EvoScientist-Multi/tests/test_web_search.py

from __future__ import annotations

import pytest


@pytest.mark.anyio
async def test_tavily_search_keeps_indexed_summary_when_source_fetch_fails(monkeypatch):
    from EvoScientist.tools import search

    class _Client:
        def search(self, *_args, **_kwargs):
            return {
                "results": [
                    {
                        "title": "AIR staff profile",
                        "url": "https://air.cas.cn/example",
                        "content": "Indexed staff-profile summary.",
                    }
                ]
            }

    recorded: list[tuple[str, str]] = []

    async def fetch_failed(_url: str, timeout: float = 10.0) -> str:
        return "Error fetching content from https://air.cas.cn/example: DNS failed"

    async def record(service: str, action: str) -> None:
        recorded.append((service, action))

    monkeypatch.setattr(search, "_get_tavily_client", lambda: _Client())
    monkeypatch.setattr(search, "fetch_webpage_content", fetch_failed)
    monkeypatch.setattr("EvoScientist.runtime_integrations.record_service_usage", record)

    result = await search.tavily_search.ainvoke({"query": "高铭 空天院"})

    assert "Indexed staff-profile summary." in result
    assert "https://air.cas.cn/example" in result
    assert "Tavily-indexed summary" in result
    assert recorded == [("tavily", "search")]


@pytest.mark.anyio
async def test_tavily_search_bounds_fetched_page_content(monkeypatch):
    from EvoScientist.tools import search

    class _Client:
        def search(self, *_args, **_kwargs):
            return {
                "results": [
                    {
                        "title": f"Result {index}",
                        "url": f"https://example.com/{index}",
                        "content": f"Indexed summary {index}",
                    }
                    for index in range(3)
                ]
            }

    async def huge_page(_url: str, timeout: float = 10.0) -> str:
        return "page-content " * 10_000

    monkeypatch.setattr(search, "_get_tavily_client", lambda: _Client())
    monkeypatch.setattr(search, "fetch_webpage_content", huge_page)

    result = await search.tavily_search.ainvoke({"query": "bounded search"})

    assert len(result) <= search.MAX_SEARCH_RESULT_CHARS
    for index in range(3):
        assert f"https://example.com/{index}" in result
    assert "[page content truncated]" in result


@pytest.mark.anyio
async def test_tavily_search_preserves_every_result_url_under_total_budget(monkeypatch):
    from EvoScientist.tools import search

    class _Client:
        def search(self, *_args, **_kwargs):
            return {
                "results": [
                    {
                        "title": f"Result {index} " + ("very-long-title " * 800),
                        "url": f"https://example.com/result-{index}",
                        "content": f"Indexed summary {index}",
                    }
                    for index in range(3)
                ]
            }

    async def page(_url: str, timeout: float = 10.0) -> str:
        return "page-content " * 1_000

    monkeypatch.setattr(search, "_get_tavily_client", lambda: _Client())
    monkeypatch.setattr(search, "fetch_webpage_content", page)

    result = await search.tavily_search.ainvoke({"query": "preserve urls"})

    assert len(result) <= search.MAX_SEARCH_RESULT_CHARS
    for index in range(3):
        assert f"https://example.com/result-{index}" in result
    assert "[search result content truncated to preserve all titles and URLs]" in result


@pytest.mark.anyio
async def test_tavily_search_bounds_maliciously_long_url(monkeypatch):
    from EvoScientist.tools import search

    long_url = "https://example.com/" + ("a" * 20_000)

    class _Client:
        def search(self, *_args, **_kwargs):
            return {
                "results": [
                    {
                        "title": "Long URL result",
                        "url": long_url,
                        "content": "Indexed summary",
                    }
                ]
            }

    async def page(_url: str, timeout: float = 10.0) -> str:
        return "page"

    monkeypatch.setattr(search, "_get_tavily_client", lambda: _Client())
    monkeypatch.setattr(search, "fetch_webpage_content", page)

    result = await search.tavily_search.ainvoke({"query": "long url"})

    assert len(result) <= search.MAX_SEARCH_RESULT_CHARS
    assert "https://example.com/" in result
    assert "[URL truncated]" in result