feat: add chat cost analytics backend (#23036)

Add cost tracking for LLM chat interactions with microdollar precision.

## Changes
- Add `chatcost` package for per-message cost calculation using
`shopspring/decimal` for intermediate arithmetic
- **Ceil rounding policy**: fractional micros round UP to next whole
micro (applied once after summing all components)
- Database migration: `total_cost_micros` BIGINT column with historical
backfill and `created_at` index
- API endpoints: per-user cost summary and admin rollup under
`/api/experimental/chats/cost/`
- SDK types: `ChatCostSummary`, `ChatCostModelBreakdown`,
`ChatCostUserRollup`
- Fix `modeloptionsgen` to handle `decimal.Decimal` as opaque numeric
type
- Update frontend pricing test fixtures for string decimal types

## Design decisions
- `NULL` = unpriced (no matching model config), `0` = free
- Reasoning tokens included in output tokens (no double-counting)
- Integer microdollars (BIGINT) for storage and API responses
- Price config uses `decimal.Decimal` for exact parsing; totals use
`int64`

Frontend: #23037
This commit is contained in:
Michael Suchacz
2026-03-13 18:30:49 +01:00
committed by GitHub
parent 1152b61ebb
commit c3b6284955
34 changed files with 2034 additions and 262 deletions
+165 -2
View File
@@ -179,7 +179,8 @@ INSERT INTO chat_messages (
cache_creation_tokens,
cache_read_tokens,
context_limit,
compressed
compressed,
total_cost_micros
) VALUES (
@chat_id::uuid,
sqlc.narg('created_by')::uuid,
@@ -195,7 +196,8 @@ INSERT INTO chat_messages (
sqlc.narg('cache_creation_tokens')::bigint,
sqlc.narg('cache_read_tokens')::bigint,
sqlc.narg('context_limit')::bigint,
COALESCE(sqlc.narg('compressed')::boolean, FALSE)
COALESCE(sqlc.narg('compressed')::boolean, FALSE),
sqlc.narg('total_cost_micros')::bigint
)
RETURNING
*;
@@ -481,3 +483,164 @@ SET
updated_at = NOW()
WHERE
chat_id = @chat_id::uuid;
-- name: GetChatCostSummary :one
-- Aggregate cost summary for a single user within a date range.
-- Only counts assistant-role messages.
SELECT
COALESCE(SUM(cm.total_cost_micros), 0)::bigint AS total_cost_micros,
COUNT(*) FILTER (
WHERE cm.total_cost_micros IS NOT NULL
)::bigint AS priced_message_count,
COUNT(*) FILTER (
WHERE cm.total_cost_micros IS NULL
AND (
cm.input_tokens IS NOT NULL
OR cm.output_tokens IS NOT NULL
OR cm.reasoning_tokens IS NOT NULL
OR cm.cache_creation_tokens IS NOT NULL
OR cm.cache_read_tokens IS NOT NULL
)
)::bigint AS unpriced_message_count,
COALESCE(SUM(cm.input_tokens), 0)::bigint AS total_input_tokens,
COALESCE(SUM(cm.output_tokens), 0)::bigint AS total_output_tokens
FROM
chat_messages cm
JOIN
chats c ON c.id = cm.chat_id
WHERE
c.owner_id = @owner_id::uuid
AND cm.role = 'assistant'
AND cm.created_at >= @start_date::timestamptz
AND cm.created_at < @end_date::timestamptz;
-- name: GetChatCostPerModel :many
-- Per-model cost breakdown for a single user within a date range.
-- Only counts assistant-role messages that have a model_config_id.
SELECT
cmc.id AS model_config_id,
cmc.display_name,
cmc.provider,
cmc.model,
COALESCE(SUM(cm.total_cost_micros), 0)::bigint AS total_cost_micros,
COUNT(*) FILTER (
WHERE cm.input_tokens IS NOT NULL
OR cm.output_tokens IS NOT NULL
OR cm.reasoning_tokens IS NOT NULL
OR cm.cache_creation_tokens IS NOT NULL
OR cm.cache_read_tokens IS NOT NULL
)::bigint AS message_count,
COALESCE(SUM(cm.input_tokens), 0)::bigint AS total_input_tokens,
COALESCE(SUM(cm.output_tokens), 0)::bigint AS total_output_tokens
FROM
chat_messages cm
JOIN
chats c ON c.id = cm.chat_id
JOIN
chat_model_configs cmc ON cmc.id = cm.model_config_id
WHERE
c.owner_id = @owner_id::uuid
AND cm.role = 'assistant'
AND cm.created_at >= @start_date::timestamptz
AND cm.created_at < @end_date::timestamptz
GROUP BY
cmc.id, cmc.display_name, cmc.provider, cmc.model
ORDER BY
total_cost_micros DESC;
-- name: GetChatCostPerChat :many
-- Per-root-chat cost breakdown for a single user within a date range.
-- Groups by root_chat_id so forked chats roll up under their root.
-- Only counts assistant-role messages.
WITH chat_costs AS (
SELECT
COALESCE(c.root_chat_id, c.id) AS root_chat_id,
COALESCE(SUM(cm.total_cost_micros), 0)::bigint AS total_cost_micros,
COUNT(*) FILTER (
WHERE cm.input_tokens IS NOT NULL
OR cm.output_tokens IS NOT NULL
OR cm.reasoning_tokens IS NOT NULL
OR cm.cache_creation_tokens IS NOT NULL
OR cm.cache_read_tokens IS NOT NULL
)::bigint AS message_count,
COALESCE(SUM(cm.input_tokens), 0)::bigint AS total_input_tokens,
COALESCE(SUM(cm.output_tokens), 0)::bigint AS total_output_tokens
FROM chat_messages cm
JOIN chats c ON c.id = cm.chat_id
WHERE c.owner_id = @owner_id::uuid
AND cm.role = 'assistant'
AND cm.created_at >= @start_date::timestamptz
AND cm.created_at < @end_date::timestamptz
GROUP BY COALESCE(c.root_chat_id, c.id)
)
SELECT
cc.root_chat_id,
COALESCE(rc.title, '') AS chat_title,
cc.total_cost_micros,
cc.message_count,
cc.total_input_tokens,
cc.total_output_tokens
FROM chat_costs cc
LEFT JOIN chats rc ON rc.id = cc.root_chat_id
ORDER BY cc.total_cost_micros DESC;
-- name: GetChatCostPerUser :many
-- Deployment-wide per-user cost rollup within a date range.
-- Only counts assistant-role messages.
WITH chat_cost_users AS (
SELECT
c.owner_id AS user_id,
u.username,
u.name,
u.avatar_url,
COALESCE(SUM(cm.total_cost_micros), 0)::bigint AS total_cost_micros,
COUNT(*) FILTER (
WHERE cm.input_tokens IS NOT NULL
OR cm.output_tokens IS NOT NULL
OR cm.reasoning_tokens IS NOT NULL
OR cm.cache_creation_tokens IS NOT NULL
OR cm.cache_read_tokens IS NOT NULL
)::bigint AS message_count,
COUNT(DISTINCT COALESCE(c.root_chat_id, c.id))::bigint AS chat_count,
COALESCE(SUM(cm.input_tokens), 0)::bigint AS total_input_tokens,
COALESCE(SUM(cm.output_tokens), 0)::bigint AS total_output_tokens
FROM
chat_messages cm
JOIN
chats c ON c.id = cm.chat_id
JOIN
users u ON u.id = c.owner_id
WHERE
cm.role = 'assistant'
AND cm.created_at >= @start_date::timestamptz
AND cm.created_at < @end_date::timestamptz
AND (
@username::text = ''
OR u.username ILIKE '%' || @username::text || '%'
)
GROUP BY
c.owner_id,
u.username,
u.name,
u.avatar_url
)
SELECT
user_id,
username,
name,
avatar_url,
total_cost_micros,
message_count,
chat_count,
total_input_tokens,
total_output_tokens,
COUNT(*) OVER()::bigint AS total_count
FROM
chat_cost_users
ORDER BY
total_cost_micros DESC,
username ASC
LIMIT
sqlc.arg('page_limit')::int
OFFSET
sqlc.arg('page_offset')::int;