Updating Vectors

1. Updating Vector Values

index.update(
    id="doc1",
    values=[0.11, 0.22, ...],  # new dense vector
)

Note: An upsert with same ID also replaces values; update() only changes the specified fields.

index.update(
    id="doc1",
    set_metadata={"status": "archived", "version": 3},
)

Param	Behavior
`set_metadata`	Merges with existing metadata (partial update)
Upsert with metadata	Replaces entire metadata object

index.update(
    id="doc1",
    sparse_values={"indices": [1, 7, 42], "values": [0.5, 0.3, 0.2]},
)

index.update(id="u1", set_metadata={"plan": "pro"}, namespace="users")

Behavior	Detail
Wrong namespace	No-op (no error)
Default	`""`

Operation	Result
`set_metadata={"k": v}`	Adds or overwrites key `k`
Other keys	Preserved unchanged
Delete a key	Not supported; re-upsert vector without it

index.upsert(vectors=[{
    "id": "doc1",
    "values": existing_values,
    "metadata": {"status": "final"},  # replaces all metadata
}])

updates = [
    {"id": v["id"], "values": v["values"], "metadata": v["metadata"]}
    for v in updated_vectors
]
index.upsert(vectors=updates)

Note: update() only handles one ID at a time. For bulk changes, prefer batched upsert.

index.update(id="d1", set_metadata={"v": 2})
got = index.fetch(ids=["d1"])
assert got["vectors"]["d1"]["metadata"]["v"] == 2

Scenario	Strategy
Concurrent updates	Last write wins; use version field
Race with delete	Idempotent: update on missing = no-op
Optimistic concurrency	Check `version` metadata before write