Avoid mutating extract tool params and validate schema JSON

Co-authored-by: Shri Sukhani <shrisukhani@users.noreply.github.com>

Author: cursoragent

hyperbrowser/tools/__init__.py

@@ -1,4 +1,7 @@
 import json
+from typing import Any, Dict, Mapping
+
+from hyperbrowser.exceptions import HyperbrowserError
 from hyperbrowser.models.agents.browser_use import StartBrowserUseTaskParams
 from hyperbrowser.models.crawl import StartCrawlJobParams
 from hyperbrowser.models.extract import StartExtractJobParams
@@ -21,6 +24,20 @@
 )
 
 
+def _prepare_extract_tool_params(params: Mapping[str, Any]) -> Dict[str, Any]:
+    normalized_params: Dict[str, Any] = dict(params)
+    schema_value = normalized_params.get("schema")
+    if isinstance(schema_value, str):
+        try:
+            normalized_params["schema"] = json.loads(schema_value)
+        except json.JSONDecodeError as exc:
+            raise HyperbrowserError(
+                "Invalid JSON string provided for `schema` in extract tool params",
+                original_error=exc,
+            ) from exc
+    return normalized_params
+
+
 class WebsiteScrapeTool:
     openai_tool_definition = SCRAPE_TOOL_OPENAI
     anthropic_tool_definition = SCRAPE_TOOL_ANTHROPIC
@@ -86,16 +103,18 @@ class WebsiteExtractTool:
 
     @staticmethod
     def runnable(hb: Hyperbrowser, params: dict) -> str:
-        if params.get("schema") and isinstance(params.get("schema"), str):
-            params["schema"] = json.loads(params["schema"])
-        resp = hb.extract.start_and_wait(params=StartExtractJobParams(**params))
+        normalized_params = _prepare_extract_tool_params(params)
+        resp = hb.extract.start_and_wait(
+            params=StartExtractJobParams(**normalized_params)
+        )
         return json.dumps(resp.data) if resp.data else ""
 
     @staticmethod
     async def async_runnable(hb: AsyncHyperbrowser, params: dict) -> str:
-        if params.get("schema") and isinstance(params.get("schema"), str):
-            params["schema"] = json.loads(params["schema"])
-        resp = await hb.extract.start_and_wait(params=StartExtractJobParams(**params))
+        normalized_params = _prepare_extract_tool_params(params)
+        resp = await hb.extract.start_and_wait(
+            params=StartExtractJobParams(**normalized_params)
+        )
         return json.dumps(resp.data) if resp.data else ""
 
 

tests/test_tools_extract.py

@@ -0,0 +1,84 @@
+import asyncio
+
+import pytest
+
+from hyperbrowser.exceptions import HyperbrowserError
+from hyperbrowser.models.extract import StartExtractJobParams
+from hyperbrowser.tools import WebsiteExtractTool
+
+
+class _Response:
+    def __init__(self, data):
+        self.data = data
+
+
+class _SyncExtractManager:
+    def __init__(self):
+        self.last_params = None
+
+    def start_and_wait(self, params: StartExtractJobParams):
+        self.last_params = params
+        return _Response({"ok": True})
+
+
+class _AsyncExtractManager:
+    def __init__(self):
+        self.last_params = None
+
+    async def start_and_wait(self, params: StartExtractJobParams):
+        self.last_params = params
+        return _Response({"ok": True})
+
+
+class _SyncClient:
+    def __init__(self):
+        self.extract = _SyncExtractManager()
+
+
+class _AsyncClient:
+    def __init__(self):
+        self.extract = _AsyncExtractManager()
+
+
+def test_extract_tool_runnable_does_not_mutate_input_params():
+    client = _SyncClient()
+    params = {
+        "urls": ["https://example.com"],
+        "schema": '{"type":"object","properties":{"name":{"type":"string"}}}',
+    }
+
+    output = WebsiteExtractTool.runnable(client, params)
+
+    assert output == '{"ok": true}'
+    assert isinstance(client.extract.last_params, StartExtractJobParams)
+    assert isinstance(client.extract.last_params.schema_, dict)
+    assert params["schema"] == '{"type":"object","properties":{"name":{"type":"string"}}}'
+
+
+def test_extract_tool_async_runnable_does_not_mutate_input_params():
+    client = _AsyncClient()
+    params = {
+        "urls": ["https://example.com"],
+        "schema": '{"type":"object","properties":{"name":{"type":"string"}}}',
+    }
+
+    async def run():
+        return await WebsiteExtractTool.async_runnable(client, params)
+
+    output = asyncio.run(run())
+
+    assert output == '{"ok": true}'
+    assert isinstance(client.extract.last_params, StartExtractJobParams)
+    assert isinstance(client.extract.last_params.schema_, dict)
+    assert params["schema"] == '{"type":"object","properties":{"name":{"type":"string"}}}'
+
+
+def test_extract_tool_runnable_raises_for_invalid_schema_json():
+    client = _SyncClient()
+    params = {
+        "urls": ["https://example.com"],
+        "schema": "{invalid-json}",
+    }
+
+    with pytest.raises(HyperbrowserError, match="Invalid JSON string provided for `schema`"):
+        WebsiteExtractTool.runnable(client, params)