Skip to main content
POST
Ingest Usage Events

Authorizations

Authorization
string
header
required

A control-plane management key using the sk-cp- prefix. Inference keys are not accepted by control-plane endpoints.

Body

application/json
Maximum array length: 1000
event_id
string<uuid>
required

Idempotency key for event ingestion

request_id
string<uuid>
required

Data-plane request ID

request_started_at
string<date-time>
required

Timestamp when the logical request began

occurred_at
string<date-time>
required

Timestamp when the attempt or denial completed

org_id
string<uuid>
required

Organization that made the request

workspace_id
string<uuid>
required

Workspace that made the request

key_id
string
required

Caller credential ID used for the request

Required string length: 1 - 255
request_source
enum<string>
required

Whether the request came from an inference key or a Playground session

Available options:
inference_key,
playground
user_id
string<uuid>
required

Principal that owned the caller credential when the request was made

requested_model_id
string
required

Original caller-requested model before routing and fallback

Required string length: 1 - 255
requested_capabilities
enum<string>[]
required

Original request capabilities before reconciliation

Available options:
tools,
reasoning,
structured_output
model_id
string
required

Caller-facing model ID

Required string length: 1 - 255
bundle_id
string<uuid>
required

Policy bundle used for the request

input_tokens
integer
required

Total input tokens

Required range: 0 <= x <= 2147483647
output_tokens
integer
required

Total output tokens

Required range: 0 <= x <= 2147483647
cost_usd
string
required

Total estimated cost in USD

Pattern: ^\d+(?:\.\d+)?$
max_output_tokens
integer | null
required

Effective upstream output-token limit

Required range: 1 <= x <= 2147483647
latency_ms
integer
required

Gateway latency in milliseconds: per attempt when routed, end-to-end for a denial before routing

Required range: 0 <= x <= 2147483647
stream
boolean
required

Whether the response was streamed

token_usage_source
string
required

No upstream token usage for a request denied before routing

Allowed value: "not_applicable"
schema_version
integer
default:1

Usage event schema version

attempt_started_at
null

No provider attempt was made

provider_id
string
default:""

No provider was selected before denial

Allowed value: ""
cost_input_usd
string
default:0

Estimated input cost in USD

Pattern: ^\d+(?:\.\d+)?$
cost_output_usd
string
default:0

Estimated output cost in USD

Pattern: ^\d+(?:\.\d+)?$
cache_read_tokens
integer
default:0

Input tokens read from a provider cache

Required range: 0 <= x <= 2147483647
cache_write_tokens
integer
default:0

Input tokens written to a provider cache

Required range: 0 <= x <= 2147483647
status
string
default:denied

The request was denied before routing

Allowed value: "denied"
credential_id
null

No provider credential was selected before denial

credential_scope
null

No provider credential scope was selected before denial

Response

Successful Response

data
EventsIngestedOut · object
required