Understanding the "Top K" Element Setting
What Is "Top K"?
Top K controls how many of the most relevant pieces of content webAI returns when you perform an embedding-based search.
When you ask a question or enter a query, WebAI searches your knowledge base and finds the Top K most similar results, based on meaning, not just keywords.
Think of it as:
"Show me the top 5, 10, or 15 chunks that are most relevant to what I just typed."
How It Works (No Code Needed)
Here's what happens behind the scenes when you search:
- You ask a question in webAI.
- webAI converts your question into a special format called an embedding (a vector that represents meaning).
- It compares this vector to all your content (which has also been embedded).
- Based on your Top K setting, it returns the closest matches.
Where to Set Top K
You'll find the Top K setting in the webAI user interface:
- Go to your Vector Retrieval Element Settings.
- Look for a setting labeled Top K
Select how many results you want returned for each search.
Choosing the Right Top K
| Goal | Suggested Top K |
|---|---|
| Focused answers | 3–5 |
| Broader overviews | 8–12 |
| Large topic exploration | 15–20 |
💡 Tip: Larger values return more content, but may also include irrelevant chunks. Start small, then increase if needed!
Example Scenario
Imagine you search:
"How do I implement Zero Trust in enterprise networks?"
- Top K = 3 → You get the 3 best matches.
- Top K = 10 → You also get related topics like identity access, network segmentation, and compliance frameworks.
How It Works with Your Embedding Model
Top K works with any embedding model you select in WebAI.
For example:
- mlx-community/multilingual-e5-small-mlx
- tasksource/ModernBERT-base
- snowflake-arctic-embed-l-v1
No matter which model you choose, WebAI will:
- Convert your query into an embedding.
- Compare it to your indexed content.
- Return the Top K closest matches.
Best Practices
- Use smaller Top K values for fast and precise queries.
- Use larger Top K values if you're summarizing long documents or exploring broadly.
- Keep in mind: Larger Top K values can increase response time and reduce model context precision.
Summary
Top K helps WebAI decide how many pieces of content to pull when searching by meaning. It's a key setting that affects:
- Answer accuracy
- Retrieval depth
- LLM context quality
Choose your Top K value based on how broad or focused you want your search results to be.