OpenAI and Azure Responses Chat Generators fail to deserialize ComponentTool

Describe the bug

OpenAIResponsesChatGenerator.from_dict() and AzureOpenAIResponsesChatGenerator.from_dict() fail to deserialize ComponentTool objects.

Because from_dict() hardcodes a check for "haystack.tools.tool.Tool", ComponentTool (which serializes with "type": "haystack.tools.component_tool.ComponentTool") is skipped and left as an un-deserialized dictionary ({'type': '...', 'data': ...}).

To Reproduce

from haystack.components.builders import PromptBuilder
from haystack.components.generators.chat.openai_responses import OpenAIResponsesChatGenerator
from haystack.tools import ComponentTool

tool = ComponentTool(
    name="prompt",
    description="Builds prompt",
    component=PromptBuilder(template="hi"),
)

generator = OpenAIResponsesChatGenerator(tools=[tool])
data = generator.to_dict()

restored = OpenAIResponsesChatGenerator.from_dict(data)

# Fails: restored.tools[0] is <class 'dict'>, not ComponentTool
assert isinstance(restored.tools[0], ComponentTool)

Expected behavior

from_dict() should deserialize ComponentTool back into a ComponentTool instance, matching the behavior of OpenAIChatGenerator.from_dict().

Root Cause

In haystack/components/generators/chat/openai_responses.py (lines 367–374) and haystack/components/generators/chat/azure_responses.py (lines 259–266):

tools = data["init_parameters"].get("tools")
if tools and (
    isinstance(tools, dict)
    and tools.get("type") == "haystack.tools.toolset.Toolset"
    or isinstance(tools, list)
    and tools[0].get("type") == "haystack.tools.tool.Tool"
):
    deserialize_tools_or_toolset_inplace(data["init_parameters"], key="tools")

The check tools[0].get("type") == "haystack.tools.tool.Tool" fails for ComponentTool, skipping deserialize_tools_or_toolset_inplace().

Fix

Check for the presence of "data" to identify serialized Haystack tools:

tools = data.get("init_parameters", {}).get("tools")
if tools and (
    isinstance(tools, dict)
    and "type" in tools
    and "data" in tools
    or isinstance(tools, list)
    and isinstance(tools[0], dict)
    and "type" in tools[0]
    and "data" in tools[0]
):
    deserialize_tools_or_toolset_inplace(data["init_parameters"], key="tools")

Additional context

• OS: Any
• Haystack version: 2.31.0 (main)
• Components: OpenAIResponsesChatGenerator and AzureOpenAIResponsesChatGenerator