Table

Ingest and search on a table

Table.add(docs: list) -> int

Ingest documents. Each item is a str or dict with a text field.

n = table.add([{"text": "Nikola Tesla AC motor", "tag": "history"}])

Table.add_arrow(arrow_table) -> int

Ingest a PyArrow table via the C Data interface. Requires pyarrow.

Table.search(query, top_k=None, offset=None, strategy=None, explain=None, graph_expand=None, depth=None, query_vector=None) -> SearchResults

query · str · required

Query text (and metadata filter syntax when supported).

top_k · int · default: "20"

Maximum hits to return.

offset · int · default: "0"

Skip first N results.

strategy · str

sparse, dense, diskann, hyde, crag, distributed, hybrid, etc.

explain · bool · default: "false"

Attach explain text to results.

graph_expand · bool · default: "false"

Expand results via graph neighbors.

depth · int · default: "2"

Graph expansion depth.

query_vector · list[float]

Dense query embedding for ANN strategies.

results = table.search("Nikola Tesla", top_k=5, strategy="hybrid")

stream_search(table, query, batch_size=128, ...)

Module helper in toradb.table — yields SearchResults batches.

from toradb.table import stream_search
for batch in stream_search(table, "Tesla", batch_size=32):
    print(batch.to_pandas())

See Search strategies.