chore(terminal-bench): merge main, retarget graduated compaction import

Merge `origin/main` (brings in the experimental->top-level capability
graduation) and update the benchmark's compaction import from the now-deprecated
`pydantic_ai_harness.experimental.compaction` shim to the canonical
`pydantic_ai_harness.compaction`. Drops the now-obsolete experimental-warning
filter. Benchmark behavior is unchanged.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Douwe Maan
2026-07-15 19:48:48 +00:00
co-authored by Claude Opus 4.8
parent e1eca45f15
commit db8e739921
3 changed files with 2 additions and 6 deletions
-4
View File
@@ -57,7 +57,3 @@ include = ['src']
[tool.pytest.ini_options]
testpaths = ['tests']
anyio_mode = 'auto'
filterwarnings = [
# Harness compaction capabilities are experimental and warn on import.
'ignore::pydantic_ai_harness.experimental.HarnessExperimentalWarning',
]
@@ -26,7 +26,7 @@ from pydantic_ai import Agent
from pydantic_ai.capabilities import AbstractCapability
from pydantic_ai.models import Model
from pydantic_ai.settings import ModelSettings
from pydantic_ai_harness.experimental.compaction import (
from pydantic_ai_harness.compaction import (
ClearToolResults,
SummarizingCompaction,
TieredCompaction,
@@ -7,7 +7,7 @@ from pydantic_ai import Agent
from pydantic_ai.messages import ModelMessage, ModelResponse, TextPart, ToolCallPart, ToolReturnPart
from pydantic_ai.models.function import AgentInfo, FunctionModel
from pydantic_ai.models.test import TestModel
from pydantic_ai_harness.experimental.compaction import TieredCompaction
from pydantic_ai_harness.compaction import TieredCompaction
from pydantic_ai_harness_terminal_bench.agent import build_agent, build_compaction
from pydantic_ai_harness_terminal_bench.tools import CommandResult, TerminalBenchDeps