Skip to content

contents rearrange still raises on id-less / unmatched trailing function responses (follow-up to orphan FR prune) #6751

Closed

Description

@a2105z

Describe the Bug

#6582 / 73e8625 added _drop_orphaned_function_responses, which correctly prunes id'd function responses with no matching function call before contents assembly.

Two gaps remain and can still permanently poison a session (raise happens in request preprocessing, before any user callback):

  1. _rearrange_events_for_latest_function_response still raises ValueError: No function call event found for function responses ids: ... when it cannot pair the trailing FR event. @GWeale noted on #6587 that this rearrange-path drop is still worth having as defense in depth (main only prunes responses that carry an id).
  2. Id-less / empty-id trailing FRs are intentionally skipped by the prune helper (if response.id), but then rearrange adds None / '' into the id set and raises — so sessions with stripped or missing FR ids still die on the next turn.

Related hole: when rearrange finds a matching FC event but the trailing FR event also carries unmatched ids (Last response event should only contain the responses...), it still raises instead of dropping the unmatched parts.

Steps to Reproduce

from google.adk.events.event import Event
from google.adk.flows.llm_flows.contents import _get_contents
from google.genai import types

events = [
  Event(author='user', content=types.Content(role='user', parts=[types.Part(text='hi')])),
  Event(author='agent', content=types.Content(role='model', parts=[types.Part(text='done')])),
  Event(
    author='agent',
    content=types.Content(
      role='user',
      parts=[types.Part(function_response=types.FunctionResponse(name='tool', id=None, response={'ok': True}))],
    ),
  ),
]
_get_contents(None, events, 'agent')  # ValueError: ... ids: {None}

Expected Behavior

Orphaned / unpairable trailing function responses are dropped (with a warning), contents assembly continues, and the session stays usable.

Observed Behavior

ValueError during contents assembly; every later turn replays the same fatal history.

Environment

  • ADK: current main (post-73e8625)
  • Related: #6582 (closed), #6587 (closed as duplicate of landed prune; rearrange follow-up requested)

Additional Context

Happy to send the rearrange-path follow-up PR referenced by @GWeale on #6587.

Activity

  1. a2105z commented on Aug 16, 2026

    @a2105z
    ContributorAuthor

    Taking this — follow-up PR: #6752

  2. added
    live[Component] This issue is related to live, voice and video chat
    on Aug 17, 2026
  3. surajksharma07 commented on Aug 17, 2026

    @surajksharma07
    Collaborator

    @a2105z The id=None trailing response slips past the orphan-prune's if response.id check and blows up in rearrange exactly as you traced.
    Tried the same drop-unpairable-trailing-FRs approach and it resolved cleanly here, id-less pairs and unmatched-extras included.
    #6752 looks like it's built the same way, could you confirm it holds against your original edge cases before this goes up for merge?

  4. a2105z commented on Aug 17, 2026

    @a2105z
    ContributorAuthor

    @surajksharma07 Thanks for double-checking — confirmed locally against #6752.

    Re-ran the original edge cases on the PR branch:

    • trailing id=None after valid FC/FR history (the prune slip you called out)
    • trailing empty-string id
    • multiple consecutive id-less trailing FRs
    • id’d orphan via the existing prune path
    • rearrange defense-in-depth when prune is bypassed
    • unmatched extras on a matched call event (strip, don’t raise)
    • id-less adjacent FC/FR pair still kept (stripped-id model families)
    • prior valid history preserved after dropping a trailing id-less orphan

    All 8 held, plus pytest tests/unittests/flows/llm_flows/test_contents_function.py tests/unittests/flows/llm_flows/test_contents.py → 56 passed. No further changes needed on this branch from what I can see.

  5. a2105z commented on Aug 26, 2026

    @a2105z
    ContributorAuthor

    Hi maintainers — could someone take another look when you have a moment, and drop the request clarification label?

    @surajksharma07 asked last week to confirm #6752 against the original edge cases before this goes up for merge. That confirmation is already on this thread (the 8 cases plus pytest tests/unittests/flows/llm_flows/test_contents_function.py tests/unittests/flows/llm_flows/test_contents.py → 56 passed). Nothing else is outstanding from the author side.

    This is the rearrange-path follow-up @GWeale requested on #6587: id-less / unmatched trailing FRs still raise after the id-only orphan prune and can poison the session.

    PR: #6752

    cc @GWeale @wuliang229 @DeanChensj @surajksharma07

  6. surajksharma07 commented on Aug 27, 2026

    @surajksharma07
    Collaborator

    @a2105z That covers exactly what I asked for. Re-checked #6752's diff too, the new drop-unmatched-trailing-FR helper plus the id-less-pair carve-out matches your description cleanly and the underlying raise is still live on current main so this is still needed.
    Dropping request clarification, over to @DeanChensj / @GWeale for a merge look, nothing further needed from your side.

  7. removed
    request clarification[Status] The maintainer need clarification or more information from the author
    on Aug 27, 2026
  8. added a commit that references this issue on Sep 29, 2026
    643df96
  9. added 2 commits that reference this issue on Sep 30, 2026
    2c16d9d
    88c7cab
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Labels

live[Component] This issue is related to live, voice and video chat

Type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions