Skip to content

strands.storage.search.keyword

Keyword search strategy using token-overlap scoring.

STOP_WORDS

Section titled “STOP_WORDS”

Common English words dropped from queries before matching.

tokenize

Section titled “tokenize”
def tokenize(text: str) -> set[str]

Defined in: src/strands/storage/search/keyword.py:151

Lowercase and split text into a set of word tokens, dropping empties.

Splits on any run of non-word characters (Unicode-aware). Ensures cross-SDK compatibility with the TypeScript /[^\p\{L}\p\{N}_]+/u regex.

Arguments:

  • text - The text to tokenize.

Returns:

A set of lowercased word tokens.

token_overlap_score

Section titled “token_overlap_score”
def token_overlap_score(query_tokens: set[str], content: str) -> int

Defined in: src/strands/storage/search/keyword.py:166

Lexical relevance score: distinct query tokens present in the content.

A higher count means more of the query’s words are present. Returns 0 when there is no overlap.

Arguments:

  • query_tokens - Pre-tokenized query terms.
  • content - The content string to score against.

Returns:

Number of distinct query tokens found in the content.

KeywordSearchStrategy

Section titled “KeywordSearchStrategy”
class KeywordSearchStrategy()

Defined in: src/strands/storage/search/keyword.py:182

Keyword search strategy using token-overlap scoring.

Tokenizes the query and each stored entry (key without its file extension + content), then scores by the number of distinct query tokens that appear. Stop words in the query (the, is, what, …) are ignored. Works on any storage backend with list() and read() — no index or embedding model required. Entries are read in small concurrent batches, and each result carries the bytes read during the scan. Entries whose read() raises StorageError are skipped.

This is the default search strategy for all shipped storage backends.

Example:

from strands.storage.search import KeywordSearchStrategy


strategy = KeywordSearchStrategy()
results = await strategy.search(storage, "dark mode toggle")

index

Section titled “index”
async def index(storage: Storage, key: str, data: bytes,
                **kwargs: Any) -> None

Defined in: src/strands/storage/search/keyword.py:205

No-op — keyword search scans storage on the fly.

Section titled “search”
async def search(storage: Storage, query: str,
                 **kwargs: Any) -> builtins.list[StorageSearchResult]

Defined in: src/strands/storage/search/keyword.py:208

Search content in storage by keyword token-overlap scoring.

Arguments:

  • storage - The storage to search over.
  • query - A natural-language string query.
  • **kwargs - Unused; accepted for protocol compatibility.

Returns:

Matched keys with relevance scores, ranked best-first.