fix: remove keywords from wrap_function_wrapper

[JP-Ellis]

The upstream wrapt library renamed the module keyword argument to target, though has kept the ordering the same. I have opted to remove the keywords entirely so that it is compatible with both recent and older versions of wrapt.

Fixes: #4009

• I have added tests that cover my changes.

  Existing tests should cover this

• If adding a new instrumentation or changing an existing one, I've added screenshots from some observability platform showing the change.

  N/A

• PR name follows conventional commits format: feat(instrumentation): ... or fix(instrumentation): ....

• (If applicable) I have updated the documentation accordingly.

  N/A

Summary by CodeRabbit

• Refactor
  • Updated OpenTelemetry instrumentation patterns for Agno and Langchain integrations to improve code consistency in function wrapper invocations.

[CLAassistant]

All committers have signed the CLA.

[coderabbitai]

📝 Walkthrough

Walkthrough

Two OpenTelemetry instrumentation modules were updated to correct their wrap_function_wrapper function call signatures. Keyword arguments (module=, wrapper=) were converted to positional arguments to match the actual function API, resolving a TypeError encountered during auto-instrumentation.

Changes

Cohort / File(s)	Summary
Agno Instrumentation API Signature Fix packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/__init__.py	Refactored six wrap_function_wrapper calls for Agent.run, Agent.arun, Team.run, Team.arun, FunctionCall.execute, and FunctionCall.aexecute by converting module= keyword argument to positional first parameter and wrapper= keyword argument to positional third parameter.
LangChain Instrumentation API Signature Fix packages/opentelemetry-instrumentation-langchain/opentelemetry/instrumentation/langchain/__init__.py	Refactored multiple wrap_function_wrapper calls across BaseCallbackManager.__init__, OpenAI LLM methods (BaseOpenAI, BaseChatOpenAI), LangGraph methods (Pregel.stream, Pregel.astream, Command.__init__), agent factory functions (create_react_agent, create_agent), and middleware hooks by removing wrapper= keyword argument and passing wrapper objects as positional third parameters.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Poem

🐰 Hop hop, the code now flows so right,
No keyword= confusion blocking the sight,
Positional args march in their place,
LangChain and Agno: instrumentation race!
TypeError banished, the rabbit did cheer,
Auto-instrument works—hooray, hooray, dear! 🥕✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 50.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The PR title clearly and concisely describes the main change: removing keyword arguments from wrap_function_wrapper calls to fix compatibility issues across wrapt versions.
Linked Issues check	✅ Passed	The PR directly addresses issue #4009 by converting wrap_function_wrapper calls from keyword arguments (module=, wrapper=) to positional arguments, which resolves the TypeError and ensures compatibility across wrapt versions.
Out of Scope Changes check	✅ Passed	All changes are directly scoped to fixing the wrap_function_wrapper API mismatch in the agno and langchain instrumentors; no unrelated modifications are present.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In
`@packages/opentelemetry-instrumentation-langchain/opentelemetry/instrumentation/langchain/__init__.py`:
- Around line 169-177: The uninstrument() implementation currently unwraps
"langchain_openai.chat_models.base" targets for
"BaseOpenAI._generate"/"_agenerate" while the instrumentation wraps
"BaseChatOpenAI._generate"/"_agenerate", leaving chat wrappers installed; update
the unwrapping calls to target the same symbols used during
wrapping—specifically change any unwrap/remove_wrapper references that mention
"BaseOpenAI._generate" and "BaseOpenAI._agenerate" to "BaseChatOpenAI._generate"
and "BaseChatOpenAI._agenerate" (also apply the same change in the other
occurrence around lines ~352-360) so that wrap_function_wrapper and the
corresponding uninstrument() use identical class/method names.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 05dc83f7-1430-4979-92ce-8629f550e3bf

📥 Commits

Reviewing files that changed from the base of the PR and between dcd5cb0 and df8ce55.

📒 Files selected for processing (2)

• packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/__init__.py
• packages/opentelemetry-instrumentation-langchain/opentelemetry/instrumentation/langchain/__init__.py

[coderabbitai]

⚠️ Potential issue | 🟡 Minor

Align BaseChatOpenAI unwrapping with the wrapped targets.

These calls wrap BaseChatOpenAI._generate and _agenerate, but _uninstrument() still unwraps BaseOpenAI... in the same module, so chat wrappers can remain installed after uninstrumenting.

🛠️ Proposed fix

-                unwrap("langchain_openai.chat_models.base", "BaseOpenAI._generate")
-                unwrap("langchain_openai.chat_models.base", "BaseOpenAI._agenerate")
-                # unwrap("langchain_openai.chat_models.base", "BaseOpenAI._stream")
-                # unwrap("langchain_openai.chat_models.base", "BaseOpenAI._astream")
+                unwrap("langchain_openai.chat_models.base", "BaseChatOpenAI._generate")
+                unwrap("langchain_openai.chat_models.base", "BaseChatOpenAI._agenerate")
+                # unwrap("langchain_openai.chat_models.base", "BaseChatOpenAI._stream")
+                # unwrap("langchain_openai.chat_models.base", "BaseChatOpenAI._astream")

Also applies to: 352-360

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In
`@packages/opentelemetry-instrumentation-langchain/opentelemetry/instrumentation/langchain/__init__.py`
around lines 169 - 177, The uninstrument() implementation currently unwraps
"langchain_openai.chat_models.base" targets for
"BaseOpenAI._generate"/"_agenerate" while the instrumentation wraps
"BaseChatOpenAI._generate"/"_agenerate", leaving chat wrappers installed; update
the unwrapping calls to target the same symbols used during
wrapping—specifically change any unwrap/remove_wrapper references that mention
"BaseOpenAI._generate" and "BaseOpenAI._agenerate" to "BaseChatOpenAI._generate"
and "BaseChatOpenAI._agenerate" (also apply the same change in the other
occurrence around lines ~352-360) so that wrap_function_wrapper and the
corresponding uninstrument() use identical class/method names.

[JP-Ellis]

Not sure exactly what it means...

[coderabbitai]

@JP-Ellis — happy to clarify! Here's the concrete problem:

During instrumentation, the code correctly wraps methods on BaseChatOpenAI inside langchain_openai.chat_models.base:

# _instrument  ✅ correct class
wrap_function_wrapper("langchain_openai.chat_models.base", "BaseChatOpenAI._generate", ...)
wrap_function_wrapper("langchain_openai.chat_models.base", "BaseChatOpenAI._agenerate", ...)

But during uninstrumentation, the unwrap calls reference a different class (BaseOpenAI) in that same module:

# _uninstrument  ❌ wrong class — BaseOpenAI lives in llms.base, not chat_models.base
unwrap("langchain_openai.chat_models.base", "BaseOpenAI._generate")
unwrap("langchain_openai.chat_models.base", "BaseOpenAI._agenerate")

BaseOpenAI and BaseChatOpenAI are two separate classes. The chat_models.base module contains BaseChatOpenAI, so these unwrap calls either silently no-op or target the wrong symbol entirely — meaning the wrappers placed on BaseChatOpenAI._generate and BaseChatOpenAI._agenerate are never removed when _uninstrument() is called.

The fix is straightforward — just change the class name in those two unwrap calls:

-                unwrap("langchain_openai.chat_models.base", "BaseOpenAI._generate")
-                unwrap("langchain_openai.chat_models.base", "BaseOpenAI._agenerate")
-                # unwrap("langchain_openai.chat_models.base", "BaseOpenAI._stream")
-                # unwrap("langchain_openai.chat_models.base", "BaseOpenAI._astream")
+                unwrap("langchain_openai.chat_models.base", "BaseChatOpenAI._generate")
+                unwrap("langchain_openai.chat_models.base", "BaseChatOpenAI._agenerate")
+                # unwrap("langchain_openai.chat_models.base", "BaseChatOpenAI._stream")
+                # unwrap("langchain_openai.chat_models.base", "BaseChatOpenAI._astream")

This is a pre-existing bug (not introduced by this PR), but it's worth fixing here since the area is already being touched.

---

🧠 Learnings used

Learnt from: duanyutong
Repo: traceloop/openllmetry PR: 3487
File: packages/opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/utils.py:177-178
Timestamp: 2025-12-02T21:09:48.690Z
Learning: The opentelemetry-instrumentation-openai and opentelemetry-instrumentation-openai-agents packages must remain independent and not share code, so code duplication between them is acceptable.

---

Learnt from: avivhalfon
Repo: traceloop/openllmetry PR: 3840
File: packages/opentelemetry-instrumentation-google-generativeai/opentelemetry/instrumentation/google_generativeai/__init__.py:206-209
Timestamp: 2026-03-23T09:16:59.824Z
Learning: In the OpenTelemetry GenAI semantic conventions, `gen_ai.operation.name` uses standard predefined values such as "chat", "text_completion", "embeddings", etc. — NOT method-specific names (e.g., not "generate_content"). The Gemini/google-generativeai and Vertex AI instrumentations in traceloop/openllmetry should use "chat" as the operation name for generate_content calls, consistent with how the LangChain instrumentation (PR `#3836`) also uses "chat".