weave auto-patch breaks OpenAIGenerators streaming

[tstadel]

Describe the bug
weave wraps OpenAI's Stream in a way that breaks class inheritance which is why our check for isinstance(response, Stream) fails which leads to an error for anyone who wants to use streaming in OpenAIChatGenerator or OpenAIGenerator together with weave. We ran into this in deepset Studio.

Although it's in general bad practice by the tracing lib providers to wrap well-know API types such as OpenAI's Stream, we've learned that more than one of these libs make use of this pattern anyways. Therefore I'd suggest to part from the type checks in question in Haystack alltogether.

Responsible Code:

haystack/haystack/components/generators/chat/openai.py

Lines 258 to 259 in 195d403

	is_streaming = isinstance(chat_completion, Stream)
	assert is_streaming or streaming_callback is None

haystack/haystack/components/generators/openai.py

Line 220 in 195d403

if isinstance(completion, Stream):

Error message

Exception has occurred: AssertionError       (note: full exception trace is shown but execution is paused at: <module>)
exception: no description
  File "/home/haystackd/.local/lib/python3.12/site-packages/haystack/components/generators/chat/openai.py", line 251, in run
    assert is_streaming or streaming_callback is None
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/haystackd/.local/lib/python3.12/site-packages/haystack/core/pipeline/pipeline.py", line 80, in _run_component
    component_output = instance.run(**component_inputs)
                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/haystackd/.local/lib/python3.12/site-packages/haystack/core/pipeline/pipeline.py", line 248, in run
    component_outputs = self._run_component(component, inputs, component_visits, parent_span=span)
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/haystackd/debug/debug_pipeline.py", line 17, in <module> (Current frame)
    result = pipeline.run(data={"query": query, "text": query, "question": query, "streaming_callback": lambda x: print(x.content, end="")})
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/runpy.py", line 88, in _run_code
    exec(code, run_globals)
  File "/usr/local/lib/python3.12/runpy.py", line 198, in _run_module_as_main
    return _run_code(code, main_globals, None,
AssertionError: 

(Stacktrace produced with Haystack 2.10.3, but current main has the same issue)

Expected behavior
Streaming of OpenAIGenerators works with tracing-libs such as weave.

Additional context
We've seen issues with type checks on tracing-lib-wrapped responses before (See deepset-ai/haystack-core-integrations#1454)

To Reproduce

1. Run weave.init(project_name) using default settings
2. Create a pipeline using OpenAIChatGenerator with streaming_callback set.
3. Run pipeline

FAQ Check

• Have you had a look at our new FAQ page?

System:

• OS:
• GPU/CPU:
• Haystack version (commit or version number): 2.10.3 (current main has the same issue)
• DocumentStore:
• Reader:
• Retriever: