test(tools): comprehensive test suite for browser_tools module (+75 tests)

[manus-use]

Summary

Comprehensive test suite for the browser_tools module (src/manus_agent/tools/browser_tools.py) — previously at zero dedicated test coverage.

What's tested (75 tests)

Module importability (4 tests)

• All expected symbols exist (21 tool functions + classes)
• BROWSER_USE_AVAILABLE flag type check
• Strands @tool decorator verification

BrowserAgentSession.init (7 tests)

• Default headless=True behavior
• Explicit headless parameter override
• Config-driven headless from browser_use section
• Explicit param priority over config
• Config without browser_use attribute fallback
• Config/None storage

BrowserAgentSession._get_llm (8 tests)

• Bedrock provider with default config
• Browser-use config overrides (model, temperature, max_tokens)
• AWS region from environment variable
• Default us-east-1 when no env var
• Unsupported provider raises ValueError
• No-config defaults
• Fallback from browser_use.provider=None to main LLM config
• Config without browser_use uses llm config directly

BrowserAgentSession.run_task (8 tests)

• ImportError when BROWSER_USE_AVAILABLE=False
• Full agent creation + run flow
• Callable extracted_content handling
• String extracted_content (direct return)
• all_results fallback (reversed search for is_done)
• str(result) fallback when no structured output
• Headless=False passed to BrowserProfile
• Headless=True passed to BrowserProfile

BrowserAgentSession.cleanup (1 test)

• No-op completion without error

get_browser_session singleton (5 tests)

• Creates new session
• Returns existing singleton
• Passes headless parameter
• Passes config parameter
• Ignores params on existing session (first-creation wins)

browser_do tool (6 tests)

• Success returns {success, result, task}
• Error returns {success=False, error, task}
• Headless from browser_use config
• Explicit headless override
• Config.from_file exception handling
• Fallback to tools.browser_headless when no browser_use

browser_cleanup tool (3 tests)

• Cleanup with no active session
• Cleanup with active session (calls cleanup, resets global)
• Exception during cleanup

Browser action tools (22 tests)

Every delegating tool verified for correct task string construction:

• browser_navigate, browser_search_google, browser_go_back
• browser_wait (default + custom seconds)
• browser_click_element (no index + with index)
• browser_input_text (no index + with index)
• browser_save_pdf (no filename + with filename)
• browser_switch_tab, browser_open_tab, browser_close_tab
• browser_extract_content (plain + with links)
• browser_get_page_info
• browser_scroll_down (default + pixels), browser_scroll_up (default + pixels)
• browser_scroll_to_text, browser_send_keys
• browser_select_dropdown, browser_drag_drop
• Error propagation through action tools

web_search tool (1 test)

• Fallback to browser_do when web_search import fails

Edge cases (4 tests)

• None config handling
• Config without browser_use attribute in browser_do
• Global session reset on cleanup
• Task preserved in error dict

BROWSER_USE_AVAILABLE guard (2 tests)

• Import error message content
• Install instructions in error

Design

• 100% mocked — no real HTTP, no real browser, no network calls
• Tests the __wrapped__ async functions directly (bypasses strands @tool decorator overhead)
• Mocks at Config.from_file + get_browser_session level for action tools
• Mocks at BrowserUseAgent/BrowserProfile/Controller/ChatBedrock level for session tests
• autouse fixture resets the global singleton between every test

Test results

1233 passed, 3 deselected, 0 failures

Baseline: 1158 existing + 75 new = 1233 total passing tests.

Open PRs checked (no overlap)

Checked all 50 open PRs (#134–#183). Closest related:

• test(tools): comprehensive test suite for browser_utils module (+107 tests) #153 (test/browser_utils) — tests browser_utils.py (different module: prepare_evaluate_script helper), NOT browser_tools.py
• test(agents): comprehensive test suite for BaseManusAgent, BrowserAgent, DataAnalysisAgent, and tools registry (+60 tests) #157 (test/base-browser-data-agents-suite) — tests agent classes, not the tools module
• test: add comprehensive tests for BrowserUseAgent.stream_async #159 (test/stream-async-browser-agent) — tests BrowserUseAgent.stream_async, not browser_tools

No open or merged PR provides dedicated test coverage for browser_tools.py.