RestApiTool ignores query parameter style/explode: object query params are sent as a Python dict repr

Aperta
#7,204 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
3/5
Tempo stimato
1-2 giorni
Idoneità per principianti
76/100
Tipo di issue
Bug
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
openapi, python

Direzione di ricerca

Start at RestApiTool's query branch and trace how Parameter values become ApiParameter instances in OpenAPIToolset. Compare the request-building path with the OpenAPI style/explode cases listed in the reproduction, then use focused OpenAPIToolset tests. Done means object and array query parameters produce the expected form, deepObject, delimited, and default encodings without changing scalar behavior.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

🔴 Required Information

Describe the Bug:

RestApiTool ignores a query parameter's OpenAPI style and explode. It passes the model's value straight to httpx:

elif param_location == "query":
  if v is not None:
    query_params[original_k] = v

httpx repeats the key for a list, which happens to match the default (style: form, explode: true) for arrays. But it serializes a dict as its Python repr, so any object-typed query parameter goes out as filter={'status': 'open'}. That's wrong for every style. Non-exploded and delimited arrays (explode: false, spaceDelimited, pipeDelimited) are also sent exploded. ApiParameter doesn't record style/explode at all, so the spec's instructions are dropped at parse time.

Object query parameters are common in enterprise APIs: search/filter objects, JSON:API filter[...], metadata[...]-style deepObject params. The API then either ignores the filter, which silently returns unfiltered data to the agent, or rejects the request.

Steps to Reproduce:

  1. pip install google-adk (reproduced on main @ 3f4bb8fa).
  2. Run the script below: a real Runner plus LlmAgent plus OpenAPIToolset, a mocked model, and an httpx.MockTransport backend.

Expected Behavior:
Per the OpenAPI 3 serialization rules: status=open&priority=P1&filter[assignee]=me&fields=id,title.

Observed Behavior:

query sent: match={'status':+'open',+'priority':+'P1'}&filter={'assignee':+'me'}&fields=id&fields=title

Environment Details:

  • ADK Library Version (pip show google-adk): main @ 3f4bb8fa
  • 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 request building)

🟡 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": "Tickets", "version": "1"},
    "servers": [{"url": "https://itsm.example.com"}],
    "paths": {"/tickets": {"get": {
        "operationId": "listTickets",
        "parameters": [
            # style: form, explode: true (the OpenAPI default for query)
            {"name": "match", "in": "query",
             "schema": {"type": "object", "properties": {"status": {"type": "string"}, "priority": {"type": "string"}}}},
            {"name": "filter", "in": "query", "style": "deepObject", "explode": True,
             "schema": {"type": "object", "properties": {"assignee": {"type": "string"}}}},
            {"name": "fields", "in": "query", "style": "form", "explode": False,
             "schema": {"type": "array", "items": {"type": "string"}}},
        ],
        "responses": {"200": {"description": "ok"}},
    }}},
}
sent = []


def handler(request: httpx.Request):
  sent.append(request.url)
  return httpx.Response(200, json={"tickets": []})


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="list_tickets", args={
        "match": {"status": "open", "priority": "P1"},
        "filter": {"assignee": "me"},
        "fields": ["id", "title"],
    }) if self.calls == 1 else types.Part.from_text(text="done"))
    yield LlmResponse(content=types.Content(role="model", parts=[part]))


agent = LlmAgent(
    name="itsm_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")
  async for _ in runner.run_async(
      user_id="u", session_id=session.id,
      new_message=types.Content(role="user", parts=[types.Part(text="my open P1 tickets")])):
    pass
  from urllib.parse import unquote
  print("query sent:", unquote(sent[0].query.decode()))


asyncio.run(main())

Suggested fix: carry style/explode from the spec's Parameter into ApiParameter (optional fields, so serialized parameters still load), and serialize query values accordingly:

  • objects are spread (form, exploded, the default), bracketed (deepObject), or k,v,... (form, not exploded);
  • non-exploded arrays are joined with ,, , or |;
  • scalars and default arrays are unchanged.

I have a PR ready with tests.

How often has this issue occurred?:

  • Always (100%): any object-typed or non-exploded query parameter.
Lingua principale
Python
Stelle
21.6k
Fork
4k
Merge medio
13h 49m
PR unite (30g)
10

Guida per i contributori

Apri la guida per i contributori

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di google/adk-python

Tutte le issue di google/adk-python

Issue simili

Altre issue su Python

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.