mirror of
https://github.com/multica-ai/multica.git
synced 2026-07-06 05:49:12 +02:00
Addresses 4 review blockers on the original PR: 1. Cron/backfill double-count race: the rollup function is now idempotent. Window calls find DIRTY KEYS via task_usage.updated_at, then RECOMPUTE each bucket from ground truth and REPLACE the daily row (no more additive ON CONFLICT). Cron and backfill can now overlap safely. 2. Silent pg_cron absence: the read path is gated behind a new USAGE_DAILY_ROLLUP_ENABLED feature flag (default off). The raw task_usage scan is preserved as the fallback. Operators flip the flag per-environment after backfill + cron are confirmed healthy (task_usage_rollup_lag_seconds() helper added for monitoring). 3. UpsertTaskUsage corrections invisible to rollup: added task_usage.updated_at column (default now(), backfilled from created_at), and bumped it on conflict. Corrections now mark the bucket dirty and the next window call recomputes it correctly. 4. CREATE INDEX blocking writes on hot table: split into separate single-statement migrations using CREATE INDEX CONCURRENTLY (074, 075), matching the 035/067 pattern. Also: cron.schedule() removed from migrations entirely. Migration 076 only enables the extension (gracefully on unsupported envs); the actual schedule is a documented operator runbook step that runs AFTER backfill. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: multica-agent <github@multica.ai>
315 lines
9.9 KiB
Go
315 lines
9.9 KiB
Go
// Code generated by sqlc. DO NOT EDIT.
|
|
// versions:
|
|
// sqlc v1.30.0
|
|
// source: runtime_usage.sql
|
|
|
|
package db
|
|
|
|
import (
|
|
"context"
|
|
|
|
"github.com/jackc/pgx/v5/pgtype"
|
|
)
|
|
|
|
const getRuntimeTaskHourlyActivity = `-- name: GetRuntimeTaskHourlyActivity :many
|
|
SELECT EXTRACT(HOUR FROM started_at)::int AS hour, COUNT(*)::int AS count
|
|
FROM agent_task_queue
|
|
WHERE runtime_id = $1 AND started_at IS NOT NULL
|
|
GROUP BY hour
|
|
ORDER BY hour
|
|
`
|
|
|
|
type GetRuntimeTaskHourlyActivityRow struct {
|
|
Hour int32 `json:"hour"`
|
|
Count int32 `json:"count"`
|
|
}
|
|
|
|
func (q *Queries) GetRuntimeTaskHourlyActivity(ctx context.Context, runtimeID pgtype.UUID) ([]GetRuntimeTaskHourlyActivityRow, error) {
|
|
rows, err := q.db.Query(ctx, getRuntimeTaskHourlyActivity, runtimeID)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
defer rows.Close()
|
|
items := []GetRuntimeTaskHourlyActivityRow{}
|
|
for rows.Next() {
|
|
var i GetRuntimeTaskHourlyActivityRow
|
|
if err := rows.Scan(&i.Hour, &i.Count); err != nil {
|
|
return nil, err
|
|
}
|
|
items = append(items, i)
|
|
}
|
|
if err := rows.Err(); err != nil {
|
|
return nil, err
|
|
}
|
|
return items, nil
|
|
}
|
|
|
|
const getRuntimeUsageByHour = `-- name: GetRuntimeUsageByHour :many
|
|
SELECT
|
|
EXTRACT(HOUR FROM tu.created_at)::int AS hour,
|
|
tu.model,
|
|
SUM(tu.input_tokens)::bigint AS input_tokens,
|
|
SUM(tu.output_tokens)::bigint AS output_tokens,
|
|
SUM(tu.cache_read_tokens)::bigint AS cache_read_tokens,
|
|
SUM(tu.cache_write_tokens)::bigint AS cache_write_tokens,
|
|
COUNT(DISTINCT tu.task_id)::int AS task_count
|
|
FROM task_usage tu
|
|
JOIN agent_task_queue atq ON atq.id = tu.task_id
|
|
WHERE atq.runtime_id = $1
|
|
AND tu.created_at >= DATE_TRUNC('day', $2::timestamptz)
|
|
GROUP BY EXTRACT(HOUR FROM tu.created_at), tu.model
|
|
ORDER BY hour, tu.model
|
|
`
|
|
|
|
type GetRuntimeUsageByHourParams struct {
|
|
RuntimeID pgtype.UUID `json:"runtime_id"`
|
|
Since pgtype.Timestamptz `json:"since"`
|
|
}
|
|
|
|
type GetRuntimeUsageByHourRow struct {
|
|
Hour int32 `json:"hour"`
|
|
Model string `json:"model"`
|
|
InputTokens int64 `json:"input_tokens"`
|
|
OutputTokens int64 `json:"output_tokens"`
|
|
CacheReadTokens int64 `json:"cache_read_tokens"`
|
|
CacheWriteTokens int64 `json:"cache_write_tokens"`
|
|
TaskCount int32 `json:"task_count"`
|
|
}
|
|
|
|
// Per-(hour, model) token aggregates (hour ∈ 0..23) for a runtime since a
|
|
// cutoff. Powers the "By hour" tab — shows when in the day this runtime is
|
|
// doing real work, with model preserved for client-side cost calculation
|
|
// (same reason as ListRuntimeUsageByAgent above). Hours with zero activity
|
|
// are omitted; the client fills the 24-bucket axis.
|
|
func (q *Queries) GetRuntimeUsageByHour(ctx context.Context, arg GetRuntimeUsageByHourParams) ([]GetRuntimeUsageByHourRow, error) {
|
|
rows, err := q.db.Query(ctx, getRuntimeUsageByHour, arg.RuntimeID, arg.Since)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
defer rows.Close()
|
|
items := []GetRuntimeUsageByHourRow{}
|
|
for rows.Next() {
|
|
var i GetRuntimeUsageByHourRow
|
|
if err := rows.Scan(
|
|
&i.Hour,
|
|
&i.Model,
|
|
&i.InputTokens,
|
|
&i.OutputTokens,
|
|
&i.CacheReadTokens,
|
|
&i.CacheWriteTokens,
|
|
&i.TaskCount,
|
|
); err != nil {
|
|
return nil, err
|
|
}
|
|
items = append(items, i)
|
|
}
|
|
if err := rows.Err(); err != nil {
|
|
return nil, err
|
|
}
|
|
return items, nil
|
|
}
|
|
|
|
const listRuntimeUsage = `-- name: ListRuntimeUsage :many
|
|
SELECT
|
|
DATE(tu.created_at) AS date,
|
|
tu.provider,
|
|
tu.model,
|
|
SUM(tu.input_tokens)::bigint AS input_tokens,
|
|
SUM(tu.output_tokens)::bigint AS output_tokens,
|
|
SUM(tu.cache_read_tokens)::bigint AS cache_read_tokens,
|
|
SUM(tu.cache_write_tokens)::bigint AS cache_write_tokens
|
|
FROM task_usage tu
|
|
JOIN agent_task_queue atq ON atq.id = tu.task_id
|
|
WHERE atq.runtime_id = $1
|
|
AND tu.created_at >= DATE_TRUNC('day', $2::timestamptz)
|
|
GROUP BY DATE(tu.created_at), tu.provider, tu.model
|
|
ORDER BY DATE(tu.created_at) DESC, tu.provider, tu.model
|
|
`
|
|
|
|
type ListRuntimeUsageParams struct {
|
|
RuntimeID pgtype.UUID `json:"runtime_id"`
|
|
Since pgtype.Timestamptz `json:"since"`
|
|
}
|
|
|
|
type ListRuntimeUsageRow struct {
|
|
Date pgtype.Date `json:"date"`
|
|
Provider string `json:"provider"`
|
|
Model string `json:"model"`
|
|
InputTokens int64 `json:"input_tokens"`
|
|
OutputTokens int64 `json:"output_tokens"`
|
|
CacheReadTokens int64 `json:"cache_read_tokens"`
|
|
CacheWriteTokens int64 `json:"cache_write_tokens"`
|
|
}
|
|
|
|
// Reads from raw `task_usage`, bucketed by DATE(tu.created_at) — usage
|
|
// report time, ~= task completion time. Since cutoff is truncated to
|
|
// start-of-day so `days=N` yields full calendar days. This is the
|
|
// always-correct fallback path; used when USAGE_DAILY_ROLLUP_ENABLED
|
|
// is false (or the rollup hasn't been deployed yet).
|
|
func (q *Queries) ListRuntimeUsage(ctx context.Context, arg ListRuntimeUsageParams) ([]ListRuntimeUsageRow, error) {
|
|
rows, err := q.db.Query(ctx, listRuntimeUsage, arg.RuntimeID, arg.Since)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
defer rows.Close()
|
|
items := []ListRuntimeUsageRow{}
|
|
for rows.Next() {
|
|
var i ListRuntimeUsageRow
|
|
if err := rows.Scan(
|
|
&i.Date,
|
|
&i.Provider,
|
|
&i.Model,
|
|
&i.InputTokens,
|
|
&i.OutputTokens,
|
|
&i.CacheReadTokens,
|
|
&i.CacheWriteTokens,
|
|
); err != nil {
|
|
return nil, err
|
|
}
|
|
items = append(items, i)
|
|
}
|
|
if err := rows.Err(); err != nil {
|
|
return nil, err
|
|
}
|
|
return items, nil
|
|
}
|
|
|
|
const listRuntimeUsageByAgent = `-- name: ListRuntimeUsageByAgent :many
|
|
SELECT
|
|
atq.agent_id,
|
|
tu.model,
|
|
SUM(tu.input_tokens)::bigint AS input_tokens,
|
|
SUM(tu.output_tokens)::bigint AS output_tokens,
|
|
SUM(tu.cache_read_tokens)::bigint AS cache_read_tokens,
|
|
SUM(tu.cache_write_tokens)::bigint AS cache_write_tokens,
|
|
COUNT(DISTINCT tu.task_id)::int AS task_count
|
|
FROM task_usage tu
|
|
JOIN agent_task_queue atq ON atq.id = tu.task_id
|
|
WHERE atq.runtime_id = $1
|
|
AND tu.created_at >= DATE_TRUNC('day', $2::timestamptz)
|
|
GROUP BY atq.agent_id, tu.model
|
|
ORDER BY atq.agent_id, tu.model
|
|
`
|
|
|
|
type ListRuntimeUsageByAgentParams struct {
|
|
RuntimeID pgtype.UUID `json:"runtime_id"`
|
|
Since pgtype.Timestamptz `json:"since"`
|
|
}
|
|
|
|
type ListRuntimeUsageByAgentRow struct {
|
|
AgentID pgtype.UUID `json:"agent_id"`
|
|
Model string `json:"model"`
|
|
InputTokens int64 `json:"input_tokens"`
|
|
OutputTokens int64 `json:"output_tokens"`
|
|
CacheReadTokens int64 `json:"cache_read_tokens"`
|
|
CacheWriteTokens int64 `json:"cache_write_tokens"`
|
|
TaskCount int32 `json:"task_count"`
|
|
}
|
|
|
|
// Per-(agent, model) token aggregates for a runtime since a cutoff. Powers
|
|
// the runtime-detail "Cost by agent" tab. task_usage only carries task_id,
|
|
// so we join the queue to expose agent_id. The model dimension is kept on
|
|
// purpose: cost is computed client-side from a per-model pricing table, so
|
|
// collapsing models server-side would erase the information needed to do
|
|
// that arithmetic. The client groups by agent_id and sums cost per agent.
|
|
func (q *Queries) ListRuntimeUsageByAgent(ctx context.Context, arg ListRuntimeUsageByAgentParams) ([]ListRuntimeUsageByAgentRow, error) {
|
|
rows, err := q.db.Query(ctx, listRuntimeUsageByAgent, arg.RuntimeID, arg.Since)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
defer rows.Close()
|
|
items := []ListRuntimeUsageByAgentRow{}
|
|
for rows.Next() {
|
|
var i ListRuntimeUsageByAgentRow
|
|
if err := rows.Scan(
|
|
&i.AgentID,
|
|
&i.Model,
|
|
&i.InputTokens,
|
|
&i.OutputTokens,
|
|
&i.CacheReadTokens,
|
|
&i.CacheWriteTokens,
|
|
&i.TaskCount,
|
|
); err != nil {
|
|
return nil, err
|
|
}
|
|
items = append(items, i)
|
|
}
|
|
if err := rows.Err(); err != nil {
|
|
return nil, err
|
|
}
|
|
return items, nil
|
|
}
|
|
|
|
const listRuntimeUsageDaily = `-- name: ListRuntimeUsageDaily :many
|
|
SELECT
|
|
bucket_date AS date,
|
|
provider,
|
|
model,
|
|
SUM(input_tokens)::bigint AS input_tokens,
|
|
SUM(output_tokens)::bigint AS output_tokens,
|
|
SUM(cache_read_tokens)::bigint AS cache_read_tokens,
|
|
SUM(cache_write_tokens)::bigint AS cache_write_tokens
|
|
FROM task_usage_daily
|
|
WHERE runtime_id = $1
|
|
AND bucket_date >= DATE(DATE_TRUNC('day', $2::timestamptz))
|
|
GROUP BY bucket_date, provider, model
|
|
ORDER BY bucket_date DESC, provider, model
|
|
`
|
|
|
|
type ListRuntimeUsageDailyParams struct {
|
|
RuntimeID pgtype.UUID `json:"runtime_id"`
|
|
Since pgtype.Timestamptz `json:"since"`
|
|
}
|
|
|
|
type ListRuntimeUsageDailyRow struct {
|
|
Date pgtype.Date `json:"date"`
|
|
Provider string `json:"provider"`
|
|
Model string `json:"model"`
|
|
InputTokens int64 `json:"input_tokens"`
|
|
OutputTokens int64 `json:"output_tokens"`
|
|
CacheReadTokens int64 `json:"cache_read_tokens"`
|
|
CacheWriteTokens int64 `json:"cache_write_tokens"`
|
|
}
|
|
|
|
// Reads from the `task_usage_daily` rollup table maintained by
|
|
// rollup_task_usage_daily() (scheduled every 5 min via pg_cron, or any
|
|
// equivalent external scheduler that calls the function). Same shape as
|
|
// ListRuntimeUsage above. Today's bucket may lag the raw table by up to
|
|
// ~10 min (5 min cron period + 5 min rollup safety lag); intentional.
|
|
//
|
|
// Only used when USAGE_DAILY_ROLLUP_ENABLED is true AND deploy has
|
|
// verified that the rollup is fresh (see task_usage_rollup_lag_seconds
|
|
// helper from migration 076).
|
|
//
|
|
// The PK on task_usage_daily already collapses to one row per
|
|
// (bucket_date, runtime_id, provider, model), but SUM/GROUP BY is kept
|
|
// so future schema changes (extra dimensions promoted into the table)
|
|
// don't silently change query semantics.
|
|
func (q *Queries) ListRuntimeUsageDaily(ctx context.Context, arg ListRuntimeUsageDailyParams) ([]ListRuntimeUsageDailyRow, error) {
|
|
rows, err := q.db.Query(ctx, listRuntimeUsageDaily, arg.RuntimeID, arg.Since)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
defer rows.Close()
|
|
items := []ListRuntimeUsageDailyRow{}
|
|
for rows.Next() {
|
|
var i ListRuntimeUsageDailyRow
|
|
if err := rows.Scan(
|
|
&i.Date,
|
|
&i.Provider,
|
|
&i.Model,
|
|
&i.InputTokens,
|
|
&i.OutputTokens,
|
|
&i.CacheReadTokens,
|
|
&i.CacheWriteTokens,
|
|
); err != nil {
|
|
return nil, err
|
|
}
|
|
items = append(items, i)
|
|
}
|
|
if err := rows.Err(); err != nil {
|
|
return nil, err
|
|
}
|
|
return items, nil
|
|
}
|