Files
Timmy-time-dashboard/tests/scripts/test_export_trajectories.py
Google Gemini 05e1196ea4
Some checks failed
Tests / lint (push) Has been cancelled
Tests / test (push) Has been cancelled
[gemini] feat: add coverage and duration strictness to pytest (#934) (#1140)
Co-authored-by: Google Gemini <gemini@hermes.local>
Co-committed-by: Google Gemini <gemini@hermes.local>
2026-03-23 18:36:01 +00:00

320 lines
11 KiB
Python

"""Unit tests for scripts/export_trajectories.py.
Tests trajectory conversion logic — no I/O, no Ollama, no mlx.
"""
from __future__ import annotations
import json
from pathlib import Path
import pytest
import scripts.export_trajectories as et
# ── Fixtures ──────────────────────────────────────────────────────────────────
@pytest.fixture()
def simple_session(tmp_path: Path) -> Path:
"""Write a minimal session JSONL file and return the logs dir."""
logs_dir = tmp_path / "logs"
logs_dir.mkdir()
entries = [
{
"type": "message",
"role": "user",
"content": "What time is it?",
"timestamp": "2026-03-01T10:00:00",
},
{
"type": "message",
"role": "timmy",
"content": "It is 10:00 AM.",
"timestamp": "2026-03-01T10:00:01",
},
{
"type": "message",
"role": "user",
"content": "Thanks!",
"timestamp": "2026-03-01T10:00:05",
},
{
"type": "message",
"role": "timmy",
"content": "You're welcome!",
"timestamp": "2026-03-01T10:00:06",
},
]
session_file = logs_dir / "session_2026-03-01.jsonl"
session_file.write_text("\n".join(json.dumps(e) for e in entries) + "\n")
return logs_dir
@pytest.fixture()
def tool_call_session(tmp_path: Path) -> Path:
"""Write a session JSONL with tool calls."""
logs_dir = tmp_path / "logs"
logs_dir.mkdir()
entries = [
{
"type": "message",
"role": "user",
"content": "Read CLAUDE.md",
"timestamp": "2026-03-01T10:00:00",
},
{
"type": "tool_call",
"tool": "read_file",
"args": {"path": "CLAUDE.md"},
"result": "# CLAUDE.md content here",
"timestamp": "2026-03-01T10:00:01",
},
{
"type": "message",
"role": "timmy",
"content": "Here is the content.",
"timestamp": "2026-03-01T10:00:02",
},
]
session_file = logs_dir / "session_2026-03-01.jsonl"
session_file.write_text("\n".join(json.dumps(e) for e in entries) + "\n")
return logs_dir
# ── _load_entries ─────────────────────────────────────────────────────────────
@pytest.mark.unit
def test_load_entries_returns_all(simple_session: Path) -> None:
entries = et._load_entries(simple_session)
assert len(entries) == 4
@pytest.mark.unit
def test_load_entries_skips_malformed(tmp_path: Path) -> None:
logs_dir = tmp_path / "logs"
logs_dir.mkdir()
session = logs_dir / "session_2026-03-01.jsonl"
session.write_text(
'{"type": "message", "role": "user", "content": "hi"}\n'
"NOT_JSON\n"
'{"type": "message", "role": "timmy", "content": "hello"}\n'
)
entries = et._load_entries(logs_dir)
assert len(entries) == 2 # malformed line skipped
@pytest.mark.unit
def test_load_entries_empty_dir(tmp_path: Path) -> None:
logs_dir = tmp_path / "logs"
logs_dir.mkdir()
entries = et._load_entries(logs_dir)
assert entries == []
@pytest.mark.unit
def test_load_entries_multiple_files(tmp_path: Path) -> None:
logs_dir = tmp_path / "logs"
logs_dir.mkdir()
for day in ("2026-03-01", "2026-03-02"):
entry = {"type": "message", "role": "user", "content": f"day {day}"}
(logs_dir / f"session_{day}.jsonl").write_text(json.dumps(entry) + "\n")
entries = et._load_entries(logs_dir)
assert len(entries) == 2
# ── _format_tool_call ─────────────────────────────────────────────────────────
@pytest.mark.unit
def test_format_tool_call_structure() -> None:
entry = {
"type": "tool_call",
"tool": "read_file",
"args": {"path": "/tmp/foo.txt"},
"result": "file contents",
}
result = et._format_tool_call(entry)
assert result.startswith("<tool_call>")
assert result.endswith("</tool_call>")
payload = json.loads(result.split("\n")[1])
assert payload["name"] == "read_file"
assert payload["arguments"]["path"] == "/tmp/foo.txt"
@pytest.mark.unit
def test_format_tool_call_missing_tool() -> None:
entry = {"type": "tool_call", "args": {}}
result = et._format_tool_call(entry)
assert "unknown" in result
# ── _group_into_turns ─────────────────────────────────────────────────────────
@pytest.mark.unit
def test_group_basic_conversation() -> None:
entries = [
{"type": "message", "role": "user", "content": "hello"},
{"type": "message", "role": "timmy", "content": "hi there"},
{"type": "message", "role": "user", "content": "bye"},
{"type": "message", "role": "timmy", "content": "goodbye"},
]
turns = et._group_into_turns(entries)
assert len(turns) == 2
assert turns[0]["user"] == "hello"
assert turns[0]["assistant"] == "hi there"
assert turns[1]["user"] == "bye"
assert turns[1]["assistant"] == "goodbye"
@pytest.mark.unit
def test_group_with_tool_call() -> None:
entries = [
{"type": "message", "role": "user", "content": "check the file"},
{"type": "tool_call", "tool": "read_file", "args": {"path": "x"}, "result": "content"},
{"type": "message", "role": "timmy", "content": "Done."},
]
turns = et._group_into_turns(entries)
assert len(turns) == 1
assert "<tool_call>" in turns[0]["assistant"]
assert "Done." in turns[0]["assistant"]
@pytest.mark.unit
def test_group_skips_user_without_response() -> None:
"""User message with no timmy response should not create a turn."""
entries = [
{"type": "message", "role": "user", "content": "hello"},
# No timmy response
{"type": "message", "role": "user", "content": "are you there?"},
{"type": "message", "role": "timmy", "content": "Yes!"},
]
turns = et._group_into_turns(entries)
assert len(turns) == 1
assert turns[0]["user"] == "are you there?"
@pytest.mark.unit
def test_group_ignores_errors_and_decisions() -> None:
entries = [
{"type": "message", "role": "user", "content": "hello"},
{"type": "error", "error": "something failed"},
{"type": "decision", "decision": "retry"},
{"type": "message", "role": "timmy", "content": "Got it."},
]
turns = et._group_into_turns(entries)
assert len(turns) == 1
assert "error" not in turns[0]["assistant"]
assert "retry" not in turns[0]["assistant"]
@pytest.mark.unit
def test_group_empty_entries() -> None:
assert et._group_into_turns([]) == []
# ── turns_to_training_examples ────────────────────────────────────────────────
@pytest.mark.unit
def test_training_examples_structure() -> None:
turns = [{"user": "hello", "assistant": "hi there, how can I help?"}]
examples = et.turns_to_training_examples(turns)
assert len(examples) == 1
msgs = examples[0]["messages"]
assert msgs[0]["role"] == "system"
assert msgs[1]["role"] == "user"
assert msgs[1]["content"] == "hello"
assert msgs[2]["role"] == "assistant"
assert msgs[2]["content"] == "hi there, how can I help?"
@pytest.mark.unit
def test_training_examples_filters_short_responses() -> None:
turns = [
{"user": "hello", "assistant": "ok"}, # too short
{"user": "hello", "assistant": "This is a longer response that passes."},
]
examples = et.turns_to_training_examples(turns, min_assistant_len=10)
assert len(examples) == 1
assert examples[0]["messages"][2]["content"] == "This is a longer response that passes."
@pytest.mark.unit
def test_training_examples_filters_empty_user() -> None:
turns = [{"user": "", "assistant": "some response here"}]
examples = et.turns_to_training_examples(turns)
assert len(examples) == 0
@pytest.mark.unit
def test_training_examples_uses_custom_system_prompt() -> None:
turns = [{"user": "hi", "assistant": "hello there!"}]
examples = et.turns_to_training_examples(turns, system_prompt="Custom prompt.")
assert examples[0]["messages"][0]["content"] == "Custom prompt."
# ── export_training_data (integration-style, uses tmp_path) ──────────────────
@pytest.mark.unit
def test_export_training_data_writes_jsonl(simple_session: Path, tmp_path: Path) -> None:
output = tmp_path / "train.jsonl"
count = et.export_training_data(logs_dir=simple_session, output_path=output)
assert count == 2
assert output.exists()
lines = [json.loads(line) for line in output.read_text().splitlines() if line.strip()]
assert len(lines) == 2
for line in lines:
assert "messages" in line
roles = [m["role"] for m in line["messages"]]
assert roles == ["system", "user", "assistant"]
@pytest.mark.unit
def test_export_training_data_with_tool_calls(tool_call_session: Path, tmp_path: Path) -> None:
output = tmp_path / "train.jsonl"
count = et.export_training_data(logs_dir=tool_call_session, output_path=output)
assert count == 1
line = json.loads(output.read_text().strip())
assistant_content = line["messages"][2]["content"]
assert "<tool_call>" in assistant_content
assert "read_file" in assistant_content
@pytest.mark.unit
def test_export_training_data_returns_zero_for_empty_logs(tmp_path: Path) -> None:
logs_dir = tmp_path / "logs"
logs_dir.mkdir()
output = tmp_path / "train.jsonl"
count = et.export_training_data(logs_dir=logs_dir, output_path=output)
assert count == 0
assert not output.exists()
# ── CLI ───────────────────────────────────────────────────────────────────────
@pytest.mark.unit
def test_cli_missing_logs_dir(tmp_path: Path) -> None:
rc = et.main(
["--logs-dir", str(tmp_path / "nonexistent"), "--output", str(tmp_path / "out.jsonl")]
)
assert rc == 1
@pytest.mark.unit
def test_cli_exports_and_returns_zero(simple_session: Path, tmp_path: Path) -> None:
output = tmp_path / "out.jsonl"
rc = et.main(
[
"--logs-dir",
str(simple_session),
"--output",
str(output),
]
)
assert rc == 0
assert output.exists()