diff --git a/litellm/proxy/route_llm_request.py b/litellm/proxy/route_llm_request.py index 536c58df65a..42ac74cae33 100644 --- a/litellm/proxy/route_llm_request.py +++ b/litellm/proxy/route_llm_request.py @@ -142,6 +142,7 @@ ROUTE_ENDPOINT_MAPPING: Final = { "acancel_run": "/evals/{eval_id}/runs/{run_id}/cancel", "adelete_run": "/evals/{eval_id}/runs/{run_id}", "acreate_batch": "/batches", + "aretrieve_batch": "/batches", } diff --git a/tests/integration/spend/test_batch_completion_accounting.py b/tests/integration/spend/test_batch_completion_accounting.py index 0cbeda934f6..4ecab10f942 100644 --- a/tests/integration/spend/test_batch_completion_accounting.py +++ b/tests/integration/spend/test_batch_completion_accounting.py @@ -2,11 +2,12 @@ from __future__ import annotations import json import uuid +from datetime import datetime, timedelta, timezone from hashlib import sha256 from typing import Final import pytest -from integration._support.client import JSON_OBJECT, Gateway, eventually, string_value +from integration._support.client import JSON_OBJECT, Gateway, eventually, object_value, string_value from integration._support.database import read_rows from integration._support.upstream import delete_scenario, register_scenario from integration.cost_calculation.cost_tracking_case import JsonResponse, RoutedResponse, TextResponse @@ -116,6 +117,28 @@ def _batch_routes(model: str) -> RoutedResponse: ) +def _team_day_endpoints(gateway: Gateway, team: str, start_date: str, end_date: str) -> dict[str, object] | None: + response: Final = gateway.request( + "GET", + "/team/daily/activity", + params={"team_ids": team, "start_date": start_date, "end_date": end_date}, + ) + if response.status_code != 200: + return None + days: Final = response.json()["results"] + if not days: + return None + return object_value(object_value(object_value(days[0])["breakdown"])["endpoints"]) + + +def _batches_total_tokens(endpoints: dict[str, object] | None) -> int | None: + if endpoints is None or "/batches" not in endpoints: + return None + metrics: Final = object_value(object_value(endpoints["/batches"])["metrics"]) + total_tokens: Final = metrics["total_tokens"] + return int(total_tokens) if isinstance(total_tokens, (int, float, str)) else None + + def _input_file(model: str) -> bytes: return ( "\n".join( @@ -200,3 +223,79 @@ def test_completed_batch_spend_row_records_reasoning_tokens_and_error_file_failu "reasoning_tokens": reasoning_tokens, "text_tokens": completion_tokens - reasoning_tokens, }, json.dumps(metadata) + + +INPUT_COST_PER_TOKEN: Final = 0.001 +OUTPUT_COST_PER_TOKEN: Final = 0.002 +BATCH_PROMPT_TOKENS: Final = FIRST_LINE["prompt_tokens"] + SECOND_LINE["prompt_tokens"] +BATCH_COMPLETION_TOKENS: Final = FIRST_LINE["completion_tokens"] + SECOND_LINE["completion_tokens"] +BATCH_SPEND: Final = (BATCH_PROMPT_TOKENS * INPUT_COST_PER_TOKEN + BATCH_COMPLETION_TOKENS * OUTPUT_COST_PER_TOKEN) / 2 + + +def test_completed_batch_spend_lands_under_batches_in_team_endpoint_activity(gateway: Gateway) -> None: + with gateway.scenario() as scenario: + scenario_id: Final = f"batch-endpoint-{uuid.uuid4().hex[:12]}" + handle: Final = register_scenario(scenario_id, _batch_routes("gpt-4o-mini")) + scenario.cleanups.callback(delete_scenario, handle) + model: Final = scenario.model( + api_base=handle.api_base(), + input_cost_per_token=INPUT_COST_PER_TOKEN, + output_cost_per_token=OUTPUT_COST_PER_TOKEN, + ) + team: Final = scenario.team(models=[model]) + key: Final = scenario.key(team_id=team, models=[model]) + file_response: Final = gateway.request_multipart( + "/v1/files", + {"purpose": "batch", "model": model}, + {"file": ("in.jsonl", _input_file(model), "application/jsonl")}, + key=key, + ) + assert file_response.status_code == 200, file_response.text + batch_response: Final = gateway.request( + "POST", + "/v1/batches", + { + "input_file_id": string_value(JSON_OBJECT.validate_json(file_response.content)["id"]), + "endpoint": "/v1/chat/completions", + "completion_window": "24h", + "model": model, + }, + key=key, + ) + assert batch_response.status_code == 200, batch_response.text + batch_id: Final = string_value(JSON_OBJECT.validate_json(batch_response.content)["id"]) + retrieval: Final = gateway.request("GET", f"/v1/batches/{batch_id}", key=key) + assert retrieval.status_code == 200, retrieval.text + assert retrieval.json()["status"] == "completed", retrieval.text + rows: Final = eventually( + lambda: read_rows( + 'SELECT spend, prompt_tokens, completion_tokens, total_tokens FROM "LiteLLM_SpendLogs" ' + "WHERE api_key=%s AND call_type='aretrieve_batch'", + (sha256(key.encode()).hexdigest(),), + ), + lambda values: len(values) == 1, + seconds=70, + ) + row: Final = rows[0] + assert float(row["spend"]) == pytest.approx(BATCH_SPEND), dict(row) + assert (row["prompt_tokens"], row["completion_tokens"]) == ( + BATCH_PROMPT_TOKENS, + BATCH_COMPLETION_TOKENS, + ), dict(row) + today: Final = datetime.now(timezone.utc) + endpoints: Final = eventually( + lambda: _team_day_endpoints( + gateway, + team, + (today - timedelta(days=1)).strftime("%Y-%m-%d"), + (today + timedelta(days=1)).strftime("%Y-%m-%d"), + ), + lambda value: _batches_total_tokens(value) == BATCH_PROMPT_TOKENS + BATCH_COMPLETION_TOKENS, + seconds=70, + return_last_on_timeout=True, + ) + assert endpoints is not None, "team daily activity returned no endpoint breakdown for the day" + assert set(endpoints) == {"/batches"}, endpoints + endpoint_metrics: Final = object_value(object_value(endpoints["/batches"])["metrics"]) + assert float(endpoint_metrics["spend"]) == pytest.approx(BATCH_SPEND), endpoints + assert endpoint_metrics["total_tokens"] == BATCH_PROMPT_TOKENS + BATCH_COMPLETION_TOKENS, endpoints diff --git a/tests/test_litellm/proxy/db/test_db_spend_update_writer.py b/tests/test_litellm/proxy/db/test_db_spend_update_writer.py index 7abb6e1ef92..7b160c055d2 100644 --- a/tests/test_litellm/proxy/db/test_db_spend_update_writer.py +++ b/tests/test_litellm/proxy/db/test_db_spend_update_writer.py @@ -1638,6 +1638,45 @@ async def test_endpoint_field_is_correctly_mapped_from_call_type(): assert transaction["custom_llm_provider"] == "openai" +@pytest.mark.asyncio +async def test_endpoint_field_maps_retrieve_batch_spend_row_to_batches_endpoint(): + writer = DBSpendUpdateWriter() + mock_prisma = MagicMock() + mock_prisma.get_request_status = MagicMock(return_value="success") + + payload = { + "request_id": "req-retrieve-batch", + "user": "test-user", + "call_type": "aretrieve_batch", + "startTime": "2024-01-01T12:00:00", + "api_key": "test-key", + "model": "gpt-4", + "custom_llm_provider": "openai", + "model_group": "gpt-4-group", + "prompt_tokens": 15, + "completion_tokens": 10, + "spend": 0.0175, + "metadata": '{"usage_object": {}}', + } + + writer.daily_spend_update_queue.add_update = AsyncMock() + + await writer.add_spend_log_transaction_to_daily_user_transaction( + payload=payload, + prisma_client=mock_prisma, + ) + + writer.daily_spend_update_queue.add_update.assert_called_once() + + call_args = writer.daily_spend_update_queue.add_update.call_args[1] + update_dict = call_args["update"] + assert len(update_dict) == 1 + + for key, transaction in update_dict.items(): + assert key == "test-user_2024-01-01_test-key_gpt-4_openai_/batches" + assert transaction["endpoint"] == "/batches" + + @pytest.mark.asyncio async def test_update_daily_spend_logs_detailed_error_on_batch_upsert_failure(): """