RestApiTool crashes the agent run on a non-UTF-8 error response body (UnicodeDecodeError in the HTTPStatusError handler)

[1aifanatic]

🔴 Required Information

Describe the Bug:

When an API call fails, RestApiTool.call() builds the error message for the model with:

except httpx.HTTPStatusError:
  error_details = response.content.decode("utf-8")

That forces UTF-8 regardless of the charset the server declares. A non-UTF-8 error body raises UnicodeDecodeError, for example a Latin-1/Windows-1252 page from an older ERP or IIS server, a message with accented characters, or a binary body. The exception is raised inside the except handler, so the sibling except ValueError can't catch it. It propagates out of the tool and ends the whole agent run, so the model never sees the HTTP error it could have recovered from.

Steps to Reproduce:

1. pip install google-adk (reproduced on main @ 3f4bb8f).
2. Run the script below: a real Runner plus LlmAgent plus OpenAPIToolset, with a mocked model and an httpx.MockTransport backend that answers 404 with a text/plain; charset=iso-8859-1 body.

Expected Behavior:
The tool returns its usual error dict, decoded with the declared charset:
... Execution Error: Status Code: 404, Commande introuvable : échec

Observed Behavior:

RUN FAILED: UnicodeDecodeError - 'utf-8' codec can't decode byte 0xe9 in position 23: invalid continuation byte

Environment Details:

• ADK Library Version (pip show google-adk): main @ 3f4bb8f
• Desktop OS: Windows 11
• Python Version (python -V): 3.12.10

Model Information:

• Are you using LiteLLM: No
• Which model is being used: N/A (mocked model; the bug is in response handling)

---

🟡 Optional Information

Minimal Reproduction Code:

import asyncio

import httpx
from google.adk.agents.llm_agent import LlmAgent
from google.adk.models.base_llm import BaseLlm
from google.adk.models.llm_response import LlmResponse
from google.adk.runners import InMemoryRunner
from google.adk.tools.openapi_tool.openapi_spec_parser.openapi_toolset import OpenAPIToolset
from google.genai import types

spec = {
    "openapi": "3.0.0",
    "info": {"title": "Legacy ERP", "version": "1"},
    "servers": [{"url": "https://erp.example.com"}],
    "paths": {"/orders/{id}": {"get": {
        "operationId": "getOrder",
        "parameters": [{"name": "id", "in": "path", "required": True, "schema": {"type": "string"}}],
        "responses": {"200": {"description": "ok"}},
    }}},
}


def handler(request: httpx.Request):
  # A legacy server answering in Windows-1252 / Latin-1, as it declares.
  return httpx.Response(
      404,
      content="Commande introuvable : échec".encode("latin-1"),
      headers={"content-type": "text/plain; charset=iso-8859-1"},
  )


class FakeModel(BaseLlm):
  model: str = "fake"
  calls: int = 0

  async def generate_content_async(self, llm_request, stream=False):
    self.calls += 1
    part = (types.Part.from_function_call(name="get_order", args={"id": "A-42"})
            if self.calls == 1 else types.Part.from_text(text="done"))
    yield LlmResponse(content=types.Content(role="model", parts=[part]))


agent = LlmAgent(
    name="erp_agent",
    model=FakeModel(),
    tools=[OpenAPIToolset(
        spec_dict=spec,
        httpx_client_factory=lambda: httpx.AsyncClient(transport=httpx.MockTransport(handler)),
    )],
)


async def main():
  runner = InMemoryRunner(agent=agent)
  session = await runner.session_service.create_session(app_name=runner.app_name, user_id="u")
  try:
    async for event in runner.run_async(
        user_id="u", session_id=session.id,
        new_message=types.Content(role="user", parts=[types.Part(text="order A-42?")])):
      for fr in event.get_function_responses():
        print("function_response:", fr.response)
  except Exception as e:
    print("RUN FAILED:", type(e).__name__, "-", e)


asyncio.run(main())

Suggested fix: use response.text, which is what the non-JSON success path a few lines below already uses. It decodes with the response's declared charset and replaces undecodable bytes instead of raising. I have a PR ready with a test.

How often has this issue occurred?:

• Always (100%): any non-UTF-8 error body.

[chelsealong]

Picking this one up now — opening a PR shortly. Flagging it here so nobody duplicates the work; if someone is already on it, say so and I will drop mine.