cmind/metering/v1/usage_service.proto

Package: cmind.metering.v1

Messages

GetTokenUsageRequest

Field	Type	Number	Description
tenantIds	repeated string	1	Tenants to include in the aggregate. Empty means all tenants.
start	google.protobuf.Timestamp	2
end	google.protobuf.Timestamp	3
step	string	4	step is the bucket width as a short code (e.g. "5m", "1h", "1d"). The server maps it from a closed set to a ClickHouse interval; unknown codes are rejected.

TokenUsagePoint

TokenUsagePoint is one time bucket's token totals across the selected tenants.

Field	Type	Number	Description
bucket	google.protobuf.Timestamp	1
inputTokens	uint64	2
outputTokens	uint64	3
embeddingTokens	uint64	4
rerankTokens	uint64	5

GetTokenUsageResponse

Field	Type	Number	Description
points	repeated TokenUsagePoint	1

GetModelTokenUsageRequest

Field	Type	Number	Description
tenantIds	repeated string	1	Tenants to include in the aggregate. Empty means all tenants.
start	google.protobuf.Timestamp	2
end	google.protobuf.Timestamp	3
limit	uint32	4	Cap on the number of models returned (highest total first). 0 = no cap.

ModelTokenUsage

ModelTokenUsage is one model's token totals over the range.

Field	Type	Number	Description
model	string	1	The metered model identifier (the usage target).
inputTokens	uint64	2
outputTokens	uint64	3
embeddingTokens	uint64	4
rerankTokens	uint64	5

GetModelTokenUsageResponse

Field	Type	Number	Description
models	repeated ModelTokenUsage	1

GetRequestVolumeRequest

Field	Type	Number	Description
tenantIds	repeated string	1	Tenants to include in the aggregate. Empty means all tenants.
start	google.protobuf.Timestamp	2
end	google.protobuf.Timestamp	3
step	string	4

RequestVolumePoint

RequestVolumePoint is one time bucket's metered-request count.

Field	Type	Number	Description
bucket	google.protobuf.Timestamp	1
requests	uint64	2

GetRequestVolumeResponse

Field	Type	Number	Description
points	repeated RequestVolumePoint	1

GetApiKeyTokenUsageRequest

Field	Type	Number	Description
tenantId	string	1
apiKeyId	string	2
start	google.protobuf.Timestamp	3
end	google.protobuf.Timestamp	4
step	string	5	step is the bucket width as a short code (e.g. "5m", "1h", "1d").

GetApiKeyTokenUsageResponse

Field	Type	Number	Description
points	repeated TokenUsagePoint	1

ExportApiKeyTokenUsageRequest

Field	Type	Number	Description
tenantId	string	1
start	google.protobuf.Timestamp	2
end	google.protobuf.Timestamp	3

ExportApiKeyTokenUsageResponse

Field	Type	Number	Description
filename	string	1
csv	bytes	2

GetApiTrafficRequest

Field	Type	Number	Description
start	google.protobuf.Timestamp	1
end	google.protobuf.Timestamp	2
step	string	3
service	string	4	Restrict to one service (the access log's service name). Empty = all.

ApiTrafficPoint

ApiTrafficPoint is one time bucket of access-log traffic: total requests, client/server error counts and latency percentiles (milliseconds).

Field	Type	Number	Description
bucket	google.protobuf.Timestamp	1
total	uint64	2
clientErrors	uint64	3
serverErrors	uint64	4
p50Ms	double	5
p90Ms	double	6
p99Ms	double	7

GetApiTrafficResponse

Field	Type	Number	Description
points	repeated ApiTrafficPoint	1
services	repeated string	2	Distinct service names seen in the window (unfiltered), for the picker.

GetPerTenantTokenUsageRequest

Field	Type	Number	Description
tenantIds	repeated string	1	Tenants to include. Empty means all tenants.
start	google.protobuf.Timestamp	2
end	google.protobuf.Timestamp	3

TenantTokenUsage

TenantTokenUsage is one tenant's token totals over the range.

Field	Type	Number	Description
tenantId	string	1
inputTokens	uint64	2
outputTokens	uint64	3
embeddingTokens	uint64	4
rerankTokens	uint64	5

GetPerTenantTokenUsageResponse

Field	Type	Number	Description
tenants	repeated TenantTokenUsage	1

Services

UsageService

UsageService is the read side of the metered usage data. The write side is cmind.cluster.metering.v1.InternalUsageService (cluster-internal ingest); this service serves platform-admin analytics over what was ingested.

GetTokenUsage

GET /cmind.metering.v1.UsageService/GetTokenUsage

GetTokenUsage returns token usage bucketed over a time range, summed across the selected tenants (all tenants when tenant_ids is empty). Platform operation.

Request: GetTokenUsageRequest

Response: GetTokenUsageResponse

GetModelTokenUsage

GET /cmind.metering.v1.UsageService/GetModelTokenUsage

GetModelTokenUsage returns per-model token totals over a time range, summed across the selected tenants (all tenants when tenant_ids is empty) and ordered by total tokens descending. Platform operation.

Request: GetModelTokenUsageRequest

Response: GetModelTokenUsageResponse

GetRequestVolume

GET /cmind.metering.v1.UsageService/GetRequestVolume

GetRequestVolume returns the count of metered inference requests bucketed over a time range (distinct trace ids in the usage data), summed across the selected tenants (all tenants when tenant_ids is empty). Platform operation.

Request: GetRequestVolumeRequest

Response: GetRequestVolumeResponse

GetApiKeyTokenUsage

GET /cmind.metering.v1.UsageService/GetApiKeyTokenUsage

GetApiKeyTokenUsage returns token usage bucketed over a time range for one API key. Authorization follows the key so its owner can inspect usage.

Request: GetApiKeyTokenUsageRequest

Response: GetApiKeyTokenUsageResponse

ExportApiKeyTokenUsage

GET /cmind.metering.v1.UsageService/ExportApiKeyTokenUsage

ExportApiKeyTokenUsage returns a backend-generated CSV containing one aggregate row per API key for a tenant and time range.

Request: ExportApiKeyTokenUsageRequest

Response: ExportApiKeyTokenUsageResponse

GetApiTraffic

GET /cmind.metering.v1.UsageService/GetApiTraffic

GetApiTraffic returns API request volume, error counts and latency percentiles bucketed over a time range, from the gateway/service access logs across the whole platform (no tenant dimension). Platform operation.

Request: GetApiTrafficRequest

Response: GetApiTrafficResponse

GetPerTenantTokenUsage

GET /cmind.metering.v1.UsageService/GetPerTenantTokenUsage

GetPerTenantTokenUsage returns per-tenant token totals over a time range (no bucketing — flat totals, one row per tenant). Tenants to include are selected by tenant_ids (empty means all tenants). Platform operation.

Request: GetPerTenantTokenUsageRequest

Response: GetPerTenantTokenUsageResponse