BerriAI/litellm · error · Exception

Filtering by 'target_model_names' is not supported when usin

Error message

Filtering by 'target_model_names' is not supported when using managed batches.

What it means

Generic Exception from list_user_batches: a non-None target_model_names filter is rejected because managed object ids encode only a hash of model name + litellm_params, which cannot be reliably mapped back to target model names. The check sits directly below the provider check in the same method.

Source

Thrown at enterprise/litellm_enterprise/proxy/hooks/managed_files.py:429

        self,
        user_api_key_dict: UserAPIKeyAuth,
        limit: Optional[int] = None,
        after: Optional[str] = None,
        provider: Optional[str] = None,
        target_model_names: Optional[str] = None,
        llm_router: Optional[Router] = None,
    ) -> Dict[str, object]:
        # Provider filtering is not supported for managed batches
        # This is because the encoded object ids stored in the managed objects table do not contain the provider information
        # To support provider filtering, we would need to store the provider information in the encoded object ids
        if provider:
            raise Exception("Filtering by 'provider' is not supported when using managed batches.")

        # Model name filtering is not supported for managed batches
        # This is because the encoded object ids stored in the managed objects table do not contain the model name
        # A hash of the model name + litellm_params for the model name is encoded as the model id. This is not sufficient to reliably map the target model names to the model ids.
        if target_model_names:
            raise Exception("Filtering by 'target_model_names' is not supported when using managed batches.")

        owner_filter = build_owner_filter(user_api_key_dict)
        if owner_filter is None:
            return build_list_page([])

        where_clause: Dict[str, object] = {"file_purpose": "batch", **owner_filter}

        if after:
            cursor_row = await _managed_object_table(self.prisma_client).find_first(
                where={**where_clause, "unified_object_id": after}
            )
            if cursor_row is None:
                raise HTTPException(
                    status_code=400,
                    detail=f"Invalid 'after' cursor: no batch found with id '{after}'.",
                )

        page_size: Final = min(limit or 20, 100)

View on GitHub (pinned to 6c2dcb801b)

Solutions

Drop target_model_names from the call and filter returned batches client-side against your own model metadata
Tag batches with model info at creation time (via metadata you store) if model-based listing is a hard requirement
Remember provider filtering is equally unsupported for managed batches

Example fix

# before
batches = await managed_files.list_user_batches(
    user_api_key_dict, limit=20, target_model_names='gpt-4o')

# after
batches = await managed_files.list_user_batches(user_api_key_dict, limit=20)
gpt4o = [b for b in batches['data'] if b.get('model') == 'gpt-4o']

Defensive patterns

Strategy: validation

Validate before calling

def clean_batch_filters(provider=None, target_model_names=None, **kw):
    if target_model_names:
        logger.warning('target_model_names unsupported for managed batches; dropped')
    kw.pop('target_model_names', None)
    return kw

Type guard

def is_supported_batch_filter(target_model_names: str | None) -> bool:
    return target_model_names is None

Try / catch

try:
    await managed_files.list_user_batches(uak, limit=20, target_model_names=m)
except Exception as e:
    if 'not supported when using managed batches' in str(e):
        page = await managed_files.list_user_batches(uak, limit=20)
        return [b for b in page['data'] if b.get('model') in m.split(',')]
    raise

Prevention

Never forward optional filter params blindly from UI to managed batch APIs
Store model metadata per batch id in your own store at creation time

When it happens

Trigger: Calling list_user_batches with target_model_names set (e.g. 'gpt-4o,claude-3-5') under managed batches. Any truthy value raises before pagination logic runs.

Common situations: Copy-pasting parameters from the non-managed batch list API; UI offering model filters that the managed backend cannot honor; SDK wrappers that always populate optional filter kwargs.

Related errors

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/86001af39129f02f. Report an issue: GitHub.