@anthropic-ai/sdk
Version:
The official TypeScript library for the Anthropic API
183 lines (164 loc) • 6.17 kB
text/typescript
import { APIResource } from '../../../../core/resource';
import * as AnalyticsAPI from './analytics';
import { BetaAnalyticsCostReportTimeBucketsReportPageCursor } from './analytics';
import { PagePromise, ReportPageCursor, type ReportPageCursorParams } from '../../../../core/pagination';
import { RequestOptions } from '../../../../internal/request-options';
import { startApiCallSpan } from '../../../../internal/tracing';
export class CostReport extends APIResource {
/**
* Get cost in USD over time across a date range.
*
* Returns cost bucketed by minute, hour, or day, optionally broken down by
* product, model, context window, inference region, speed, cost type, or token
* type. Available to organizations on a Claude Enterprise plan. Requires an API
* key with the `read:analytics` scope.
*
* @example
* ```ts
* // Automatically fetches more pages as needed.
* for await (const betaAnalyticsCostReportTimeBucket of client.beta.organization.analytics.costReport.list(
* { starting_at: '2019-12-27T18:11:19.117Z' },
* )) {
* // ...
* }
* ```
*/
list(
query: CostReportListParams,
options?: RequestOptions,
): PagePromise<
BetaAnalyticsCostReportTimeBucketsReportPageCursor,
AnalyticsAPI.BetaAnalyticsCostReportTimeBucket
> {
return this._client.getAPIList(
'/v1/organizations/analytics/cost_report?beta=true',
ReportPageCursor<AnalyticsAPI.BetaAnalyticsCostReportTimeBucket>,
{
query,
...options,
__span: startApiCallSpan(this._client, 'anthropic.organization.analytics.cost_report.list', options, {
route: '/v1/organizations/analytics/cost_report',
queryParams: [
'bucket_width',
'claude_tag_categories',
'claude_tag_user_ids',
'context_windows',
'ending_at',
'group_by',
'inference_geos',
'limit',
'models',
'page',
'products',
'rbac_group_ids',
'slack_channel_ids',
'speeds',
'starting_at',
'user_ids',
],
}),
},
);
}
}
export interface CostReportListParams extends ReportPageCursorParams {
/**
* Start of range, inclusive. RFC 3339 tz-aware. Must be within the last 365 days
* and no earlier than 2026-01-01T00:00:00Z.
*/
starting_at: string;
/**
* Time bucket granularity.
*/
bucket_width?: '1d' | '1h' | '1m';
/**
* Filter to Claude Tag (Claude in Slack) usage in specific spend categories. Usage
* with no category never matches. `dm` usage is reported under the user's product
* rather than `claude-tag`, so combining this filter with `products[]=claude-tag`
* excludes it. Use `group_by[]=claude_tag_category` to break out per-category
* values.
*/
claude_tag_categories?: Array<AnalyticsAPI.BetaAnalyticsClaudeTagCategory> | null;
/**
* Filter to Claude Tag (Claude in Slack) usage attributed to specific Slack users,
* by Slack user ID (for example `U0123ABCDEF`), not claude.ai user ID. Usage that
* is not Claude Tag, and Claude Tag usage not attributed to a single user, never
* matches. Use `group_by[]=claude_tag_user_id` to break out per-user values.
*/
claude_tag_user_ids?: Array<string> | null;
/**
* Filter to specific context-window pricing tiers. Use `group_by[]=context_window`
* to break out per-tier values.
*/
context_windows?: Array<AnalyticsAPI.BetaAnalyticsContextWindow> | null;
/**
* End of range, exclusive. When omitted, defaults to the earlier of now and
* `starting_at` + 31 days. The range may span at most 31 days.
*/
ending_at?: string | null;
/**
* Dimensions to break each time bucket out by. Defaults to no grouping (one total
* per bucket). Each bucket reports at most its top 100 groups; a group beyond that
* cap has no row in that bucket (there is no remainder row), so grouped buckets
* are not exhaustive when a dimension has more than 100 distinct values.
*/
group_by?: Array<
| 'claude_tag_category'
| 'claude_tag_user_id'
| 'context_window'
| 'cost_type'
| 'inference_geo'
| 'model'
| 'product'
| 'rbac_group_id'
| 'slack_channel_id'
| 'speed'
| 'token_type'
> | null;
/**
* Filter to specific inference regions. `not_available` matches rows where the
* region is unset. Use `group_by[]=inference_geo` to break out per-region values.
*/
inference_geos?: Array<AnalyticsAPI.BetaAnalyticsInferenceGeoFilter> | null;
/**
* Maximum number of time buckets per page. Defaults and caps vary by
* `bucket_width` (`1d`: default 7, max 31; `1h`: default 24, max 168; `1m`:
* default 60, max 256).
*/
limit?: number | null;
/**
* Models to include. Defaults to all models. Use `group_by[]=model` to break out
* per-model values.
*/
models?: Array<string> | null;
/**
* Product surfaces to include. Defaults to all products. Use `group_by[]=product`
* to break out per-product values.
*/
products?: Array<AnalyticsAPI.BetaAnalyticsProductFilter> | null;
/**
* Filter to usage attributed to specific RBAC groups. Accepts tagged RBAC group
* IDs (`rbac_group_...`) or bare group UUIDs. A row matches when the user belonged
* to any of the listed groups on the (UTC) day the usage occurred; usage with no
* group attribution never matches.
*/
rbac_group_ids?: Array<string> | null;
/**
* Filter to usage originating from specific Slack channels. Use
* `group_by[]=slack_channel_id` to break out per-channel values.
*/
slack_channel_ids?: Array<string> | null;
/**
* Filter to fast or standard inference mode. Use `group_by[]=speed` to break out
* per-mode values.
*/
speeds?: Array<'fast' | 'standard'> | null;
/**
* Filter to specific users by tagged user ID.
*/
user_ids?: Array<string> | null;
}
export declare namespace CostReport {
export { type CostReportListParams as CostReportListParams };
}
export { type BetaAnalyticsCostReportTimeBucketsReportPageCursor };