langchain-ai/deepagents · error · TypeError

Unsupported hook-specific output: {type(specific).__name__}

Error message

Unsupported hook-specific output: {type(specific).__name__}

What it means

`_merge_specific` is a singledispatch function in `libs/code/deepagents_code/hooks/reducer.py` that folds handler-specific output (e.g. session-start context, decision fields) into the reduction state. The fallback raises `TypeError` when handed a `HookSpecificOutput` subtype with no registered merger, indicating output that the reducer cannot incorporate.

Source

Thrown at libs/code/deepagents_code/hooks/reducer.py:298

                    f"Ignored Stop continuation after {MAX_STOP_CONTINUATIONS} "
                    "consecutive attempts"
                ),
            )
        )
        return
    state.continue_loop = True
    state.feedback.append(message)


@singledispatch
def _merge_specific(
    specific: HookSpecificOutput,
    _invocation: HookInvocation,
    _state: _Reduction,
    _handler_id: str,
) -> None:
    msg = f"Unsupported hook-specific output: {type(specific).__name__}"
    raise TypeError(msg)


@_merge_specific.register
def _merge_session_start(
    specific: SessionStartSpecificOutput,
    _invocation: HookInvocation,
    state: _Reduction,
    handler_id: str,
) -> None:
    _append(state.context, specific.additional_context)
    for attr, wire_name in _UNSUPPORTED_SESSION_START_FIELDS:
        value = getattr(specific, attr)
        if value not in (None, False, [], ""):
            _diagnose_unsupported_field(state, handler_id, wire_name)


@_merge_specific.register
def _merge_user_prompt_submit(

View on GitHub (pinned to a1af029e6e)

Solutions

1. Return a supported HookSpecificOutput type (e.g. `SessionStartSpecificOutput`) from the handler
2. Register a merger via `@_merge_specific.register` for the custom output type
3. Update/align library versions so handler outputs and reducer registrations match

Example fix

// before
class MyOutput(HookSpecificOutput): ...
return MyOutput()
// after
return SessionStartSpecificOutput(...)  # or register @_merge_specific.register for MyOutput

if not isinstance(output, (SessionStartSpecificOutput,)):  # supported types
    raise TypeError(f"unsupported hook-specific output: {type(output).__name__}")

from typing import TypeGuard

def is_mergeable(o: HookSpecificOutput) -> TypeGuard[SessionStartSpecificOutput]:
    return isinstance(o, SessionStartSpecificOutput)

try:
    decision = reduce_hook_results(results, event)
except TypeError as e:
    logger.warning("dropping unmergeable hook output: %s", e)

When it happens

Trigger: A hook handler returns a `HookSpecificOutput` subclass that has no `@_merge_specific.register` handler — typically a custom output type, a mocked output in tests, or output from a newer library version not matched by the reducer table.

Common situations: Writing a custom hook handler that returns a hand-rolled HookSpecificOutput subclass; test doubles implementing the output protocol; version skew between hook handler code and the reducer.

Related errors

• Unsupported hook event: {type(event).__name__}
• Unsupported hook event: {event}
• Expected UserPromptSubmitDecision, got {type(decision).__nam
• Expected PreCompactDecision, got {type(decision).__name__}
• Expected PermissionRequestDecision, got {type(decision).__na

AI-assisted analysis of langchain-ai/deepagents@a1af029e6e (2026-08-29). Data as JSON: /api/errors/fe45545964e86adf. Report an issue: GitHub.