hermes-agent/tests/tools/test_browser_lightpanda.py
Teknium 39975613b1
test: prune wave 2 + speed fixes — 28,106 → 19,757 test functions, suite wall 315s → 294s
Second, deeper pass over tools/gateway/hermes_cli plus first pass over
the trees wave 1 missed (acp, acp_adapter, skills, computer_use, docker,
dashboard, conformance, monitoring, secret_sources, hermes_state,
providers). Same rubric as wave 1 (AGENTS.md test policy); security,
alternation/caching invariants, issue-number regressions, and E2E kept.

Real test-quality fixes found and rooted out along the way:
- tests/tools/test_command_guards.py made real auxiliary-LLM HTTPS calls
  (DEFAULT_CONFIG smart-approval leaked in) — pinned approval
  mode=manual via autouse fixture: 17.4s → 0.4s.
- test_model_switch_custom_providers.py / test_user_providers_model_switch.py
  silently probed live provider catalogs (~2s/test) — stubbed
  cached_provider_model_ids/provider_model_ids/fetch_api_models.
- test_telegram_noise_filter.py: 15-platform copy-paste matrix over
  shared gateway.run logic → 3 representative platforms (55s → 3.9s).
- test_gateway_shutdown.py: stop()'s 5s interrupt-deadline loop spun on
  MagicMock agents — interrupt.side_effect now clears _running_agents
  (22s → 1.0s).
- test_gateway_inactivity_timeout.py poll-harness timings shrunk 3-5x
  (24s → 1.1s); test_mcp_stability.py backoff/SIGTERM-grace sleeps
  patched (15.4s → 2.5s); test_async_delegation.py negative-drain wait
  5s → 0.5s.
- test_telegram_init_deadline.py: loop-block margin restored to 1.0s
  with rationale comment — the watchdog-dump assertion needs the loop
  blocked well past deadline+grace under parallel load (flaked once in
  the 40-worker verification run at a 0.2s margin).

Verification: full hermetic suite via scripts/run_tests.sh —
2,438 files, 21,718 tests passed, 0 failed, 293.9s wall.
Suite totals vs original baseline: 46,820 → 19,757 test functions
(−57.8%), wall 583.5s → 293.9s (−50%), subprocess CPU 13,564s → 11,623s.
2026-07-29 13:39:40 -07:00

430 lines
19 KiB
Python

"""Tests for Lightpanda engine support in browser_tool.py."""
import json
import os
from unittest.mock import MagicMock, patch
import pytest
# ---------------------------------------------------------------------------
# Helpers
# ---------------------------------------------------------------------------
def _reset_engine_cache():
"""Reset the module-level engine cache so tests start clean."""
import tools.browser_tool as bt
bt._cached_browser_engine = None
bt._browser_engine_resolved = False
@pytest.fixture(autouse=True)
def _clean_engine_cache():
"""Reset engine cache before and after each test."""
_reset_engine_cache()
yield
_reset_engine_cache()
# ---------------------------------------------------------------------------
# _get_browser_engine
# ---------------------------------------------------------------------------
class TestGetBrowserEngine:
"""Test engine resolution from config and env vars."""
def test_default_is_auto(self):
"""With no config or env var, engine defaults to 'auto'."""
from tools.browser_tool import _get_browser_engine
with patch.dict(os.environ, {}, clear=False):
os.environ.pop("AGENT_BROWSER_ENGINE", None)
with patch("hermes_cli.config.read_raw_config", return_value={}):
assert _get_browser_engine() == "auto"
def test_config_lightpanda(self):
"""Config browser.engine = 'lightpanda' is respected."""
from tools.browser_tool import _get_browser_engine
cfg = {"browser": {"engine": "lightpanda"}}
with patch("hermes_cli.config.read_raw_config", return_value=cfg):
assert _get_browser_engine() == "lightpanda"
def test_caching(self):
"""Result is cached — second call doesn't re-read config."""
from tools.browser_tool import _get_browser_engine
mock_read = MagicMock(return_value={"browser": {"engine": "lightpanda"}})
with patch("hermes_cli.config.read_raw_config", mock_read):
assert _get_browser_engine() == "lightpanda"
assert _get_browser_engine() == "lightpanda"
mock_read.assert_called_once()
# ---------------------------------------------------------------------------
# _should_inject_engine
# ---------------------------------------------------------------------------
class TestShouldInjectEngine:
"""Test whether --engine flag is injected based on mode."""
def test_auto_never_injects(self):
from tools.browser_tool import _should_inject_engine
assert _should_inject_engine("auto") is False
def test_lightpanda_injects_in_local_mode(self):
from tools.browser_tool import _should_inject_engine
with patch("tools.browser_tool._is_camofox_mode", return_value=False), \
patch("tools.browser_tool._get_cdp_override", return_value=""), \
patch("tools.browser_tool._get_cloud_provider", return_value=None):
assert _should_inject_engine("lightpanda") is True
def test_chrome_injects_in_local_mode(self):
from tools.browser_tool import _should_inject_engine
with patch("tools.browser_tool._is_camofox_mode", return_value=False), \
patch("tools.browser_tool._get_cdp_override", return_value=""), \
patch("tools.browser_tool._get_cloud_provider", return_value=None):
assert _should_inject_engine("chrome") is True
def test_no_inject_in_camofox_mode(self):
from tools.browser_tool import _should_inject_engine
with patch("tools.browser_tool._is_camofox_mode", return_value=True):
assert _should_inject_engine("lightpanda") is False
def test_no_inject_with_cdp_override(self):
from tools.browser_tool import _should_inject_engine
with patch("tools.browser_tool._is_camofox_mode", return_value=False), \
patch("tools.browser_tool._get_cdp_override_raw", return_value="ws://localhost:9222"):
assert _should_inject_engine("lightpanda") is False
# ---------------------------------------------------------------------------
# _needs_lightpanda_fallback
# ---------------------------------------------------------------------------
class TestNeedsLightpandaFallback:
"""Test fallback detection for Lightpanda results."""
def test_non_lightpanda_never_falls_back(self):
from tools.browser_tool import _needs_lightpanda_fallback
result = {"success": False, "error": "timeout"}
assert _needs_lightpanda_fallback("chrome", "open", result) is False
assert _needs_lightpanda_fallback("auto", "open", result) is False
def test_failed_command_triggers_fallback(self):
from tools.browser_tool import _needs_lightpanda_fallback
result = {"success": False, "error": "page.goto: Timeout"}
assert _needs_lightpanda_fallback("lightpanda", "open", result) is True
def test_empty_snapshot_triggers_fallback(self):
from tools.browser_tool import _needs_lightpanda_fallback
result = {"success": True, "data": {"snapshot": ""}}
assert _needs_lightpanda_fallback("lightpanda", "snapshot", result) is True
def test_unknown_command_does_not_trigger_fallback(self):
"""Commands not in the whitelist should not trigger fallback."""
from tools.browser_tool import _needs_lightpanda_fallback
result = {"success": False, "error": "nope"}
assert _needs_lightpanda_fallback("lightpanda", "some_future_cmd", result) is False
# ---------------------------------------------------------------------------
# Config integration
# ---------------------------------------------------------------------------
class TestConfigIntegration:
"""Verify engine config is in DEFAULT_CONFIG."""
def test_engine_in_default_config(self):
from hermes_cli.config import DEFAULT_CONFIG
assert "engine" in DEFAULT_CONFIG["browser"]
assert DEFAULT_CONFIG["browser"]["engine"] == "auto"
def test_env_var_registered(self):
from hermes_cli.config import OPTIONAL_ENV_VARS
assert "AGENT_BROWSER_ENGINE" in OPTIONAL_ENV_VARS
entry = OPTIONAL_ENV_VARS["AGENT_BROWSER_ENGINE"]
assert entry["category"] == "tool"
assert entry["advanced"] is True
class TestLightpandaRequirements:
"""Lightpanda should expose browser tools without local Chromium."""
def test_lightpanda_local_mode_does_not_require_chromium(self):
import tools.browser_tool as bt
with patch("tools.browser_tool._is_camofox_mode", return_value=False), \
patch("tools.browser_tool._get_cdp_override", return_value=""), \
patch("tools.browser_tool._find_agent_browser", return_value="/usr/bin/agent-browser"), \
patch("tools.browser_tool._requires_real_termux_browser_install", return_value=False), \
patch("tools.browser_tool._get_cloud_provider", return_value=None), \
patch("tools.browser_tool._get_browser_engine", return_value="lightpanda"), \
patch("tools.browser_tool._chromium_installed", return_value=False):
assert bt.check_browser_requirements() is True
def test_chrome_local_mode_still_requires_chromium(self):
import tools.browser_tool as bt
with patch("tools.browser_tool._is_camofox_mode", return_value=False), \
patch("tools.browser_tool._get_cdp_override", return_value=""), \
patch("tools.browser_tool._find_agent_browser", return_value="/usr/bin/agent-browser"), \
patch("tools.browser_tool._requires_real_termux_browser_install", return_value=False), \
patch("tools.browser_tool._get_cloud_provider", return_value=None), \
patch("tools.browser_tool._get_browser_engine", return_value="auto"), \
patch("tools.browser_tool._chromium_installed", return_value=False):
assert bt.check_browser_requirements() is False
# ---------------------------------------------------------------------------
# cleanup_all_browsers resets engine cache
# ---------------------------------------------------------------------------
class TestCleanupResetsEngineCache:
"""Verify cleanup_all_browsers resets engine-related globals."""
def test_engine_cache_reset(self):
import tools.browser_tool as bt
# Seed the cache
bt._cached_browser_engine = "lightpanda"
bt._browser_engine_resolved = True
# cleanup should reset them
bt.cleanup_all_browsers()
assert bt._cached_browser_engine is None
assert bt._browser_engine_resolved is False
# ---------------------------------------------------------------------------
# fallback warning annotation
# ---------------------------------------------------------------------------
class TestLightpandaFallbackWarning:
"""Verify Chrome fallback results are annotated for users."""
def test_fallback_result_gets_user_visible_warning(self):
from tools.browser_tool import _annotate_lightpanda_fallback
result = {"success": True, "data": {"snapshot": "- heading \"Hello\" [ref=e1]"}}
annotated = _annotate_lightpanda_fallback(
result,
"Lightpanda returned an empty/too-short snapshot; retried with Chrome.",
)
assert annotated["browser_engine"] == "chrome"
assert "Lightpanda fallback" in annotated["fallback_warning"]
assert annotated["browser_engine_fallback"] == {
"from": "lightpanda",
"to": "chrome",
"reason": "Lightpanda returned an empty/too-short snapshot; retried with Chrome.",
}
assert annotated["data"]["fallback_warning"] == annotated["fallback_warning"]
assert annotated["data"]["browser_engine"] == "chrome"
def test_browser_navigate_surfaces_fallback_warning(self):
import json
import tools.browser_tool as bt
result = bt._annotate_lightpanda_fallback(
{"success": True, "data": {"title": "Fallback OK", "url": "https://example.com/"}},
"synthetic Lightpanda failure; retried with Chrome.",
)
with patch("tools.browser_tool._is_local_backend", return_value=True), \
patch("tools.browser_tool._get_cloud_provider", return_value=None), \
patch("tools.browser_tool._get_session_info", return_value={
"session_name": "test", "_first_nav": False, "features": {"local": True, "proxies": True}
}), \
patch("tools.browser_tool._run_browser_command", side_effect=[
result,
{"success": True, "data": {"snapshot": "- heading \"Fallback OK\" [ref=e1]", "refs": {"e1": {}}}},
]):
response = json.loads(bt.browser_navigate("https://example.com", task_id="warn-test"))
assert response["success"] is True
assert response["browser_engine"] == "chrome"
assert "Lightpanda fallback" in response["fallback_warning"]
assert response["browser_engine_fallback"]["from"] == "lightpanda"
assert response["browser_engine_fallback"]["to"] == "chrome"
bt._last_active_session_key.pop("warn-test", None)
def test_browser_vision_lightpanda_response_has_structured_fallback(self, tmp_path):
import json
import tools.browser_tool as bt
chrome_shot = tmp_path / "chrome-structured.png"
chrome_shot.write_bytes(b"\x89PNG" + b"0" * 128)
class _Msg:
content = "Example Domain screenshot"
class _Choice:
message = _Msg()
class _Response:
choices = [_Choice()]
with patch("tools.browser_tool._get_browser_engine", return_value="lightpanda"), \
patch("tools.browser_tool._should_inject_engine", return_value=True), \
patch("tools.browser_tool._chrome_fallback_screenshot", return_value={
"success": True, "data": {"path": str(chrome_shot)}
}), \
patch("hermes_constants.get_hermes_dir", return_value=tmp_path), \
patch("tools.browser_tool.call_llm", return_value=_Response()):
response = json.loads(bt.browser_vision("what is this?", task_id="vision-structured"))
assert response["success"] is True
assert response["browser_engine"] == "chrome"
assert response["browser_engine_fallback"] == {
"from": "lightpanda",
"to": "chrome",
"reason": "Lightpanda has no graphical renderer for screenshots; used Chrome for vision capture.",
}
# ---------------------------------------------------------------------------
# _engine_override parameter
# ---------------------------------------------------------------------------
class TestEngineOverride:
"""Verify _engine_override bypasses the cached engine."""
@patch("tools.browser_tool._get_session_info")
@patch("tools.browser_tool._find_agent_browser", return_value="/usr/bin/agent-browser")
@patch("tools.browser_tool._is_local_mode", return_value=True)
@patch("tools.browser_tool._chromium_installed", return_value=True)
@patch("tools.browser_tool._get_cloud_provider", return_value=None)
@patch("tools.browser_tool._get_cdp_override", return_value="")
@patch("tools.browser_tool._is_camofox_mode", return_value=False)
def test_override_prevents_engine_injection(
self, _camofox, _cdp, _cloud, _chromium, _local, _find, _session
):
"""When _engine_override='auto', --engine flag is NOT injected."""
import tools.browser_tool as bt
# Set the global cache to lightpanda
bt._cached_browser_engine = "lightpanda"
bt._browser_engine_resolved = True
_session.return_value = {"session_name": "test-sess"}
# Track the cmd_parts that Popen receives
captured_cmds = []
mock_proc = MagicMock()
mock_proc.wait.return_value = None
mock_proc.returncode = 0
def capture_popen(cmd, **kwargs):
captured_cmds.append(cmd)
return mock_proc
# We need to mock the file operations too
with patch("subprocess.Popen", side_effect=capture_popen), \
patch("os.open", return_value=99), \
patch("os.close"), \
patch("os.unlink"), \
patch("os.makedirs"), \
patch("builtins.open", MagicMock(return_value=MagicMock(
__enter__=MagicMock(return_value=MagicMock(read=MagicMock(return_value='{"success": true, "data": {}}'))),
__exit__=MagicMock(return_value=False),
))), \
patch("tools.interrupt.is_interrupted", return_value=False), \
patch("tools.browser_tool._write_owner_pid"):
bt._run_browser_command("task1", "snapshot", [], _engine_override="auto")
# Should NOT contain "--engine" since override is "auto"
assert len(captured_cmds) == 1
assert "--engine" not in captured_cmds[0]
@patch("tools.browser_tool._get_session_info")
@patch("tools.browser_tool._find_agent_browser", return_value="/usr/bin/agent-browser")
@patch("tools.browser_tool._is_local_mode", return_value=True)
@patch("tools.browser_tool._chromium_installed", return_value=True)
@patch("tools.browser_tool._get_cloud_provider", return_value=None)
@patch("tools.browser_tool._get_cdp_override", return_value="")
@patch("tools.browser_tool._is_camofox_mode", return_value=False)
def test_no_override_uses_cached_engine(
self, _camofox, _cdp, _cloud, _chromium, _local, _find, _session
):
"""Without _engine_override, the cached engine is used."""
import tools.browser_tool as bt
bt._cached_browser_engine = "lightpanda"
bt._browser_engine_resolved = True
_session.return_value = {"session_name": "test-sess"}
captured_cmds = []
mock_proc = MagicMock()
mock_proc.wait.return_value = None
mock_proc.returncode = 0
def capture_popen(cmd, **kwargs):
captured_cmds.append(cmd)
return mock_proc
# Return a substantive snapshot so the LP fallback does NOT trigger.
mock_stdout = '{"success": true, "data": {"snapshot": "- heading \\"Hello\\" [ref=e1]", "refs": {"e1": {}}}}'
with patch("subprocess.Popen", side_effect=capture_popen), \
patch("os.open", return_value=99), \
patch("os.close"), \
patch("os.unlink"), \
patch("os.makedirs"), \
patch("builtins.open", MagicMock(return_value=MagicMock(
__enter__=MagicMock(return_value=MagicMock(read=MagicMock(return_value=mock_stdout))),
__exit__=MagicMock(return_value=False),
))), \
patch("tools.interrupt.is_interrupted", return_value=False), \
patch("tools.browser_tool._write_owner_pid"):
bt._run_browser_command("task1", "snapshot", [])
# SHOULD contain "--engine lightpanda"
assert len(captured_cmds) == 1
assert "--engine" in captured_cmds[0]
engine_idx = captured_cmds[0].index("--engine")
assert captured_cmds[0][engine_idx + 1] == "lightpanda"
def test_hybrid_local_sidecar_injects_engine_even_with_cloud_provider(self):
"""A task::local sidecar is local even when global cloud config exists."""
import tools.browser_tool as bt
bt._cached_browser_engine = "lightpanda"
bt._browser_engine_resolved = True
captured_cmds = []
mock_provider = MagicMock()
mock_proc = MagicMock()
mock_proc.wait.return_value = None
mock_proc.returncode = 0
def capture_popen(cmd, **kwargs):
captured_cmds.append(cmd)
return mock_proc
mock_stdout = json.dumps({
"success": True,
"data": {"snapshot": '- heading "Hello" [ref=e1]', "refs": {"e1": {}}},
})
with patch("tools.browser_tool._get_session_info", return_value={"session_name": "local-sidecar"}), \
patch("tools.browser_tool._find_agent_browser", return_value="/usr/bin/agent-browser"), \
patch("tools.browser_tool._is_local_mode", return_value=False), \
patch("tools.browser_tool._chromium_installed", return_value=True), \
patch("tools.browser_tool._get_cloud_provider", return_value=mock_provider), \
patch("tools.browser_tool._get_cdp_override", return_value=""), \
patch("tools.browser_tool._is_camofox_mode", return_value=False), \
patch("subprocess.Popen", side_effect=capture_popen), \
patch("os.open", return_value=99), \
patch("os.close"), \
patch("os.unlink"), \
patch("os.makedirs"), \
patch("builtins.open", MagicMock(return_value=MagicMock(
__enter__=MagicMock(return_value=MagicMock(read=MagicMock(return_value=mock_stdout))),
__exit__=MagicMock(return_value=False),
))), \
patch("tools.interrupt.is_interrupted", return_value=False), \
patch("tools.browser_tool._write_owner_pid"):
bt._run_browser_command("task::local", "snapshot", [])
assert len(captured_cmds) == 1
assert "--engine" in captured_cmds[0]
assert captured_cmds[0][captured_cmds[0].index("--engine") + 1] == "lightpanda"