chore: Release 1.7.11 (#4217)

# Changelog

## New Features:

- **In Memory Storage**: Added `InMemoryStorage` for a simpler storage
option useful in rapid prototyping.
- **Trafilatura Tools**: Added `TrafilaturaTools` SDK for web scraping
and text extraction using the Trafilatura library, with 15 comprehensive
cookbook examples.
- **Qwen and Dashscope Native Models**: Use Qwen models via `DashScope`
integration.

## Improvements:

- **Async Support for Milvus Hybrid Search**
- **Bedrock Files Support**: Now supporting `File` for compatible AWS
Bedrock models.

## Bug Fixes:

- **MarkdownKnowledgeBase Custom Chunking Strategy**: Fixed issues with
setting a custom chunking strategy.

Author: dirkbrnd

cookbook/agent_concepts/user_control_flows/external_tool_execution_async_responses.py

@@ -27,7 +27,12 @@ def execute_shell_command(command: str) -> str:
     Returns:
         str: The output of the shell command
     """
-    if command.startswith("ls ") or command == "ls" or command.startswith("cat ") or command.startswith("head "):
+    if (
+        command.startswith("ls ")
+        or command == "ls"
+        or command.startswith("cat ")
+        or command.startswith("head ")
+    ):
         return subprocess.check_output(command, shell=True).decode("utf-8")
     raise Exception(f"Unsupported command: {command}")
 
@@ -41,7 +46,9 @@ def execute_shell_command(command: str) -> str:
 run_response = asyncio.run(agent.arun("What files do I have in my current directory?"))
 
 # Keep executing externally-required tools until the run completes
-while run_response.is_paused and len(run_response.tools_awaiting_external_execution) > 0:
+while (
+    run_response.is_paused and len(run_response.tools_awaiting_external_execution) > 0
+):
     for tool in run_response.tools_awaiting_external_execution:
         if tool.tool_name == execute_shell_command.name:
             print(f"Executing {tool.tool_name} with args {tool.tool_args} externally")
@@ -57,5 +64,3 @@ def execute_shell_command(command: str) -> str:
 
 # Or for simple debug flow
 # agent.print_response("What files do I have in my current directory?")
-
-

libs/agno/agno/models/aws/bedrock.py

@@ -31,6 +31,7 @@
 BEDROCK_SUPPORTED_VIDEO_FORMATS = ["mp4", "mov", "mkv", "webm", "flv", "mpeg", "mpg", "wmv", "three_gp"]
 BEDROCK_SUPPORTED_FILE_FORMATS = ["pdf", "csv", "doc", "docx", "xls", "xlsx", "html", "txt", "md"]
 
+
 @dataclass
 class AwsBedrock(Model):
     """

libs/agno/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "agno"
-version = "1.7.10"
+version = "1.7.11"
 description = "Agno: a lightweight library for building Multi-Agent Systems"
 requires-python = ">=3.7,<4"
 readme = "README.md"

libs/agno/tests/integration/models/anthropic/test_tool_use.py

@@ -255,5 +255,5 @@ def test_tool_call_list_parameters():
             tool_calls.extend(msg.tool_calls)
     for call in tool_calls:
         if call.get("type", "") == "function":
-            assert call["function"]["name"] in ["get_contents", "exa_answer", "search_exa"]
+            assert call["function"]["name"] in ["get_contents", "exa_answer", "search_exa", "find_similar"]
     assert response.content is not None

libs/agno/tests/integration/models/openai/responses/test_tool_use.py

@@ -5,7 +5,6 @@
 
 from agno.agent import Agent, RunResponse  # noqa
 from agno.models.openai import OpenAIResponses
-from agno.tools.duckduckgo import DuckDuckGoTools
 from agno.tools.exa import ExaTools
 from agno.tools.yfinance import YFinanceTools
 
@@ -153,22 +152,29 @@ def test_parallel_tool_calls():
 
 def test_multiple_tool_calls():
     """Test multiple different tool types with the responses API."""
+
+    def get_the_weather(city: str):
+        return f"It is currently 70 degrees and cloudy in {city}"
+
+    def get_favourite_city():
+        return "Tokyo"
+
     agent = Agent(
         model=OpenAIResponses(id="gpt-4.1-mini"),
-        tools=[YFinanceTools(cache_results=True), DuckDuckGoTools(cache_results=True)],
+        tools=[get_the_weather, get_favourite_city],
         markdown=True,
         telemetry=False,
         monitoring=False,
     )
 
-    response = agent.run("Find the latest price of TSLA. Then, based on the price, find the latest news about it.")
+    response = agent.run("Find my favourite city. Then, get the weather in that city.")
 
     # Verify tool usage
     tool_calls = [msg.tool_calls for msg in response.messages if msg.tool_calls]
-    assert len(tool_calls) >= 2  # At least two messages have tool calls
-    assert sum(len(calls) for calls in tool_calls) >= 2  # Total of 2 tool calls made
+    assert len(tool_calls) >= 2  # At least one message has tool calls
+    assert sum(len(calls) for calls in tool_calls) >= 1  # Total of 1 tool calls made
     assert response.content is not None
-    assert "TSLA" in response.content and "latest news" in response.content.lower()
+    assert "Tokyo" in response.content and "70" in response.content
 
 
 def test_tool_call_custom_tool_no_parameters():

libs/agno/tests/integration/models/xai/test_basic.py

@@ -29,17 +29,15 @@ def test_basic():
 
 
 def test_basic_stream():
-    agent = Agent(model=xAI(id="grok-3-mini-fast"), markdown=True, telemetry=False, monitoring=False)
+    agent = Agent(model=xAI(id="grok-3-mini"), markdown=True, telemetry=False, monitoring=False)
 
     response_stream = agent.run("Share a 2 sentence horror story", stream=True)
 
     # Verify it's an iterator
     assert hasattr(response_stream, "__iter__")
 
-    responses = list(response_stream)
-    assert len(responses) > 0
-    for response in responses:
-        assert response.content is not None
+    for response in response_stream:
+        assert response.content is not None or response.reasoning_content is not None
 
     _assert_metrics(agent.run_response)
 
@@ -63,7 +61,7 @@ async def test_async_basic_stream():
     response_stream = await agent.arun("Share a 2 sentence horror story", stream=True)
 
     async for response in response_stream:
-        assert response.content is not None
+        assert response.content is not None or response.reasoning_content is not None
 
     _assert_metrics(agent.run_response)
 

libs/agno/tests/unit/models/openai/test_openai_responses_id_handling.py

@@ -1,9 +1,9 @@
 from typing import Any, Dict, List, Optional
 
+from agno.models.base import MessageData
 from agno.models.message import Message
 from agno.models.openai.responses import OpenAIResponses
 from agno.models.response import ModelResponse
-from agno.models.base import MessageData
 
 
 class _FakeError:
@@ -47,7 +47,15 @@ def __init__(self, *, _id: str, call_id: Optional[str], name: str, arguments: st
 
 
 class _FakeStreamEvent:
-    def __init__(self, *, type: str, item: Optional[_FakeStreamItem] = None, delta: str = "", response: Any = None, annotation: Any = None):
+    def __init__(
+        self,
+        *,
+        type: str,
+        item: Optional[_FakeStreamItem] = None,
+        delta: str = "",
+        response: Any = None,
+        annotation: Any = None,
+    ):
         self.type = type
         self.item = item
         self.delta = delta
@@ -66,15 +74,22 @@ def test_format_messages_maps_tool_output_fc_to_call_id():
                 "id": "fc_abc123",
                 "call_id": "call_def456",
                 "type": "function",
-                "function": {"name": "execute_shell_command", "arguments": "{\"command\": \"ls -la\"}"},
+                "function": {"name": "execute_shell_command", "arguments": '{"command": "ls -la"}'},
             }
         ],
     )
 
     # Tool output referring to the fc_* id should be normalized to call_*
     tool_output = Message(role="tool", tool_call_id="fc_abc123", content="ok")
 
-    fm = model._format_messages(messages=[Message(role="system", content="s"), Message(role="user", content="u"), assistant_with_tool_call, tool_output])
+    fm = model._format_messages(
+        messages=[
+            Message(role="system", content="s"),
+            Message(role="user", content="u"),
+            assistant_with_tool_call,
+            tool_output,
+        ]
+    )
 
     # Expect one function_call and one function_call_output normalized
     fc_items = [x for x in fm if x.get("type") == "function_call"]
@@ -114,7 +129,10 @@ def test_process_stream_response_builds_tool_calls():
     stream_data = MessageData()
 
     # Simulate function_call added and then completed
-    added = _FakeStreamEvent(type="response.output_item.added", item=_FakeStreamItem(_id="fc_abc123", call_id="call_def456", name="execute", arguments="{}"))
+    added = _FakeStreamEvent(
+        type="response.output_item.added",
+        item=_FakeStreamItem(_id="fc_abc123", call_id="call_def456", name="execute", arguments="{}"),
+    )
     mr, tool_use = model._process_stream_response(added, assistant_message, stream_data, {})
     assert mr is None
 
@@ -155,9 +173,14 @@ def test_reasoning_previous_response_skips_prior_function_call_items(monkeypatch
         ],
     )
 
-    fm = model._format_messages(messages=[Message(role="system", content="s"), Message(role="user", content="u"), assistant_with_prev, assistant_with_tool_call])
+    fm = model._format_messages(
+        messages=[
+            Message(role="system", content="s"),
+            Message(role="user", content="u"),
+            assistant_with_prev,
+            assistant_with_tool_call,
+        ]
+    )
 
     # Expect no re-sent function_call when previous_response_id is present for reasoning models
     assert all(x.get("type") != "function_call" for x in fm)
-
-