wehub-resource-sync
63 lines
2.0 KiB
Python
63 lines
2.0 KiB
Python
"""Tests for shared harness-bench session-item parsing."""
|
|
|
|
from tests.harness_bench.session_items import (
|
|
assistant_text,
|
|
contains_user_text,
|
|
function_calls,
|
|
item_role,
|
|
item_type,
|
|
tool_output_states,
|
|
)
|
|
|
|
|
|
def test_item_type_accepts_top_level_and_nested_shapes() -> None:
|
|
assert item_type({"type": "message"}) == "message"
|
|
assert item_type({"data": {"type": "function_call"}}) == "function_call"
|
|
assert item_role({"data": {"role": "assistant"}}) == "assistant"
|
|
|
|
|
|
def test_assistant_text_reads_both_item_shapes() -> None:
|
|
items = [
|
|
{"role": "assistant", "content": [{"type": "text", "text": "one"}]},
|
|
{
|
|
"data": {
|
|
"role": "assistant",
|
|
"content": [{"type": "output_text", "text": "two"}],
|
|
}
|
|
},
|
|
]
|
|
assert assistant_text(items) == "one\ntwo"
|
|
assert assistant_text(items, separator=" ") == "one two"
|
|
|
|
|
|
def test_assistant_text_accepts_untyped_text_and_ignores_non_text_blocks() -> None:
|
|
item = {
|
|
"role": "assistant",
|
|
"content": [
|
|
{"text": "plain"},
|
|
{"type": "reasoning", "text": "internal"},
|
|
{"type": "image", "text": "metadata"},
|
|
],
|
|
}
|
|
assert assistant_text(item) == "plain"
|
|
|
|
|
|
def test_function_calls_and_output_states_are_normalized() -> None:
|
|
items = [
|
|
{"data": {"type": "function_call", "call_id": "c1", "name": "Bash"}},
|
|
{"type": "function_call_output", "data": {"output": "ok"}},
|
|
{"type": "function_call_output", "data": {"status": "blocked", "output": "no"}},
|
|
]
|
|
assert function_calls(items) == [{"call_id": "c1", "name": "Bash", "arguments": None}]
|
|
assert tool_output_states(items, deny_marker="bench-deny") == (True, True)
|
|
|
|
|
|
def test_contains_user_text_reads_message_blocks() -> None:
|
|
items = [
|
|
{
|
|
"type": "message",
|
|
"data": {"role": "user", "content": [{"type": "text", "text": "interrupted"}]},
|
|
}
|
|
]
|
|
assert contains_user_text(items, "interrupt")
|