[gjija]

Please ensure you have read the contribution guide before creating a pull request.

Link to Issue or Description of Change

1. Link to an existing issue (if applicable):

• Closes: N/A
• Related: N/A

2. Or, if no issue exists, describe the change:

Problem:

LangGraphAgent._run_async_impl takes final_state['messages'][-1].content and passes it straight to types.Part.from_text(text=...). In LangChain, BaseMessage.content is str | list[str | dict]. Chat models return the list form in common setups: Gemini 3 models through ChatGoogleGenerativeAI (langchain-google-genai 4.x returns text blocks for them; langchain-ai/langchain-google#1460 shows the streamed form), Vertex with a thinking budget (langchain-ai/langchain-google#1621), and any model set to output_version="v1". Then Part.from_text fails validation and the agent raises on every turn instead of answering:

pydantic_core._pydantic_core.ValidationError: 1 validation error for Part
text
  Input should be a valid string [type=string_type, input_value=[{'type': 'reasoning', 'r...'text': 'Sunny today.'}], input_type=list]

Repro (a real compiled graph through InMemoryRunner; the node returns what such a model returns):

import asyncio

from google.adk.agents.langgraph_agent import LangGraphAgent
from google.adk.runners import InMemoryRunner
from google.genai import types
from langchain_core.messages import AIMessage
from langgraph.graph import END, START, MessagesState, StateGraph


def reply(state):
  blocks = [
      {"type": "reasoning", "reasoning": "Checking the forecast."},
      {"type": "text", "text": "Sunny today."},
  ]
  return {"messages": [AIMessage(content=blocks)]}


async def main():
  builder = StateGraph(MessagesState)
  builder.add_node("reply", reply)
  builder.add_edge(START, "reply")
  builder.add_edge("reply", END)
  agent = LangGraphAgent(name="weather", graph=builder.compile())
  runner = InMemoryRunner(agent=agent, app_name="app")
  session = await runner.session_service.create_session(app_name="app", user_id="u")
  async for event in runner.run_async(
      user_id="u", session_id=session.id, new_message=types.UserContent("Weather?")
  ):
    print("reply:", event.content.parts[0].text)


asyncio.run(main())

On main it raises the ValidationError above.

Solution:

When the content is a list, use the message's .text property. It keeps plain strings and type == "text" blocks and joins them into a str. String content is passed through as before, so graphs that work today are unchanged.

.text is a property on every langchain-core 1.x, which ADK's langgraph>=1.0.10 floor requires. On langchain-core 0.3, which installs from when ADK required langgraph 0.x can still have, .text is a method; string content never reaches it there, so those installs keep working.

Non-text blocks (reasoning, images) are left out of the reply. Mapping reasoning blocks to ADK thought parts would be a separate change.

With this change the repro prints:

reply: Sunny today.

Testing Plan

Unit Tests:

• I have added or updated unit tests for my change.
• All unit tests pass locally.

New test test_langgraph_agent_reads_text_from_content_blocks in tests/unittests/agents/test_langgraph_agent.py, next to test_langgraph_agent_runs_real_compiled_state_graph. It runs a real compiled StateGraph(MessagesState) whose node returns a reasoning block and a text block.

Without the source change it fails with the ValidationError above (1 failed, 10 passed). With it:

$ pytest tests/unittests/agents/test_langgraph_agent.py
11 passed

The whole tests/unittests/agents folder: 975 passed, 1 xfailed (974 passed on main). I also ran the file on langgraph 0.4.8 with langchain-core 0.3.79: the 10 existing tests pass with the change, as on main; the new list test fails there on both, since 0.3 has no .text property. pyink and isort leave both files unchanged.

Manual End-to-End (E2E) Tests:

The repro above stands in for a manual run: a real compiled LangGraph graph through InMemoryRunner, with the node returning the block content a Gemini 3 or output_version="v1" model gives, in place of a live model call. Versions: langgraph 1.2.20, langchain-core 1.7.6.

Checklist

• I have read the CONTRIBUTING.md document.
• I have performed a self-review of my own code.
• I have commented my code, particularly in hard-to-understand areas.
• I have added tests that prove my fix is effective or that my feature works.
• New and existing unit tests pass locally with my changes.
• I have manually tested my changes end-to-end.
• Any dependent changes have been merged and published in downstream modules.

Additional context

A list with no text blocks gives an empty text part instead of a crash.