mirror of
https://github.com/NousResearch/hermes-agent.git
synced 2026-07-30 19:09:28 +00:00
6b81590c55
Systematic prune per AGENTS.md test policy, one pass over every major test tree (gateway, hermes_cli, tools, agent, run_agent, plugins, cli, cron, tui_gateway, honcho/openviking, root-level): - DELETE: source-reading tests (read_text/getsource on prod files), change-detector tests (exact catalog counts, model-name snapshots, config version literals), mock-echo tests (assert a mock returns what it was told), assertion-free/trivial tests, near-duplicate parametrizations (boundaries + one representative kept), async/sync twin duplicates, cosmetic within-file variations. - KEEP (mandatory): security/redaction/approval guards, message-role alternation invariants, prompt-caching/deterministic-call-id invariants, issue-number regression tests (deduped), E2E tests. - 6 test files deleted outright (script-style/no-assert or fully redundant); conftest.py, fakes/, fixtures/ untouched. - tests/acp/conftest.py added: autouse fixture stubs the live models.dev/GitHub/Copilot/Anthropic inventory fetches that ACP server tests performed on every session create — test_server.py 147s → 3.4s, and the tests are now genuinely hermetic. - Sleep-based slowness shrunk where safe (codex_ttfb_watchdog, compression_concurrent_fork, etc.); no wall-clock assertion tightened. Verification: full hermetic suite via scripts/run_tests.sh — 2439 files, 31,130 tests passed, 0 failed, 0 flaky retries, 315s wall (baseline: 583s wall, 13,564s subprocess CPU).
321 lines
11 KiB
Python
321 lines
11 KiB
Python
import time
|
|
from datetime import datetime, timedelta
|
|
from types import SimpleNamespace
|
|
from unittest.mock import MagicMock, patch
|
|
|
|
import cli as cli_mod
|
|
from cli import HermesCLI
|
|
|
|
|
|
def _make_cli(model: str = "anthropic/claude-sonnet-4-20250514"):
|
|
cli_obj = HermesCLI.__new__(HermesCLI)
|
|
cli_obj.model = model
|
|
cli_obj.session_start = datetime.now() - timedelta(minutes=14, seconds=32)
|
|
cli_obj.conversation_history = [{"role": "user", "content": "hi"}]
|
|
cli_obj.agent = None
|
|
return cli_obj
|
|
|
|
|
|
def _attach_agent(
|
|
cli_obj,
|
|
*,
|
|
input_tokens: int | None = None,
|
|
output_tokens: int | None = None,
|
|
cache_read_tokens: int = 0,
|
|
cache_write_tokens: int = 0,
|
|
prompt_tokens: int,
|
|
completion_tokens: int,
|
|
total_tokens: int,
|
|
api_calls: int,
|
|
context_tokens: int,
|
|
context_length: int,
|
|
compressions: int = 0,
|
|
):
|
|
cli_obj.agent = SimpleNamespace(
|
|
model=cli_obj.model,
|
|
provider="anthropic" if cli_obj.model.startswith("anthropic/") else None,
|
|
base_url="",
|
|
session_input_tokens=input_tokens if input_tokens is not None else prompt_tokens,
|
|
session_output_tokens=output_tokens if output_tokens is not None else completion_tokens,
|
|
session_cache_read_tokens=cache_read_tokens,
|
|
session_cache_write_tokens=cache_write_tokens,
|
|
session_prompt_tokens=prompt_tokens,
|
|
session_completion_tokens=completion_tokens,
|
|
session_total_tokens=total_tokens,
|
|
session_api_calls=api_calls,
|
|
get_rate_limit_state=lambda: None,
|
|
context_compressor=SimpleNamespace(
|
|
last_prompt_tokens=context_tokens,
|
|
context_length=context_length,
|
|
compression_count=compressions,
|
|
),
|
|
)
|
|
return cli_obj
|
|
|
|
|
|
class TestCLIStatusBar:
|
|
def test_context_style_thresholds(self):
|
|
cli_obj = _make_cli()
|
|
|
|
assert cli_obj._status_bar_context_style(None) == "class:status-bar-dim"
|
|
assert cli_obj._status_bar_context_style(10) == "class:status-bar-good"
|
|
assert cli_obj._status_bar_context_style(50) == "class:status-bar-warn"
|
|
assert cli_obj._status_bar_context_style(81) == "class:status-bar-bad"
|
|
assert cli_obj._status_bar_context_style(95) == "class:status-bar-critical"
|
|
|
|
def test_build_status_bar_text_for_wide_terminal(self):
|
|
cli_obj = _attach_agent(
|
|
_make_cli(),
|
|
prompt_tokens=10_230,
|
|
completion_tokens=2_220,
|
|
total_tokens=12_450,
|
|
api_calls=7,
|
|
context_tokens=12_450,
|
|
context_length=200_000,
|
|
)
|
|
|
|
text = cli_obj._build_status_bar_text(width=120)
|
|
|
|
assert "claude-sonnet-4-20250514" in text
|
|
assert "12.4K/200K" in text
|
|
assert "6%" in text
|
|
assert "$0.06" not in text # cost hidden by default
|
|
assert "15m" in text
|
|
|
|
|
|
def test_input_height_counts_prompt_only_on_first_wrapped_row(self):
|
|
# Regression for prompt_toolkit classic CLI resize glitches: the prompt
|
|
# is inserted by BeforeInput only on logical line 0. At three terminal
|
|
# cells, "⚔ " leaves one cell for the first input character, but
|
|
# wrapped continuation rows use the full three cells. Estimating every
|
|
# wrapped row as one-cell wide over-allocates the TextArea and can leave
|
|
# stale prompt/input cells visible after resize.
|
|
assert cli_mod._estimate_tui_input_height(["abcdef"], "⚔ ", 3) == 3
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def test_compression_count_shown_in_wide_status_bar(self):
|
|
cli_obj = _attach_agent(
|
|
_make_cli(),
|
|
prompt_tokens=10_230,
|
|
completion_tokens=2_220,
|
|
total_tokens=12_450,
|
|
api_calls=7,
|
|
context_tokens=12_450,
|
|
context_length=200_000,
|
|
compressions=3,
|
|
)
|
|
|
|
text = cli_obj._build_status_bar_text(width=120)
|
|
|
|
assert "🗜️ 3" in text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def test_minimal_tui_chrome_threshold(self):
|
|
cli_obj = _make_cli()
|
|
|
|
assert cli_obj._use_minimal_tui_chrome(width=63) is True
|
|
assert cli_obj._use_minimal_tui_chrome(width=64) is False
|
|
|
|
|
|
|
|
|
|
def test_scheduled_unsuppress_debounces_resize_storm(self):
|
|
"""A fresh resize cancels the pending unsuppress and restarts it."""
|
|
cli_obj = _make_cli()
|
|
cli_obj._status_bar_unsuppress_timer = None
|
|
cli_obj._status_bar_suppressed_after_resize = True
|
|
app = MagicMock()
|
|
app.loop = None
|
|
|
|
# First schedule (long delay) then a second should cancel the first.
|
|
cli_obj._schedule_status_bar_unsuppress(app, delay=5.0)
|
|
first_timer = cli_obj._status_bar_unsuppress_timer
|
|
assert first_timer is not None
|
|
cli_obj._schedule_status_bar_unsuppress(app, delay=0.01)
|
|
assert first_timer is not cli_obj._status_bar_unsuppress_timer
|
|
assert not first_timer.is_alive() or first_timer.finished.is_set()
|
|
time.sleep(0.1)
|
|
assert cli_obj._status_bar_suppressed_after_resize is False
|
|
|
|
|
|
|
|
|
|
def test_spinner_height_uses_display_width_for_wide_characters(self):
|
|
cli_obj = _make_cli()
|
|
cli_obj._spinner_text = "你" * 40
|
|
cli_obj._tool_start_time = 0
|
|
|
|
assert cli_obj._spinner_widget_height(width=64) == 2
|
|
|
|
|
|
def test_voice_status_bar_compacts_on_narrow_terminals(self):
|
|
cli_obj = _make_cli()
|
|
cli_obj._voice_mode = True
|
|
cli_obj._voice_recording = False
|
|
cli_obj._voice_processing = False
|
|
cli_obj._voice_tts = True
|
|
cli_obj._voice_continuous = True
|
|
|
|
fragments = cli_obj._get_voice_status_fragments(width=50)
|
|
|
|
assert fragments == [("class:voice-status", " 🎤 Ctrl+B ")]
|
|
|
|
|
|
# Round-13 Copilot review regressions on #19835. The label in voice
|
|
# status bar / recording hint / placeholder must render the
|
|
# configured ``voice.record_key`` — not hardcoded Ctrl+B. Pinning
|
|
# the cache (``set_voice_record_key_cache``) keeps display in sync
|
|
# with the prompt_toolkit binding without re-reading config on
|
|
# every render.
|
|
def test_voice_status_bar_renders_configured_ctrl_letter(self):
|
|
cli_obj = _make_cli()
|
|
cli_obj._voice_mode = True
|
|
cli_obj._voice_recording = False
|
|
cli_obj._voice_processing = False
|
|
cli_obj._voice_tts = False
|
|
cli_obj._voice_continuous = False
|
|
cli_obj.set_voice_record_key_cache("ctrl+o")
|
|
|
|
wide = cli_obj._get_voice_status_fragments(width=120)
|
|
assert any("Ctrl+O to record" in text for _cls, text in wide)
|
|
|
|
compact = cli_obj._get_voice_status_fragments(width=50)
|
|
assert compact == [("class:voice-status", " 🎤 Ctrl+O ")]
|
|
|
|
|
|
|
|
|
|
|
|
class TestCLIUsageReport:
|
|
def test_show_usage_omits_cost_reporting(self, capsys):
|
|
cli_obj = _attach_agent(
|
|
_make_cli(),
|
|
prompt_tokens=10_230,
|
|
completion_tokens=2_220,
|
|
total_tokens=12_450,
|
|
api_calls=7,
|
|
context_tokens=12_450,
|
|
context_length=200_000,
|
|
compressions=1,
|
|
)
|
|
cli_obj.verbose = False
|
|
|
|
cli_obj._show_usage()
|
|
output = capsys.readouterr().out
|
|
|
|
# Token counts and session metadata still shown.
|
|
assert "Model:" in output
|
|
assert "Input tokens:" in output
|
|
assert "Output tokens:" in output
|
|
assert "Total tokens:" in output
|
|
assert "Session duration:" in output
|
|
assert "Compressions:" in output
|
|
# Cost and cache-hit reporting is removed everywhere.
|
|
assert "Total cost:" not in output
|
|
assert "Cost status:" not in output
|
|
assert "Cost source:" not in output
|
|
assert "Cache read tokens:" not in output
|
|
assert "Cache write tokens:" not in output
|
|
|
|
|
|
class TestStatusBarWidthSource:
|
|
"""Ensure status bar fragments don't overflow the terminal width."""
|
|
|
|
def _make_wide_cli(self):
|
|
cli_obj = _attach_agent(
|
|
_make_cli(),
|
|
prompt_tokens=100_000,
|
|
completion_tokens=5_000,
|
|
total_tokens=105_000,
|
|
api_calls=20,
|
|
context_tokens=100_000,
|
|
context_length=200_000,
|
|
)
|
|
cli_obj._status_bar_visible = True
|
|
return cli_obj
|
|
|
|
def test_fragments_fit_within_announced_width(self):
|
|
"""Total fragment text length must not exceed the width used to build them."""
|
|
from unittest.mock import MagicMock, patch
|
|
cli_obj = self._make_wide_cli()
|
|
|
|
for width in (40, 52, 76, 80, 120, 200):
|
|
mock_app = MagicMock()
|
|
mock_app.output.get_size.return_value = MagicMock(columns=width)
|
|
|
|
with patch("prompt_toolkit.application.get_app", return_value=mock_app):
|
|
frags = cli_obj._get_status_bar_fragments()
|
|
|
|
total_text = "".join(text for _, text in frags)
|
|
display_width = cli_obj._status_bar_display_width(total_text)
|
|
assert display_width <= width + 4, ( # +4 for minor padding chars
|
|
f"At width={width}, fragment total {display_width} cells overflows "
|
|
f"({total_text!r})"
|
|
)
|
|
|
|
def test_fragments_use_pt_width_over_shutil(self):
|
|
"""When prompt_toolkit reports a width, shutil.get_terminal_size must not be used."""
|
|
from unittest.mock import MagicMock, patch
|
|
cli_obj = self._make_wide_cli()
|
|
|
|
mock_app = MagicMock()
|
|
mock_app.output.get_size.return_value = MagicMock(columns=120)
|
|
|
|
with patch("prompt_toolkit.application.get_app", return_value=mock_app) as mock_get_app, \
|
|
patch("shutil.get_terminal_size") as mock_shutil:
|
|
cli_obj._get_status_bar_fragments()
|
|
|
|
mock_shutil.assert_not_called()
|
|
|
|
|
|
def test_build_status_bar_text_uses_pt_width(self):
|
|
"""_build_status_bar_text() must also prefer prompt_toolkit width."""
|
|
from unittest.mock import MagicMock, patch
|
|
cli_obj = self._make_wide_cli()
|
|
|
|
mock_app = MagicMock()
|
|
mock_app.output.get_size.return_value = MagicMock(columns=80)
|
|
|
|
with patch("prompt_toolkit.application.get_app", return_value=mock_app), \
|
|
patch("shutil.get_terminal_size") as mock_shutil:
|
|
text = cli_obj._build_status_bar_text() # no explicit width
|
|
|
|
mock_shutil.assert_not_called()
|
|
assert isinstance(text, str)
|
|
assert len(text) > 0
|
|
|
|
|
|
|
|
class TestIdleSinceLastTurn:
|
|
"""Time-since-last-final-agent-response read-out on the status bar."""
|
|
|
|
def test_hidden_before_first_turn(self):
|
|
assert HermesCLI._format_idle_since(None, turn_live=False) == ""
|
|
|
|
def test_hidden_while_turn_is_live(self):
|
|
assert HermesCLI._format_idle_since(time.time() - 30, turn_live=True) == ""
|
|
|
|
def test_shows_compact_idle_time_after_turn(self):
|
|
label = HermesCLI._format_idle_since(time.time() - 42, turn_live=False)
|
|
assert label.startswith("✓ ")
|
|
assert label == "✓ 42s"
|
|
|
|
|
|
def test_snapshot_carries_idle_since(self):
|
|
cli_obj = _make_cli()
|
|
cli_obj._last_turn_finished_at = time.time() - 10
|
|
cli_obj._prompt_start_time = None
|
|
cli_obj._prompt_duration = 5.0
|
|
snapshot = cli_obj._get_status_bar_snapshot()
|
|
assert snapshot["idle_since"].startswith("✓ ")
|
|
|
|
|