Text Retriever

TextRetriever is the default text-oriented retrieval implementation. It embeds queries on the application side, calls the raw GraphDB search primitives, and deduplicates overlapping hits.

grawiki.retrieval.text.TextRetriever

Bases: Retriever

Embed queries, call DB primitives, and compose results.

The retriever is the single place where query-side embedding happens. The DB layer receives only pre-computed vectors and raw text; ranking and deduplication are concerns of this class.

Parameters:

Name	Type	Description	Default
db	GraphDB	Storage engine adapter used for raw search primitives.	required
embedding	Embedding	Shared embedding client instance. The same instance should be reused by the ingestion path to avoid loading the model twice.	required

retrieve async

retrieve(query, limit=5, *args, **kwargs)

Run a retrieval query and return a list of hits.

Parameters:

Name	Type	Description	Default
query	str	Raw query text.	required
limit	int | None	Maximum hits per label. Overrides the default limit if provided.	5

Returns:

Type	Description
list[NodeHit]	Flat, deduplicated hit list across all requested labels.

fulltext async

fulltext(query, *, labels, limit=10)

Run a full-text search across the given node labels.

Parameters:

Name	Type	Description	Default
query	str	Raw query text forwarded to the DB full-text index.	required
labels	list[str]	Node labels to search.	required
limit	int	Maximum hits per label.	10

Returns:

Type	Description
list[NodeHit]	Flat, deduplicated hit list across all requested labels.

vector async

vector(query, *, labels, limit=10)

Embed a query and run a vector similarity search.

Parameters:

Name	Type	Description	Default
query	str	Raw query text. The retriever embeds it using :attr:embedding.	required
labels	list[str]	Node labels to search.	required
limit	int	Maximum hits per label.	10

Returns:

Type	Description
list[NodeHit]	Flat, deduplicated hit list with higher scores representing more relevant matches.

Raises:

Type	Description
ValueError	Raised when the embedding returns an empty result.