Merge pull request #181 from nextcloud/feat/web-fetch

feat(tools): Add web_fetch tool

Author: marcelklehr

ex_app/lib/agent.py

@@ -127,7 +127,7 @@ async def call_model(
 At the end of each message to the user, if you have carried out a task or answered a question, suggest up to three actions for things you can do for the user based on the tools you have available and details of the previous task. For example: If the user wants to know the weather for some location, they might be planning an event, you can suggest to create an event for them, or if they searched for a file, they may want to share it with others, suggest to create a share link for them, if they want a summary of something, you can suggest them to send the summary to somebody.
 """
 		if tool_enabled("duckduckgo_results_json"):
-			system_prompt_text += "Only use the duckduckgo_results_json tool if the user explicitly asks for a web search.\n"
+			system_prompt_text += "Use the duckduckgo_results_json tool if the user explicitly asks for a web search or you don't know about a topic or concept that the user is referencing.\n"
 		if tool_enabled("list_talk_conversations"):
 			system_prompt_text += "Use the list_talk_conversations tool to check which conversations exist.\n"
 		if tool_enabled("list_calendars"):
@@ -140,8 +140,10 @@ async def call_model(
 			system_prompt_text += "Use the find_details_of_current_user tool to find the current user's location and timezone.\n"
 		if tool_enabled("list_mails"):
 			system_prompt_text += "Always check for the mail account id before requesting a folder list.\n"
+		if tool_enabled("web_fetch"):
+			system_prompt_text += "Use the web_fetch tool to fetch web content. You can fetch the complete page content of a duckduckgo search result using web_fetch as well.\n"
 
-		if task['input'].get('memories', None) is not None:
+		if task['input'].get('memories', None) is not None and task['input'].get('memories', None) is not []:
 			system_prompt_text += "You can remember things from other conversations with the user. If relevant, take into account the following memories:\n\n" + "\n".join(task['input']['memories']) + "\n\n"
 		# this is similar to customizing the create_react_agent with state_modifier, but is a lot more flexible
 		system_prompt = SystemMessage(

ex_app/lib/all_tools/calendar.py

@@ -165,7 +165,7 @@ def find_free_time_slot_in_calendar_sync(participants: list[str], slot_duration:
 																	   headers={
 																		   "Content-Type": "text/calendar; charset=utf-8",
 																		   "Depth": "0",
-																	   }, content=freebusyRequest)
+																	   }, data=freebusyRequest)
 			print(freebusyRequest)
 			print(response.text)
 

ex_app/lib/all_tools/contacts.py

@@ -70,7 +70,7 @@ async def find_person_in_contacts(name: str) -> list[dict[str, typing.Any]]:
 			response = await nc._session._create_adapter(True).request('REPORT', f"{nc.app_cfg.endpoint}{link}", headers={
 				"Content-Type": "application/xml; charset=utf-8",
 				"Depth": "1",
-			}, content=xml_body)
+			}, data=xml_body)
 
 			if response.status_code != 207:  # Multi-Status
 				raise Exception(f"Error: {response.status_code} - {response.reason_phrase}")

ex_app/lib/all_tools/files.py

@@ -15,7 +15,7 @@ async def get_tools(nc: AsyncNextcloudApp):
 	@safe_tool
 	async def get_file_content(file_path: str):
 		"""
-		Get the content of a file
+		Get the content of a nextcloud-internal file of the current user
 		:param file_path: the path of the file
 		:return:
 		"""
@@ -33,7 +33,7 @@ async def get_file_content(file_path: str):
 	async def get_file_content_by_file_link(file_url: str):
 		"""
 		Get the content of a file given an internal Nextcloud link (e.g., https://host/index.php/f/12345)
-		:param file_url: the internal file URL
+		:param file_url: the nextcloud-internal file URL
 		:return: text content of the file
 		"""
 

ex_app/lib/all_tools/web.py

@@ -0,0 +1,31 @@
+# SPDX-FileCopyrightText: 2025 Nextcloud GmbH and Nextcloud contributors
+# SPDX-License-Identifier: AGPL-3.0-or-later
+import niquests
+from langchain_core.tools import tool
+from nc_py_api import AsyncNextcloudApp
+
+from ex_app.lib.all_tools.lib.decorator import safe_tool
+
+
+async def get_tools(nc: AsyncNextcloudApp):
+
+	@tool
+	@safe_tool
+	async def web_fetch(url: str) -> str:
+		"""
+		Get the contents of a web page via HTTP
+		:param url: The HTTP URL to the web page (e.g. https://nextcloud.com/team/ )
+		:return: the web page content
+		"""
+		res = await niquests.get(url)
+		return res.text()
+
+	return [
+		web_fetch,
+	]
+
+def get_category_name():
+	return "Web access"
+
+async def is_available(nc: AsyncNextcloudApp):
+	return True

ex_app/lib/nc_model.py

@@ -37,7 +37,7 @@ class ChatWithNextcloud(BaseChatModel):
 	tools: Sequence[
 		Union[typing.Dict[str, Any], type, Callable, BaseTool]] = []
 	TIMEOUT: int = 60 * 30 # 30 minutes
-	MAX_MESSAGE_HISTORY: int = 13
+	MAX_MESSAGE_HISTORY: int = 42
 
 	def _generate(self, messages: list[BaseMessage], stop: Optional[list[str]] = None, run_manager: Optional[CallbackManagerForLLMRun] = None, **kwargs: Any):
 		raise Exception("Use _agenerate instead")