New Data Streams.

2. Enter Your Data Streams Info
Fill in the required fields:

• Name – Give your stream a clear, searchable name like "US Sports" or "Finance – No Crypto".
• Description – Optionally describe what the Data Streams is for (e.g., "Daily digest for clean sports content excluding fantasy and betting").

3. Choose Data Streams Type
Select one of the following:

• Article – Scores and surfaces the strongest individual articles based on your filters. Great for Agents needing citations.
• Inspiration – Clusters and ranks groups of articles by trending topic, helping agents surface broader themes or brainstorms.

4. Set Your Time Frames

When to use what:

• Use short time frames (24–72 hours) for breaking news or daily updates.
• Use 7–14 days when targeting deeper trends, analysis, or slower verticals like research or policy.
• Set cluster time to 72 hours or more for better topic grouping in Inspiration Streams.

5. Add Filter Groups

Scroll to the Filtering & Scoring section and click Add Filter Group. You’ll use this to shape the stream.

You can add:

• Included Items
• Boosted Items (score 1 to 1000)
• Penalized Items (score –1 to –1000)
• Excluded Items
• Included Sources
• Boosted Sources
• Penalized Sources
• Excluded Sources

Use keyword search to quickly find relevant Items or Sources. You can select as many as needed.

6. Preview the Results

Once your Data Stream has a name and type selected, the Preview button becomes active. Clicking it will open a preview window where you can inspect the Stream’s performance in real time.

In this view, you can:

• See the top articles returned by your current filters
• View each article’s source, headline, and publish date
• Examine the keywords and metadata tags associated with the article
• Check the article’s score bar, which reflects how strongly it matches your filter criteria
• Identify outliers, duplicates, or irrelevant entries you may want to penalize or exclude

The preview is one of the most important parts of the setup process.
It allows you to test your logic, catch noise early, and refine the stream before linking it to an Agent.

If the results are aligned with your expectations, you can proceed and click Create.
If not, go back and add or adjust filter groups until the content feels right.

7. Create and Assign

Click Create when you’re satisfied.

Now you can assign this Data Stream to any compatible Agent by selecting it under the Source Input field during configuration.

Understanding the Metadata Filters

Trial, Error & Iteration

Data Streams get better the more you refine them.

• Always Preview First – Don’t guess. Check actual output.
• Spot Unwanted Noise – Seeing clickbait? Go back and exclude those Items or Sources.
• Duplicate to Experiment – Use the Duplicate feature to test new filters without overwriting the original.
• Name Clearly – Add suffixes like – No Politics or + Trend Boost to track versions.
• Refine Weekly – Adjust filters as your editorial goals evolve.

Pro Tips

• Keep Included Items minimal — 2–4 high-value tags is ideal.
• Use penalties instead of exclusions to reduce noise gently.
• Combine boosts with exclusions to steer tone and coverage.
• Use Inspiration for brainstorms, and Article for citation-heavy writing.
• Use ChatGPT to debug – paste in your preview results and ask for improvement tips.

Real-World Data Streams Examples

Finance Trends Data Streams

Health & Wellbeing Data Streams

General U.S. News Data Streams

Sports Highlights Data Streams