feat: search multiple query vectors as one query (#1811)

Allows users to pass multiple query vector as part of a single query
plan. This just runs the queries in parallel without any further
optimization. It's mostly a convenience.

Previously, I think this was only handled by the sync Python remote API.
This makes it common across all SDKs.

Closes https://github.com/lancedb/lancedb/issues/1803

```python
>>> import lancedb
>>> import asyncio
>>> 
>>> async def main():
...     db = await lancedb.connect_async("./demo")
...     table = await db.create_table("demo", [{"id": 1, "vector": [1, 2, 3]}, {"id": 2, "vector": [4, 5, 6]}], mode="overwrite")
...     return await table.query().nearest_to([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0], [4.0, 5.0, 6.0]]).limit(1).to_pandas()
... 
>>> asyncio.run(main())
   query_index  id           vector  _distance
0            2   2  [4.0, 5.0, 6.0]        0.0
1            1   2  [4.0, 5.0, 6.0]        0.0
2            0   1  [1.0, 2.0, 3.0]        0.0
```
This commit is contained in:
Will Jones
2024-11-13 16:05:16 -08:00
committed by GitHub
parent 0fd8a50bd7
commit abd75e0ead
9 changed files with 366 additions and 64 deletions

View File

@@ -492,6 +492,42 @@ export class VectorQuery extends QueryBase<NativeVectorQuery> {
super.doCall((inner) => inner.bypassVectorIndex());
return this;
}
/*
* Add a query vector to the search
*
* This method can be called multiple times to add multiple query vectors
* to the search. If multiple query vectors are added, then they will be searched
* in parallel, and the results will be concatenated. A column called `query_index`
* will be added to indicate the index of the query vector that produced the result.
*
* Performance wise, this is equivalent to running multiple queries concurrently.
*/
addQueryVector(vector: IntoVector): VectorQuery {
if (vector instanceof Promise) {
const res = (async () => {
try {
const v = await vector;
const arr = Float32Array.from(v);
//
// biome-ignore lint/suspicious/noExplicitAny: we need to get the `inner`, but js has no package scoping
const value: any = this.addQueryVector(arr);
const inner = value.inner as
| NativeVectorQuery
| Promise<NativeVectorQuery>;
return inner;
} catch (e) {
return Promise.reject(e);
}
})();
return new VectorQuery(res);
} else {
super.doCall((inner) => {
inner.addQueryVector(Float32Array.from(vector));
});
return this;
}
}
}
/** A builder for LanceDB queries. */