Skip to content

Personal AI, On Personal Devices

Python

fix(tools): return labeled page content from web_search (#480)

Jon Saad-Falconf922811cba5d
diff --git a/src/openjarvis/tools/web_search.py b/src/openjarvis/tools/web_search.py
index fcf5fe1a..a90d0cec 100644
--- a/src/openjarvis/tools/web_search.py
+++ b/src/openjarvis/tools/web_search.py
@@ -120,12 +120,15 @@ class WebSearchTool(BaseTool):
from ddgs import DDGS
ddgs = DDGS()
- results = list(ddgs.text(query, max_results=max_results))
- formatted = "\n\n".join(
- f"**{r.get('title', 'Untitled')}**\n"
- f"{r.get('href', '')}\n{r.get('body', '')}"
- for r in results
- )
+ raw_results = list(ddgs.text(query, max_results=max_results))
+ results = []
+ for r in raw_results:
+ title = r.get("title", "Untitled")
+ url = r.get("href", "")
+ snippet = r.get("body", "")
+ results.append(f"### {title}\nSource: {url}\nSummary: {snippet}")
+
+ formatted = "\n\n---\n\n".join(results)
return formatted
def execute(self, **params: Any) -> ToolResult:
@@ -161,13 +164,20 @@ class WebSearchTool(BaseTool):
from tavily import TavilyClient
client = TavilyClient(api_key=self._api_key)
- response = client.search(query, max_results=max_results)
- results = response.get("results", [])
- formatted = "\n\n".join(
- f"**{r.get('title', 'Untitled')}**\n"
- f"{r.get('url', '')}\n{r.get('content', '')}"
- for r in results
+ response = client.search(
+ query, max_results=max_results, search_depth="advanced"
)
+ results = response.get("results", [])
+ formatted_parts = []
+ for r in results:
+ title = r.get("title", "Untitled")
+ url = r.get("url", "")
+ content = r.get("content", "") or r.get("snippet", "")
+ formatted_parts.append(
+ f"### {title}\nSource: {url}\nSummary: {content}"
+ )
+
+ formatted = "\n\n---\n\n".join(formatted_parts)
return ToolResult(
tool_name="web_search",
content=formatted or "No results found.",
diff --git a/tests/tools/test_web_search.py b/tests/tools/test_web_search.py
index cdfdfc21..570c3019 100644
--- a/tests/tools/test_web_search.py
+++ b/tests/tools/test_web_search.py
@@ -170,7 +170,9 @@ class TestWebSearchTool:
tool = WebSearchTool(api_key="test-key", max_results=3)
tool.execute(query="test", max_results=7)
- mock_client.search.assert_called_once_with("test", max_results=7)
+ mock_client.search.assert_called_once_with(
+ "test", max_results=7, search_depth="advanced"
+ )
def test_to_openai_function(self):
tool = WebSearchTool(api_key="test-key")
@@ -231,6 +233,82 @@ class TestWebSearchTool:
ToolRegistry.register_value("web_search", WebSearchTool)
assert ToolRegistry.contains("web_search")
+ def test_tavily_results_use_labeled_content_format(self, monkeypatch):
+ """Regression for #390: results expose page CONTENT under labeled
+ Source/Summary headings (so agents synthesize content, not echo
+ URLs), and Tavily is queried with search_depth='advanced'."""
+ import builtins
+
+ mock_client = MagicMock()
+ mock_client.search.return_value = {
+ "results": [
+ {
+ "title": "Result 1",
+ "url": "https://example.com/1",
+ "content": "Content about test.",
+ },
+ ]
+ }
+ mock_tavily_module = MagicMock()
+ mock_tavily_module.TavilyClient.return_value = mock_client
+ original_import = builtins.__import__
+
+ def _mock_import(name, *args, **kwargs):
+ if name == "tavily":
+ return mock_tavily_module
+ if name == "tavily.errors":
+ mock_errors = MagicMock()
+ mock_errors.UsageLimitExceededError = Exception
+ return mock_errors
+ return original_import(name, *args, **kwargs)
+
+ monkeypatch.setattr(builtins, "__import__", _mock_import)
+
+ tool = WebSearchTool(api_key="test-key")
+ result = tool.execute(query="test query")
+ assert result.success is True
+ # Labeled structure with the page content surfaced.
+ assert "### Result 1" in result.content
+ assert "Source: https://example.com/1" in result.content
+ assert "Summary: Content about test." in result.content
+ # search_depth='advanced' is what pulls richer content from Tavily.
+ _, kwargs = mock_client.search.call_args
+ assert kwargs.get("search_depth") == "advanced"
+
+ def test_tavily_falls_back_to_snippet_when_no_content(self, monkeypatch):
+ """When a Tavily result lacks 'content', the 'snippet' field is used
+ for the Summary rather than rendering an empty summary."""
+ import builtins
+
+ mock_client = MagicMock()
+ mock_client.search.return_value = {
+ "results": [
+ {
+ "title": "Snippet Only",
+ "url": "https://example.com/s",
+ "snippet": "Fallback snippet text.",
+ },
+ ]
+ }
+ mock_tavily_module = MagicMock()
+ mock_tavily_module.TavilyClient.return_value = mock_client
+ original_import = builtins.__import__
+
+ def _mock_import(name, *args, **kwargs):
+ if name == "tavily":
+ return mock_tavily_module
+ if name == "tavily.errors":
+ mock_errors = MagicMock()
+ mock_errors.UsageLimitExceededError = Exception
+ return mock_errors
+ return original_import(name, *args, **kwargs)
+
+ monkeypatch.setattr(builtins, "__import__", _mock_import)
+
+ tool = WebSearchTool(api_key="test-key")
+ result = tool.execute(query="test query")
+ assert "Summary: Fallback snippet text." in result.content
+
# ---------------------------------------------------------------------------
# URL detection and fetching tests