Search Tab
The Search tab allows you to test semantic search against your indexed documents directly within the application. This is useful for verifying that your documents are properly indexed and for testing search queries before integrating with your RAG application.
Performing a Search
- Navigate to Playground from the sidebar
- Ensure your desired project is selected (shown below the header)
- Enter a natural language query in the search box
- Example: "How do I reset my password?"
- Example: "What are the pricing tiers?"
- Optionally adjust Top K (5, 10, 15, or 20 results)
- Click Search or press Enter
Understanding Search Results
Search results are displayed as cards, each showing:
| Element | Description |
|---|---|
| Rank | Position in results (#1, #2, etc.) |
| Similarity Score | Percentage match (higher = more relevant) |
| Content Preview | Extracted text from the matched chunk |
| Document Title | Name of the source document |
| Open Link | Direct link to the original source (if available) |
| Chunk Number | Which chunk within the document matched |
Similarity Score Colors
| Score Range | Color | Meaning |
|---|---|---|
| 80%+ | Green | High relevance - strong semantic match |
| 50-79% | Yellow | Moderate relevance - partial match |
| Below 50% | Gray | Low relevance - weak match |
Search Metadata
After each search, a metadata footer displays:
- Vectors searched: Total number of vector embeddings in your current project
- Search time: How long the search took (in milliseconds)
- Embedding model: The model used to generate the query embedding
- Dimensions: Vector dimensionality (e.g., 1536 for OpenAI text-embedding-3-small)
- Vector store: Which provider stored your embeddings (Supabase, Pinecone, etc.)