mirror of
https://github.com/BerriAI/litellm.git
synced 2026-10-24 05:45:11 +00:00
a6245c1bbb
* test(perf): measure SpendLog-backed UI query latency * test(perf): validate every numeric row and both counter snapshots * test(perf): isolate read API cases and report request metrics --------- Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
48 lines
2.7 KiB
SQL
48 lines
2.7 KiB
SQL
INSERT INTO "LiteLLM_SpendLogs"
|
|
SELECT (jsonb_populate_record(NULL::"LiteLLM_SpendLogs", to_jsonb(template) || jsonb_build_object(
|
|
'request_id', 'perf-ui-' || g,
|
|
'litellm_call_id', md5('perf-ui-call-' || g),
|
|
'api_key', %(api_key)s::text,
|
|
'team_id', 'perf-team-0',
|
|
'user', 'perf-user-' || g %% 200,
|
|
'end_user', 'perf-end-user-' || g %% 200,
|
|
'model', 'perf-chat', 'model_id', 'perf-model-0', 'model_group', 'perf-chat',
|
|
'custom_llm_provider', 'openai',
|
|
'startTime', %(anchor)s::timestamp - g * interval '10 seconds',
|
|
'endTime', %(anchor)s::timestamp - g * interval '10 seconds' + interval '1 second',
|
|
'completionStartTime', %(anchor)s::timestamp - g * interval '10 seconds' + interval '100 milliseconds',
|
|
'session_id', CASE WHEN g %% 2 = 0 THEN 'perf-ui-session' ELSE 'perf-ui-session-' || g END,
|
|
'status', CASE WHEN g %% 10 = 0 THEN 'failure' ELSE 'success' END,
|
|
'cache_hit', CASE WHEN g %% 10 = 1 THEN 'True' ELSE 'False' END,
|
|
'spend', 0.01, 'prompt_tokens', 1000, 'completion_tokens', 100, 'total_tokens', 1100,
|
|
'messages', '{}', 'response', '{}', 'proxy_server_request', '{}',
|
|
'metadata', jsonb_build_object(
|
|
'used_client_oauth_token', false,
|
|
'user_api_key_alias', 'perf-key-0', 'user_api_key_team_alias', 'perf-team-0',
|
|
'usage_object', jsonb_build_object('prompt_tokens', 1000, 'completion_tokens', 100,
|
|
'cache_read_input_tokens', 500, 'cache_creation_input_tokens', 100),
|
|
%(gateway_marker)s::text, 'perf-model-0',
|
|
'error_information', CASE WHEN g %% 10 = 0 THEN
|
|
jsonb_build_object('error_code', '429', 'error_message', 'Synthetic upstream rate limit') ELSE '{}'::jsonb END,
|
|
'guardrail_information', jsonb_build_array(jsonb_build_object(
|
|
'guardrail_name', 'perf-ui-guardrail', 'guardrail_status', 'success', 'duration', 0.01))
|
|
)
|
|
))).*
|
|
FROM "LiteLLM_SpendLogs" template CROSS JOIN generate_series(1, 320) g
|
|
WHERE template.request_id = 'perf-seed-1'
|
|
ON CONFLICT (request_id) DO NOTHING;
|
|
|
|
INSERT INTO "LiteLLM_GuardrailsTable" (guardrail_id, guardrail_name, litellm_params, updated_at)
|
|
VALUES ('perf-ui-guardrail', 'perf-ui-guardrail',
|
|
'{"guardrail":"generic_guardrail_api","mode":"pre_call","api_base":"http://127.0.0.1:1"}', now())
|
|
ON CONFLICT (guardrail_id) DO NOTHING;
|
|
|
|
INSERT INTO "LiteLLM_SpendLogGuardrailIndex" (request_id, guardrail_id, start_time)
|
|
SELECT request_id, 'perf-ui-guardrail', "startTime" FROM "LiteLLM_SpendLogs"
|
|
WHERE request_id = ANY(%(request_ids)s)
|
|
ON CONFLICT DO NOTHING;
|
|
|
|
INSERT INTO "LiteLLM_SpendLogToolIndex" (request_id, tool_name, start_time)
|
|
SELECT request_id, 'perf-ui-tool', "startTime" FROM "LiteLLM_SpendLogs"
|
|
WHERE request_id = ANY(%(request_ids)s)
|
|
ON CONFLICT DO NOTHING;
|