headroom/tests/test_output_turn_policy.py
JD Davis c904a70d4e
refactor(proxy): isolate output turn policy (#1962)
## Description

Extracts output-shaper turn classification into a pure
`output_turn_policy` module. The shaper still owns request mutation and
labels, while Anthropic-style and OpenAI Responses structural turn
classification now live in a deterministic policy boundary.

Closes #

## Type of Change

- [ ] Bug fix (non-breaking change that fixes an issue)
- [ ] New feature (non-breaking change that adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to change)
- [ ] Documentation update
- [ ] Performance improvement
- [x] Code refactoring (no functional changes)

## Changes Made

- Added `headroom.proxy.output_turn_policy` with `TurnKind`,
`classify_turn`, and `classify_openai_responses_input`.
- Updated `output_shaper` to import and re-export the classifiers,
preserving existing import behavior.
- Added direct policy tests for Anthropic tool-result turns and OpenAI
Responses input classification.
- Included the current LiteLLM callback signature compatibility shim
required for repo-wide mypy on main-based slices.

## Testing

- [x] Unit tests pass (`pytest`)
- [x] Linting passes (`ruff check .`)
- [x] Type checking passes (`mypy headroom`)
- [x] New tests added for new functionality
- [ ] Manual testing performed

### Test Output

```text
python -m pytest tests/test_output_turn_policy.py tests/test_output_shaper.py tests/test_litellm_callback.py -q
60 passed in 6.25s

python -m ruff check .
All checks passed!

python -m ruff format --check .
1095 files already formatted

python -m mypy headroom --ignore-missing-imports
Success: no issues found in 409 source files

gitleaks protect --staged --no-banner --redact
no leaks found
```

## Real Behavior Proof

- Environment: Windows, Python 3.13.13, clean worktree based on
`headroomlabs/main`.
- Exact command / steps: targeted pytest, ruff, format check, repo-wide
mypy, staged gitleaks scan.
- Observed result: output turn policy/shaper/callback tests pass; static
checks pass; no staged secrets detected.
- Not tested: live provider calls; this slice only moves structural
classification logic and preserves existing shaper behavior.

## Review Readiness

- [x] I have performed a self-review
- [x] This PR is ready for human review

## Checklist

- [x] My code follows the project's style guidelines
- [x] I have performed a self-review of my code
- [x] I have commented my code, particularly in hard-to-understand areas
- [ ] I have made corresponding changes to the documentation
- [x] My changes generate no new warnings
- [x] I have added tests that prove my fix is effective or that my
feature works
- [x] New and existing unit tests pass locally with my changes
- [ ] I have updated the CHANGELOG.md if applicable

## Screenshots (if applicable)

N/A

## Additional Notes

Documentation and changelog updates are not applicable for this internal
architecture slice. PR-specific GHAS checks will be monitored after
opening.
2026-07-11 21:33:48 -05:00

102 lines
3.3 KiB
Python

"""Tests for pure output turn classification policy."""
from __future__ import annotations
from typing import Any
from headroom.proxy.output_turn_policy import (
TurnKind,
classify_openai_responses_input,
classify_turn,
)
def _tool_result(is_error: bool = False) -> dict[str, Any]:
block: dict[str, Any] = {"type": "tool_result", "content": "ok"}
if is_error:
block["is_error"] = True
return block
def test_anthropic_text_user_message_is_new_ask() -> None:
assert classify_turn([{"role": "user", "content": "explain this"}]) is TurnKind.NEW_USER_ASK
def test_anthropic_clean_tool_results_are_mechanical() -> None:
messages = [{"role": "user", "content": [_tool_result(), _tool_result()]}]
assert classify_turn(messages) is TurnKind.MECHANICAL_CONTINUATION
def test_anthropic_error_tool_result_is_error_continuation() -> None:
messages = [{"role": "user", "content": [_tool_result(), _tool_result(is_error=True)]}]
assert classify_turn(messages) is TurnKind.ERROR_CONTINUATION
def test_anthropic_user_media_or_text_block_is_new_ask() -> None:
assert (
classify_turn([{"role": "user", "content": [{"type": "image", "source": {}}]}])
is TurnKind.NEW_USER_ASK
)
assert (
classify_turn(
[{"role": "user", "content": [_tool_result(), {"type": "text", "text": "also"}]}]
)
is TurnKind.NEW_USER_ASK
)
def test_anthropic_unknown_shapes_are_unknown() -> None:
assert classify_turn([]) is TurnKind.UNKNOWN
assert classify_turn([{"role": "assistant", "content": "done"}]) is TurnKind.UNKNOWN
assert classify_turn([{"role": "user", "content": []}]) is TurnKind.UNKNOWN
assert classify_turn([{"role": "user", "content": [{}]}]) is TurnKind.UNKNOWN
def test_openai_responses_string_input_is_new_ask() -> None:
assert classify_openai_responses_input("explain this") is TurnKind.NEW_USER_ASK
assert classify_openai_responses_input(" ") is TurnKind.UNKNOWN
def test_openai_responses_tool_outputs_only_are_mechanical() -> None:
assert (
classify_openai_responses_input(
[
{"type": "function_call_output", "call_id": "call_1", "output": "ok"},
{"type": "local_shell_call_output", "call_id": "call_2", "output": "ok"},
]
)
is TurnKind.MECHANICAL_CONTINUATION
)
def test_openai_responses_user_message_or_input_media_is_new_ask() -> None:
assert (
classify_openai_responses_input(
[
{
"type": "message",
"role": "user",
"content": [{"type": "input_text", "text": "also check foo.py"}],
}
]
)
is TurnKind.NEW_USER_ASK
)
assert (
classify_openai_responses_input(
[{"type": "message", "role": "user", "content": [{"type": "input_image"}]}]
)
is TurnKind.NEW_USER_ASK
)
def test_openai_responses_unknown_mixed_with_tool_output_is_unknown() -> None:
assert (
classify_openai_responses_input(
[
{"type": "function_call_output", "call_id": "call_1", "output": "ok"},
{"type": "unrecognized_event"},
]
)
is TurnKind.UNKNOWN
)