Skip to main content
Semantic search is an optional mode in the CMD‑K terminal that turns natural‑language questions into structured Rover queries, so you do not have to start by thinking in building blocks and operators. Iss

​
What semantic search does

When the semantic search toggle is on, the terminal interprets your free‑form text and builds the corresponding legos for you. It understands entities like ENS names, token symbols, function selectors, addresses, blocks, and views, then snaps them into the right filters with appropriate operators and value types. Instead of manually assembling Transaction Fee | is at or above | 0.01 ETH, you can type “transactions with fees above 0.01 ETH,” and semantic search generates the filters needed to match that intent.

​
Working with on‑chain entities

Semantic search recognizes common on‑chain identifiers and metadata to speed up query construction.
  • ENS names with autocomplete and resolution, so typing “vitalik” suggests vitalik.eth and resolves the address.
  • Popular tokens from the canonical token registry (stablecoins, wrapped tokens, DeFi, meme, and exchange tokens), so “USDC transactions” becomes an Asset | is | USDC filter with the correct logo and address.
  • Function selectors and names, so phrases like “transactions calling function 0xa9059cbb,” “function equals transfer,” or “function contains swap” become function filters in the Transactions category.
Because these entities are interpreted before filters are created, the resulting legos use canonical addresses, symbols, and metadata.

​
Smarter filters, ranges, and logic

Semantic search can infer comparison operators, numeric ranges, and boolean logic directly from your wording.
  • Range phrases such as “blocks with base fee between 50 and 150 gwei” are turned into two filters: one “at or above” 50 and one “at or below” 150.
  • Queries like “blocks with miner reward above 2 ETH,” “blocks with gas used over 30,000,000,” or “blocks with over 200 transactions” map to greater‑than filters on the correct block fields.
  • Statements such as “USDC transactions but NOT from 0x123…” or “(failed or value above 10 ETH) and gas below 30 gwei” are turned into grouped logic using AND, OR, and NOT.
Semantic search also distinguishes between similarly named properties on different categories (for example, block gas vs transaction gas, block base fee vs transaction base fee) so filters land where you expect. Iss Example

​
Using @ for views and ENS

The @ symbol provides a fast way to reference views and ENS names in your query.
  • Type @ to search for and insert a view as input to a building block, such as Participants @Engineering Wallets, instead of listing individual addresses.
  • Type @ followed by an ENS name, like @vitalik.eth, to resolve and reference that address directly in your filters.
This keeps complex searches readable while letting you reuse existing address sets and identities. At Symbol

​
Example queries

Semantic search supports both simple and complex patterns. Transaction filters
  • “Transactions from 0xd8da6bf26964af9d7eed9e03e53415d37aa96045.”
  • “Show transactions to 0xd8da6bf26964af9d7eed9e03e53415d37aa96045.”
  • “Get transactions sent by 0x742d35Cc6634C0532925a3b844Bc9e59….”
  • “Get transactions received by 0x742d35Cc6634C0532925a3b844Bc9e59….”
Each becomes a Transactions query with From, To, or Participants filters set to the specified address. Address + balance with views
  • “Find addresses from Metamask view with balance over $10,000.”
  • “Show Metamask addresses with balance above $5,000.”
  • “Get addresses from Metamask Swap view with balance exceeding $1,000.”
  • “Addresses in Metamask view with more than $50,000.”
These are interpreted as Address queries that apply the named view and combine it with a USD balance range filter. You can mix these patterns with token, block, and function phrases. For example, “USDC transactions above $10,000” or “blocks with base fee above 100 gwei” and semantic search will create the corresponding structured filters.

​
Performance and instant patterns

For very simple inputs, semantic search returns results immediately by recognizing hashes, addresses, or “show all” patterns without extra steps. As legos are generated, they will appear one by one in the query bar, making it easier to understand how your plain‑language request maps to Rover’s structured search model.