Implementation:BerriAI Litellm Check Responses Cost

Attribute	Value
Sources	enterprise/litellm_enterprise/proxy/common_utils/check_responses_cost.py
Domains	Cost Tracking, Responses API, Background Jobs, Enterprise
Last Updated	2026-02-15 16:00 GMT

Overview

CheckResponsesCost is a background polling utility that checks the completion status of managed Responses API jobs and relies on litellm.aget_responses() to automatically track their cost.

Description

This class polls the LiteLLM_ManagedObjectTable database table for response jobs that are in queued or in_progress status with file_purpose="response". For each active job, it:

1. Retrieves the stored response object to extract the model name.
2. Decrypts the unified response ID using ResponsesIDSecurity._decrypt_response_id.
3. Calls litellm.aget_responses() with the decrypted response ID and metadata. Cost tracking is handled automatically by this call.
4. Checks the response status:
   • "completed" -- Marks the job as completed.
   • "failed" or "cancelled" -- Also marks the job as completed (terminal state).
   • Other statuses -- Skips (job remains in progress).
5. Updates all completed jobs to "completed" status in the database via a batch update.

Unlike CheckBatchCost, this class does not need to manually calculate cost because litellm.aget_responses() handles cost tracking through the standard logging pipeline.

Usage

Instantiate and call check_responses_cost periodically (e.g., via a background scheduler) to ensure background Responses API jobs are tracked after completion.

Code Reference

Source Location

enterprise/litellm_enterprise/proxy/common_utils/check_responses_cost.py

Signature

class CheckResponsesCost:
    def __init__(
        self,
        proxy_logging_obj: ProxyLogging,
        prisma_client: PrismaClient,
        llm_router: Router,
    ): ...

    async def check_responses_cost(self): ...

Import

from litellm_enterprise.proxy.common_utils.check_responses_cost import CheckResponsesCost

I/O Contract

Inputs

Parameter	Type	Description
proxy_logging_obj	ProxyLogging	Proxy logging object (currently unused but maintained for consistency).
prisma_client	PrismaClient	Database client to query managed object table.
llm_router	Router	LLM router (currently unused; responses use litellm.aget_responses directly).

Outputs

Output	Type	Description
Cost logging	Side effect	Cost automatically tracked by litellm.aget_responses().
Status update	Side effect	Completed/failed/cancelled jobs updated to "completed" in the database.

Usage Examples

from litellm_enterprise.proxy.common_utils.check_responses_cost import CheckResponsesCost

checker = CheckResponsesCost(
    proxy_logging_obj=proxy_logging_obj,
    prisma_client=prisma_client,
    llm_router=llm_router,
)

# Run as a periodic background task
await checker.check_responses_cost()

Related Pages

• BerriAI_Litellm_Check_Batch_Cost -- Similar polling for batch job cost tracking
• BerriAI_Litellm_Managed_Files -- Managed files hook used for file operations