BerriAI/litellm · error · Exception

LiteLLM Managed {accessor_key} with id={retrieve_object_id}

Error message

LiteLLM Managed {accessor_key} with id={retrieve_object_id} is invalid - does not contain encoded model_id.

What it means

After access checks pass for a managed batch/object id, the hook decodes the unified id to extract the encoded model_id via get_model_id_from_unified_batch_id. If the base64 payload does not contain the expected model_id segment (corrupt, truncated, or hand-crafted id), a plain Exception is raised stating the managed object is invalid. This is a data-integrity failure of the unified id, not an auth failure.

Source

Thrown at enterprise/litellm_enterprise/proxy/hooks/managed_files.py:693

            if accessor_key:
                retrieve_object_id = cast(Optional[str], data.get(accessor_key))

            potential_llm_object_id = (
                _is_base64_encoded_unified_file_id(retrieve_object_id) if retrieve_object_id else False
            )
            if potential_llm_object_id and retrieve_object_id:
                ## VALIDATE USER HAS ACCESS TO THE OBJECT ##
                if not await self.can_user_call_unified_object_id(retrieve_object_id, user_api_key_dict):
                    raise HTTPException(
                        status_code=403,
                        detail=f"User {user_api_key_dict.user_id} does not have access to the object {retrieve_object_id}",
                    )

                ## for managed batch id - get the model id
                potential_model_id = get_model_id_from_unified_batch_id(potential_llm_object_id)
                if potential_model_id is None:
                    raise Exception(
                        f"LiteLLM Managed {accessor_key} with id={retrieve_object_id} is invalid - does not contain encoded model_id."
                    )
                data["model"] = potential_model_id
                data[accessor_key] = get_batch_id_from_unified_batch_id(potential_llm_object_id)
        elif call_type == CallTypes.acreate_fine_tuning_job.value:
            input_file_id = cast(Optional[str], data.get("training_file"))
            if input_file_id:
                model_file_id_mapping = await self.get_model_file_id_mapping(
                    [input_file_id], user_api_key_dict.parent_otel_span
                )

        return data

    async def async_filter_deployments(
        self,
        model: str,
        healthy_deployments: List,
        messages: Optional[List[AllMessageValues]],

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Re-fetch the id from the creation response or /v1/batches list instead of copying it from logs or external storage
2. Ensure ids round-trip byte-for-byte (no trimming of base64 padding, no URL-unescaping twice)
3. If ids were created by an older proxy version, re-create the batch after upgrading so the id carries the current encoding
4. Check for middleware/gateway code that rewrites or truncates long header/query values containing the id

Example fix

# before: id copied from truncated log line
await client.batches.retrieve("ZmlsZSxsbG1fbW9kZWxfaWQ" )  # missing segments

# after: use the id exactly as returned
job = await client.batches.create(...)
await client.batches.retrieve(job.id)

import base64

def unified_id_has_model_segment(unified_id: str) -> bool:
    try:
        decoded = base64.urlsafe_b64encode(unified_id.encode()).decode() if False else unified_id
        # ids arrive base64-encoded; validate round-trip and marker presence
        raw = base64.urlsafe_b64decode(unified_id + "=" * (-len(unified_id) % 4)).decode()
        return "llm_model_id," in raw or "model_id" in raw
    except Exception:
        return False

assert unified_id_has_model_segment(batch_id), "corrupt unified id; refetch it"

def is_wellformed_unified_batch_id(bid: str) -> bool:
    import base64
    try:
        pad = bid + "=" * (-len(bid) % 4)
        raw = base64.urlsafe_b64decode(pad).decode()
        return "model_id" in raw and ";" in raw
    except Exception:
        return False

try:
    await client.batches.retrieve(batch_id)
except Exception as e:
    if "does not contain encoded model_id" in str(e):
        # corrupt id: refetch from creation response, do not retry same id
        raise ValueError(f"corrupt unified id {batch_id}") from e
    raise

When it happens

Trigger: Passing a partially-copied or truncated base64 batch id (URL clipping, manual transcription); a unified id minted by an older proxy version with a different encoding scheme; client code that re-encodes or mangles ids between submit and retrieve.

Common situations: Ids stored in text columns that silently truncate; logging pipelines that strip '=' padding from base64; upgrading LiteLLM enterprise versions where the unified-id format changed; hand-modified ids in tests.

Related errors

• Unified id does not contain {marker!r}: {file_id[:80]!r}
• LiteLLM Managed File object with id={file_id} has no file_ob
• LiteLLM Managed File object with id={file_id} not found
• Failed to retrieve file {file_id} from provider: {str(e)}
• LiteLLM Managed File object with id={file_id} not found. Che

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/649af98863d81430. Report an issue: GitHub.