JiuwenSwarm is an intelligent AI Agent built on openJiuwen. It extends the powerful capabilities of large language models directly to your fingertips through various communication apps you use daily.
Python
# Copyright (c) Huawei Technologies Co., Ltd. 2026. All rights reserved.
"""裸 ``task_interaction`` controller_output → 前端提问卡 的解析测试。
背景:原生 harness 的 ask_user 中断以裸 task_interaction controller_output
到达解析层,stream_utils 此前仅识别 ``__interaction__`` 载荷,该 chunk 被静默
丢弃——前端既收不到提问卡也收不到 chat.final。现需识别嵌入的 ask_user
中断值并转换为 ``chat.ask_user_question``,无法解析时下发兜底输入卡。
"""
from __future__ import annotations
from types import SimpleNamespace
from jiuwenswarm.server.runtime.agent_adapter.interface_deep import (
JiuWenSwarmDeepAdapter,
)
from jiuwenswarm.server.utils import stream_utils
from jiuwenswarm.server.utils.stream_utils import parse_stream_chunk
from jiuwenswarm.server.runtime.agent_adapter.deep.stream_parse import _parse_stream_chunk
def _ask_user_value(**overrides) -> dict:
value = {
"tool_name": "ask_user",
"tool_call_id": "call_ask_user_1",
"questions": [
{
"question": "文件存放到哪个目录?",
"header": "Location",
"options": [],
"multi_select": False,
}
],
}
value.update(overrides)
return value
def _task_interaction_chunk(payload: dict) -> SimpleNamespace:
return SimpleNamespace(type="controller_output", payload=payload)
def test_parse_task_interaction_ask_user_card_from_bare_controller_output() -> None:
"""裸 task_interaction(无 __interaction__ 标记)必须转出提问卡,不再被丢弃。"""
chunk = _task_interaction_chunk(
{"type": "task_interaction", "data": [_ask_user_value()]}
)
parsed = parse_stream_chunk(chunk)
assert parsed is not None
assert parsed["event_type"] == "chat.ask_user_question"
assert parsed["request_id"] == "call_ask_user_1"
assert parsed["source"] == "ask_user_interrupt"
assert parsed["questions"][0]["question"] == "文件存放到哪个目录?"
def test_parse_task_interaction_ask_user_card_appends_other_option() -> None:
"""带选项的问题转换为卡片时追加 __other__ 自定义输入项。"""
value = _ask_user_value(
questions=[
{
"question": "选择目录",
"options": [
{"label": "当前项目目录", "description": ""},
{"label": "桌面", "description": ""},
],
}
]
)
chunk = _task_interaction_chunk({"type": "task_interaction", "data": [value]})
parsed = parse_stream_chunk(chunk)
assert parsed is not None
options = parsed["questions"][0]["options"]
assert len(options) == 3
assert options[-1]["label"] == "Other"
def test_parse_task_interaction_from_pydantic_model_dump() -> None:
"""ToolCallInterruptRequest 等模型对象经 model_dump 展开后仍可识别。"""
class _FakeInterruptRequest:
def __init__(self, dump: dict) -> None:
self._dump = dump
def model_dump(self, mode: str = "python") -> dict:
return dict(self._dump)
chunk = _task_interaction_chunk(
{
"type": "task_interaction",
"result": _FakeInterruptRequest(
_ask_user_value(tool_call_id="call_dump_1")
),
}
)
parsed = parse_stream_chunk(chunk)
assert parsed is not None
assert parsed["event_type"] == "chat.ask_user_question"
assert parsed["request_id"] == "call_dump_1"
def test_parse_task_interaction_questions_in_tool_args_json_string() -> None:
"""questions 藏在 tool_args JSON 字符串(StructuredAskUserRail 形态)也可解析。"""
value = {
"tool_name": "ask_user",
"tool_call_id": "call_json",
"tool_args": '{"questions": [{"question": "目录?"}]}',
}
chunk = _task_interaction_chunk({"type": "task_interaction", "data": [value]})
parsed = parse_stream_chunk(chunk)
assert parsed is not None
assert parsed["request_id"] == "call_json"
assert parsed["questions"][0]["question"] == "目录?"
def test_parse_task_interaction_prefers_structured_questions_over_plain() -> None:
"""结构化 questions 候选优先于 plain query,即使 plain 排在前面。"""
plain = {
"tool_name": "ask_user",
"tool_call_id": "call_plain",
"message": "直接回答问题",
}
structured = _ask_user_value(tool_call_id="call_struct")
chunk = _task_interaction_chunk(
{"type": "task_interaction", "data": [plain, structured]}
)
parsed = parse_stream_chunk(chunk)
assert parsed is not None
assert parsed["request_id"] == "call_struct"
assert parsed["questions"][0]["question"] == "文件存放到哪个目录?"
def test_parse_task_interaction_request_id_prefers_value_over_candidates() -> None:
"""request_id 取中断值自身的 tool_call_id,而非外层收集到的候选。"""
payload = {
"type": "task_interaction",
"tool_call_id": "outer_candidate",
"data": [_ask_user_value(tool_call_id="inner_call")],
}
parsed = stream_utils.parse_task_interaction_payload(payload)
assert parsed is not None
assert parsed["request_id"] == "inner_call"
def test_parse_task_interaction_request_id_falls_back_to_collected_candidates() -> None:
"""中断值无 id 时回退到外层收集的候选(如 data 兄弟节点的 tool_call_id)。"""
value = {
k: v for k, v in _ask_user_value().items() if k != "tool_call_id"
}
payload = {
"type": "task_interaction",
"data": [value, {"tool_call_id": "sibling_call"}],
}
parsed = stream_utils.parse_task_interaction_payload(payload)
assert parsed is not None
assert parsed["request_id"] == "sibling_call"
def test_parse_task_interaction_without_ask_user_emits_fallback_card() -> None:
"""无法解析出 ask_user 值时必须下发兜底输入卡,前端不能没有输入入口。"""
chunk = _task_interaction_chunk(
{"type": "task_interaction", "data": [{"tool_call_id": "call_x", "status": "pending"}]}
)
parsed = parse_stream_chunk(chunk)
assert parsed is not None
assert parsed["event_type"] == "chat.ask_user_question"
assert parsed["request_id"] == "call_x"
assert parsed["source"] == "task_interaction_fallback"
assert parsed["questions"][0]["question"]
assert parsed["questions"][0]["options"] == []
assert parsed["questions"][0]["multi_select"] is False
def test_fallback_card_generates_request_id_when_missing() -> None:
card = stream_utils._fallback_awaiting_input_card("")
assert card["request_id"].startswith("task_interaction_")
card = stream_utils._fallback_awaiting_input_card("keep_me")
assert card["request_id"] == "keep_me"
def test_parse_task_interaction_falls_through_unconverted_interaction_payloads() -> None:
"""__interaction__ 载荷存在但不可转换时,不得静默丢弃——落到
task_interaction 分支继续解析嵌入的 ask_user 值。"""
chunk = _task_interaction_chunk(
{
"type": "task_interaction",
"data": [
{"type": "__interaction__", "payload": _ask_user_value()}
],
}
)
parsed = parse_stream_chunk(chunk)
assert parsed is not None
assert parsed["event_type"] == "chat.ask_user_question"
assert parsed["request_id"] == "call_ask_user_1"
def test_collect_request_id_candidates_gathers_ids_from_shapes() -> None:
payload = {
"tool_call_id": "outer_call",
"interrupt_ids": ["int_1", "int_2"],
"data": [
{"request_id": "inner_req"},
{"interrupt_id": "inner_int"},
],
}
ids = stream_utils._collect_request_id_candidates(payload)
assert ids == ["outer_call", "int_1", "int_2", "inner_req", "inner_int"]
def test_iter_ask_user_interrupt_values_terminates_on_cycles() -> None:
cyclic: list = []
cyclic.append(cyclic)
container = {"tool_name": "ask_user", "questions": [], "nested": cyclic}
values = list(stream_utils._iter_ask_user_interrupt_values(container))
assert values == [container]
def test_interface_deep_parse_stream_chunk_builds_task_interaction_card() -> None:
"""interface_deep 镜像解析:task_interaction 同样转出提问卡。"""
chunk = _task_interaction_chunk(
{"type": "task_interaction", "data": [_ask_user_value()]}
)
parsed = _parse_stream_chunk(chunk)
assert parsed is not None
assert parsed["event_type"] == "chat.ask_user_question"
assert parsed["request_id"] == "call_ask_user_1"
def test_interface_deep_parse_stream_chunk_unparsable_task_interaction_drops() -> None:
"""interface_deep 镜像解析:不可解析的 task_interaction 归为控制面元数据
(返回 None),兜底卡由外层 stream_utils 负责。"""
chunk = _task_interaction_chunk(
{"type": "task_interaction", "data": [{"status": "pending"}]}
)
assert _parse_stream_chunk(chunk) is None