Metis InfluxDB Data Dictionary

Document Information

Version: v1.0 Updated: 2026-09-11

This document is intended for third-party users who query InfluxDB from Grafana to build statistical dashboards. It describes the available measurements, tags, fields, time semantics, and SQL query examples. Environment-specific values such as the endpoint, database name, token, and credentials are provided separately.

1. Connection Parameters

Item	Value
Endpoint	Provided by the deployment environment
Query Path	/api/v3/query_sql
Database	Provided by the deployment environment
Query Precision	nanosecond

2. General Conventions

Item	Description
Time column	InfluxDB SQL queries use time; timestamps are stored with nanosecond precision.
Query language	InfluxDB 3 SQL.
Snapshot measurements	time is the timestamp of the snapshot batch.
Event measurements	time is the event timestamp.
Cumulative measurements	credit_usage_daily and credit_usage_mtd_daily may contain invalidated records; business queries must filter with active = true.
Aggregate queries	If an aggregate query does not select time, do not sort it by time.
Dimension types	All tags are stored as strings. Boolean-like dimensions such as is_public and is_aggregated are also string tags.
Optional columns	Empty fields are not written as empty strings or explicit NULL. A missing column means that the point does not contain that tag or field.

2.1 Instance Monitoring

Instance monitoring measurements provide instance-level, card-level, and hourly aggregate data. deploy_mode distinguishes single-instance, multi-instance, and cluster deployments. is_cluster is a string tag whose value is true or false. Raw monitoring values are stored as received and are not automatically unit-converted.

3. Measurement Overview

Measurement	Description	Time semantics	Recommended Grafana use
tenant_profile_snapshot	Tenant profile snapshot	Snapshot batch time	Query the latest tenant profile.
tenant_credit_snapshot	Tenant credit balance snapshot	Snapshot batch time	Query tenant credit balances.
tenant_overview_snapshot	Tenant overview snapshot	Snapshot batch time	Query tenant profile and credit balance data.
model_source_snapshot	Model provider snapshot	Snapshot batch time	Query model provider information.
model_snapshot	Model snapshot	Snapshot batch time	Query model, status, pricing, and billing configuration.
model_metrics_1m	One-minute model metrics	Metric collection time	Query short-range model metrics.
model_call_event	Model call detail events	Call event time	Query call details and errors.
credit_transaction_event	Credit transaction details	Credit event time	Query individual credit transactions.
credit_usage_daily	Daily credit usage aggregate	UTC day start	Query daily usage; always filter with active = true.
credit_usage_mtd_daily	Month-to-date credit usage aggregate	UTC day start	Query month-to-date usage; always filter with active = true.
model_call_1h	Hourly model call aggregate	UTC hour start	Build long-range call trend dashboards.
model_call_daily	Daily model call aggregate	UTC day start	Build long-range call trend dashboards.
model_metrics_1h	Hourly model metrics aggregate	UTC hour start	Build long-range model metric dashboards.
tenant_model_usage_daily	Daily tenant-model usage aggregate	UTC day start	Build tenant usage dashboards.
provider_model_usage_daily	Daily provider-model usage aggregate	UTC day start	Build provider and model-source usage dashboards.
model_instance_statistics	Instance monitoring data	Metric collection time	Query instance status, CPU, and memory metrics.
model_instance_card_statistics	Card-level instance monitoring data	Metric collection time	Query accelerator-card metrics.
model_instance_statistics_1h	Hourly instance monitoring aggregate	Completed UTC hour start	Build instance trend dashboards.
model_instance_card_statistics_1h	Hourly card-level monitoring aggregate	Completed UTC hour start	Build card-level trend dashboards.
sync_diagnostic	InfluxDB diagnostic data	Diagnostic write time	Do not use for business dashboards.

4. Field Reference

Each tag and field is listed separately. A required field is marked in the Constraint column.

4.1 tenant_profile_snapshot

Category	Field	Description	Type	Constraint
Tag	tenant_id	Tenant ID	String	-
Field	display_name	Tenant display name	String	-
Field	admin_user_id	Tenant administrator user ID	String	-
Field	admin_username	Tenant administrator username	String	-
Field	tenant_type	Tenant type: operator, creator, or eu	String	-
Field	is_superadmin	Whether the tenant is a super-admin tenant	Boolean	-

4.2 tenant_credit_snapshot

Category	Field	Description	Type	Constraint
Tag	tenant_id	Tenant ID	String	-
Field	credit_balance	Tenant credit balance	Double	-
Field	biz_scope_id	Business scope identifier	String	-
Field	category_type	Account category	String	-

4.3 tenant_overview_snapshot

Category	Field	Description	Type	Constraint
Tag	tenant_id	Tenant ID	String	-
Field	display_name	Tenant display name	String	-
Field	admin_username	Tenant administrator username	String	-
Field	credit_balance	Tenant credit balance	Double	-

4.4 model_source_snapshot

Category	Field	Description	Type	Constraint
Tag	model_source_id	Model source or provider ID	String	-
Field	model_source_name	Model source or provider name	String	-
Field	status	Current value is active	String	-

4.5 model_snapshot

Category	Field	Description	Type	Constraint
Tag	model_id	Model ID	String	-
Tag	meta_model_author	Meta-model author identifier	String	-
Tag	model_type	Normalized model type	String	-
Tag	is_public	Visibility: public or private	String	-
Tag	is_aggregated	Whether the model is aggregated	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	model_source_id	Model source or provider ID	String	-
Tag	status	online, scheduled_offline, scheduled_online, or offline	String	-
Field	model_name	Model name	String	-
Field	provider_name	Provider name	String	-
Field	source_model_id	Source model ID	String	-
Field	related_sub_model_ids_csv	Related sub-model IDs in CSV format	String	-
Field	related_sub_model_count	Number of related sub-models	Long	-
Field	price_input_tokens	Input-token unit price	Double	-
Field	price_output_tokens	Output-token unit price	Double	-
Field	price_input_cache_tokens	Cached input-token unit price	Double	-
Field	price_output_cache_tokens	Cached output-token unit price	Double	-
Field	billing_rule_type	Model billing rule type	Long	-
Field	bill_config	Billing configuration JSON used to reconstruct the price matrix	String	-
Field	billing_main_block	Main billing block: token_billing, image_billing, character_billing, duration_billing, or none_token_billing	String	-
Field	billing_price_matrix	Complete price matrix JSON	String	-
Field	billing_price_compare_enabled	Whether price comparison is enabled	Boolean	-
Field	token_cache_enabled	Whether cached-token pricing is enabled	Boolean	-
Field	token_tier_enabled	Whether tiered token pricing is enabled	Boolean	-
Field	token_image_enabled	Whether image-token pricing is enabled	Boolean	-
Field	websearch_billing_enabled	Whether WebSearch billing is enabled	Boolean	-
Field	billing_unit_price	Generic unit price for the main billing block	Double	-
Field	billing_original_unit_price	Generic original unit price	Double	-
Field	billing_unit	Unit: per_1m_tokens, per_image, per_1m_characters, or per_second	String	-
Field	ext_info	Model extension information	String	-
Field	raw_status_code	Raw status code	Long	-

4.6 model_metrics_1m

Category	Field	Description	Type	Constraint
Tag	model_id	Model ID	String	-
Tag	meta_model_author	Meta-model author identifier	String	-
Tag	model_type	Model type	String	-
Tag	is_public	Whether the model is public	String	-
Tag	is_aggregated	Whether the model is aggregated	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	model_source_id	Model source ID	String	-
Tag	provider_name	Provider name	String	-
Tag	model_source_name	Model source name	String	-
Tag	model_name	Model name	String	-
Field	request_count	Request count	Double	-
Field	error_count	Error count	Double	-
Field	success_rate	Success rate	Double	-
Field	input_tokens	Input token count	Double	-
Field	output_tokens	Output token count	Double	-
Field	first_token_use_total_ms	Total time to first token, in milliseconds	Double	-
Field	finish_token_use_total_ms	Total completion time, in milliseconds	Double	-
Field	first_token_use_avg_ms	Average time to first token, in milliseconds	Double	-
Field	finish_token_use_avg_ms	Average completion time, in milliseconds	Double	-
Field	fail_service_error	Service-error count	Double	-
Field	fail_timeout	Timeout count	Double	-
Field	fail_balance	Insufficient-balance count	Double	-
Field	health_status	Health status aggregate	Double	-
Field	operation	Write operation type	String	Required

4.7 model_call_event

Category	Field	Description	Type	Constraint
Tag	model_id	Model ID	String	-
Tag	meta_model_id	Meta-model ID	String	-
Tag	meta_model_author	Meta-model author identifier	String	-
Tag	model_type	Model type	String	-
Tag	is_public	Visibility	String	-
Tag	is_aggregated	Whether the model is aggregated	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	model_source_id	Model source ID	String	-
Tag	call_tenant_id	Calling tenant ID	String	-
Tag	call_user_id	Calling user ID	String	-
Tag	status	success, rate_limited, or fail	String	-
Tag	error_type	Error type	String	-
Tag	source	Call source	String	-
Tag	provider_name	Provider name	String	-
Tag	model_source_name	Model source name	String	-
Tag	model_name	Model name	String	-
Tag	call_tenant_name	Calling tenant name	String	-
Field	request_id	Request ID	String	-
Field	call_user_name	Calling username	String	-
Field	invoke_model_id	Actually invoked model ID	String	-
Field	related_sub_model_ids_csv	Related sub-model IDs	String	-
Field	input_tokens	Input token count	Double	-
Field	output_tokens	Output token count	Double	-
Field	input_cache_tokens	Cached input-token count	Double	-
Field	output_cache_tokens	Cached output-token count	Double	-
Field	first_token_time_ms	Time to first token, in milliseconds	Double	-
Field	finish_token_time_ms	Completion time, in milliseconds	Double	-
Field	call_duration_ms	Total call duration, in milliseconds	Double	-
Field	call_start_time	Call start time	String	-
Field	call_end_time	Call end time	String	-
Field	error_message	Error message	String	-
Field	operation	Write operation type	String	Required

4.8 credit_transaction_event

Category	Field	Description	Type	Constraint
Tag	transaction_type	income or deduction	String	-
Tag	sub_type	free or charge	String	-
Tag	model_id	Model ID	String	-
Tag	meta_model_id	Meta-model ID	String	-
Tag	meta_model_author	Meta-model author identifier	String	-
Tag	model_type	Model type	String	-
Tag	is_public	Visibility	String	-
Tag	is_aggregated	Whether the model is aggregated	String	-
Tag	model_source_id	Model source ID	String	-
Tag	owner_user_id	Owner user ID	String	-
Tag	owner_tenant_id	Owner tenant ID	String	-
Tag	caller_user_id	Calling user ID	String	-
Tag	caller_tenant_id	Calling tenant ID	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	owner_tenant_name	Owner tenant name	String	-
Tag	caller_tenant_name	Calling tenant name	String	-
Tag	provider_name	Provider name	String	-
Tag	model_source_name	Model source name	String	-
Tag	model_name	Model name	String	-
Field	record_id	Credit record ID	String	-
Field	billing_fact_id	Stable billing fact ID	String	-
Field	owner_user_name	Owner username; may be empty	String	-
Field	caller_user_name	Calling username; may be empty	String	-
Field	input_token_credits	Credits for input tokens	Double	-
Field	output_token_credits	Credits for output tokens	Double	-
Field	total_credits	Total credits	Double	-
Field	deducted_credits	Credits actually deducted	Double	-
Field	round_off_credits	Rounded-off credits	Double	-
Field	input_cache_token_credits	Credits for cached input tokens	Double	-
Field	output_cache_token_credits	Credits for cached output tokens	Double	-
Field	cache_hit_tokens	Cached token count	Double	-
Field	total_tokens	Total token count	Double	-
Field	prompt_tokens	Input token count	Double	-
Field	completion_tokens	Output token count	Double	-
Field	pricing_specifications	Pricing rule description	String	-
Field	pricing_snapshot	Billing price snapshot JSON	String	-
Field	usage_detail	Billing usage detail JSON	String	-
Field	free_total_tokens	Free token count	Double	-
Field	charge_total_tokens	Billable token count	Double	-
Field	internal_call	Whether the call is internal	Boolean	-
Field	operation	Write operation type	String	Required

4.9 credit_usage_daily

Category	Field	Description	Type	Constraint
Tag	product_type	Current value is model	String	-
Tag	product_id	Product or model ID	String	-
Tag	tenant_id	Calling tenant ID	String	-
Tag	user_id	Calling user ID	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	meta_model_id	Meta-model ID	String	-
Tag	meta_model_author	Meta-model author identifier	String	-
Field	product_name	Product or model name	String	-
Field	tenant_name	Calling tenant name	String	-
Field	user_name	Calling username	String	-
Field	provider_name	Provider name	String	-
Field	consume_credits_day	Cumulative credits consumed today	Double	-
Field	consume_deducted_credits_day	Cumulative credits actually deducted today	Double	-
Field	consume_round_off_day	Cumulative rounded-off consumption today	Double	-
Field	income_credits_day	Cumulative credits earned today	Double	-
Field	income_deducted_credits_day	Cumulative settled income today	Double	-
Field	income_round_off_day	Cumulative rounded-off income today	Double	-
Field	active	Whether this is the current valid point	Boolean	-

4.10 credit_usage_mtd_daily

Category	Field	Description	Type	Constraint
Tag	stat_month	Statistics month in YYYY-MM format	String	-
Tag	product_type	Current value is model	String	-
Tag	product_id	Product or model ID	String	-
Tag	tenant_id	Calling tenant ID	String	-
Tag	user_id	Calling user ID	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	meta_model_id	Meta-model ID	String	-
Tag	meta_model_author	Meta-model author identifier	String	-
Field	product_name	Product or model name	String	-
Field	tenant_name	Calling tenant name	String	-
Field	user_name	Calling username	String	-
Field	provider_name	Provider name	String	-
Field	consume_credits_mtd	Cumulative credits consumed in the month	Double	-
Field	consume_deducted_credits_mtd	Cumulative credits actually deducted in the month	Double	-
Field	consume_round_off_mtd	Cumulative rounded-off consumption in the month	Double	-
Field	income_credits_mtd	Cumulative credits earned in the month	Double	-
Field	income_deducted_credits_mtd	Cumulative settled income in the month	Double	-
Field	income_round_off_mtd	Cumulative rounded-off income in the month	Double	-
Field	active	Whether this is the current valid point	Boolean	-

4.11 model_call_1h

Category	Field	Description	Type	Constraint
Tag	model_id	Model ID	String	-
Tag	meta_model_id	Meta-model ID	String	-
Tag	meta_model_author	Meta-model author	String	-
Tag	model_type	Model type	String	-
Tag	is_public	Whether the model is public	String	-
Tag	is_aggregated	Whether the model is aggregated	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	model_source_id	Model source ID	String	-
Tag	call_tenant_id	Calling tenant ID	String	-
Tag	status	success, rate_limited, or fail	String	-
Tag	error_type	Error type	String	-
Tag	source	Call source	String	-
Tag	provider_name	Provider name	String	-
Tag	model_source_name	Model source name	String	-
Tag	model_name	Model name	String	-
Tag	call_tenant_name	Calling tenant name	String	-
Field	request_count	Request count	Double	-
Field	success_count	Successful request count	Double	-
Field	error_count	Non-success count	Double	-
Field	rate_limited_count	Rate-limited request count	Double	-
Field	fail_count	Ordinary failure count	Double	-
Field	success_rate	Success rate	Double	-
Field	input_tokens	Total input tokens	Double	-
Field	output_tokens	Total output tokens	Double	-
Field	input_cache_tokens	Total cached input tokens	Double	-
Field	output_cache_tokens	Total cached output tokens	Double	-
Field	avg_first_token_time_ms	Average time to first token	Double	-
Field	avg_finish_token_time_ms	Average completion time	Double	-
Field	avg_call_duration_ms	Average call duration	Double	-

4.12 model_call_daily

The fields are the same as model_call_1h, with the addition of the following tag:

Category	Field	Description	Type	Constraint
Tag	stat_date	Statistics date	String	-

4.13 model_metrics_1h

The dimensions and fields are the same as model_metrics_1m, except that the values are aggregated by UTC hour. The metric fields are:

Category	Field	Description	Type	Constraint
Field	request_count	Request count	Double	-
Field	success_count	Successful request count	Double	-
Field	error_count	Error count	Double	-
Field	success_rate	Success rate	Double	-
Field	input_tokens	Total input tokens	Double	-
Field	output_tokens	Total output tokens	Double	-
Field	first_token_use_total_ms	Total time to first token	Double	-
Field	finish_token_use_total_ms	Total completion time	Double	-
Field	first_token_use_avg_ms	Average time to first token	Double	-
Field	finish_token_use_avg_ms	Average completion time	Double	-
Field	fail_service_error	Service-error count	Double	-
Field	fail_timeout	Timeout count	Double	-
Field	fail_balance	Insufficient-balance count	Double	-
Field	health_status	Health status aggregate	Double	-

4.14 tenant_model_usage_daily

Category	Field	Description	Type	Constraint
Tag	stat_date	Statistics date	String	-
Tag	tenant_id	Calling tenant ID	String	-
Tag	model_id	Model ID	String	-
Tag	meta_model_id	Meta-model ID	String	-
Tag	meta_model_author	Meta-model author	String	-
Tag	model_type	Model type	String	-
Tag	is_public	Whether the model is public	String	-
Tag	is_aggregated	Whether the model is aggregated	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	model_source_id	Model source ID	String	-
Tag	tenant_name	Tenant name	String	-
Tag	provider_name	Provider name	String	-
Tag	model_source_name	Model source name	String	-
Tag	model_name	Model name	String	-
Field	request_count	Request count	Double	-
Field	success_count	Successful request count	Double	-
Field	error_count	Non-success count	Double	-
Field	rate_limited_count	Rate-limited request count	Double	-
Field	fail_count	Ordinary failure count	Double	-
Field	success_rate	Success rate	Double	-
Field	input_tokens	Total input tokens	Double	-
Field	output_tokens	Total output tokens	Double	-
Field	input_cache_tokens	Total cached input tokens	Double	-
Field	output_cache_tokens	Total cached output tokens	Double	-
Field	avg_first_token_time_ms	Average time to first token	Double	-
Field	avg_finish_token_time_ms	Average completion time	Double	-
Field	avg_call_duration_ms	Average call duration	Double	-

4.15 provider_model_usage_daily

This measurement uses provider dimensions instead of tenant dimensions. Its metric fields are:

Category	Field	Description	Type	Constraint
Tag	stat_date	Statistics date	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	model_id	Model ID	String	-
Tag	meta_model_id	Meta-model ID	String	-
Tag	meta_model_author	Meta-model author	String	-
Tag	model_type	Model type	String	-
Tag	is_public	Whether the model is public	String	-
Tag	is_aggregated	Whether the model is aggregated	String	-
Tag	model_source_id	Model source ID	String	-
Tag	provider_name	Provider name	String	-
Tag	model_source_name	Model source name	String	-
Tag	model_name	Model name	String	-
Field	request_count	Request count	Double	-
Field	success_count	Successful request count	Double	-
Field	error_count	Error count	Double	-
Field	success_rate	Success rate	Double	-
Field	input_tokens	Total input tokens	Double	-
Field	output_tokens	Total output tokens	Double	-
Field	first_token_use_total_ms	Total time to first token	Double	-
Field	finish_token_use_total_ms	Total completion time	Double	-
Field	first_token_use_avg_ms	Average time to first token	Double	-
Field	finish_token_use_avg_ms	Average completion time	Double	-
Field	fail_service_error	Service-error count	Double	-
Field	fail_timeout	Timeout count	Double	-
Field	fail_balance	Insufficient-balance count	Double	-
Field	health_status	Health status aggregate	Double	-

4.16 model_instance_statistics

Category	Field	Description	Type	Constraint
Tag	related_model_id	Related model ID	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	model_instance_id	Model instance ID	String	-
Tag	deploy_mode	single_instance, multi_instance, or cluster	String	-
Tag	is_cluster	true for cluster mode; otherwise false	String	-
Tag	type	Metric source type identifier	String	-
Field	related_model_name	Related model name	String	-
Field	provider_tenant_name	Provider tenant name	String	-
Field	model_instance_name	Model instance name	String	-
Field	flavor_cpu	CPU specification	Double	-
Field	flavor_mem	Memory specification	Double	-
Field	flavor_ai_card	Accelerator-card specification	Double	-
Field	flavor_ai_card_type	Accelerator-card type	String	-
Field	status	Instance status value	Double	-
Field	cpu_util	Raw CPU utilization; no automatic unit conversion	Double	-
Field	mem_util	Raw memory utilization; no automatic unit conversion	Double	-

status, cpu_util, and mem_util are merged by timestamp. A point may contain only some of these fields.

4.17 model_instance_card_statistics

Category	Field	Description	Type	Constraint
Tag	related_model_id	Related model ID	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	model_instance_id	Model instance ID	String	-
Tag	card_no	Card number, for example nvidia0	String	-
Tag	deploy_mode	single_instance, multi_instance, or cluster	String	-
Tag	is_cluster	true for cluster mode; otherwise false	String	-
Tag	type	Metric source type identifier	String	-
Field	related_model_name	Related model name	String	-
Field	provider_tenant_name	Provider tenant name	String	-
Field	model_instance_name	Model instance name	String	-
Field	xpu_util	Raw accelerator utilization; no automatic unit conversion	Double	-
Field	xpu_mem_util	Raw accelerator-memory utilization; no automatic unit conversion	Double	-
Field	xpu_gpu_power_usage	Raw GPU power usage; no automatic unit conversion	Double	-
Field	xpu_gpu_temperature	Raw GPU temperature; no automatic unit conversion	Double	-

Different metrics for the same timestamp and card_no are merged into one point.

4.18 model_instance_statistics_1h

Category	Field	Description	Type	Constraint
Tag	related_model_id	Related model ID	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	model_instance_id	Model instance ID	String	-
Tag	deploy_mode	single_instance, multi_instance, or cluster	String	-
Tag	is_cluster	true for cluster mode; otherwise false	String	-
Tag	type	Metric source type identifier	String	-
Tag	meta_model_id	Meta-model ID	String	-
Field	related_model_name	Related model name	String	-
Field	provider_tenant_name	Provider tenant name	String	-
Field	model_instance_name	Model instance name	String	-
Field	flavor_cpu	CPU specification	Double	-
Field	flavor_mem	Memory specification	Double	-
Field	flavor_ai_card	Accelerator-card specification	Double	-
Field	flavor_ai_card_type	Accelerator-card type	String	-
Field	status_last	Last finite status value in the completed UTC hour	Double	-
Field	cpu_util_avg	Average CPU utilization in the completed UTC hour	Double	-
Field	cpu_util_max	Maximum CPU utilization in the completed UTC hour	Double	-
Field	mem_util_avg	Average memory utilization in the completed UTC hour	Double	-
Field	mem_util_max	Maximum memory utilization in the completed UTC hour	Double	-
Field	sample_count	Number of distinct valid sample timestamps in the hour	Long	-

4.19 model_instance_card_statistics_1h

Category	Field	Description	Type	Constraint
Tag	related_model_id	Related model ID	String	-
Tag	provider_tenant_id	Provider tenant ID	String	-
Tag	model_instance_id	Model instance ID	String	-
Tag	card_no	Card number, for example nvidia0	String	-
Tag	deploy_mode	single_instance, multi_instance, or cluster	String	-
Tag	is_cluster	true for cluster mode; otherwise false	String	-
Tag	type	Metric source type identifier	String	-
Tag	meta_model_id	Meta-model ID	String	-
Field	related_model_name	Related model name	String	-
Field	provider_tenant_name	Provider tenant name	String	-
Field	model_instance_name	Model instance name	String	-
Field	xpu_util_avg	Average accelerator utilization in the completed UTC hour	Double	-
Field	xpu_util_max	Maximum accelerator utilization in the completed UTC hour	Double	-
Field	xpu_mem_util_avg	Average accelerator-memory utilization in the completed UTC hour	Double	-
Field	xpu_mem_util_max	Maximum accelerator-memory utilization in the completed UTC hour	Double	-
Field	xpu_gpu_power_usage_avg	Average GPU power usage in the completed UTC hour	Double	-
Field	xpu_gpu_power_usage_max	Maximum GPU power usage in the completed UTC hour	Double	-
Field	xpu_gpu_temperature_avg	Average GPU temperature in the completed UTC hour	Double	-
Field	xpu_gpu_temperature_max	Maximum GPU temperature in the completed UTC hour	Double	-
Field	sample_count	Number of distinct valid sample timestamps for the card in the hour	Long	-

4.20 sync_diagnostic

Category	Field	Description	Type	Constraint
Tag	diag_id	Diagnostic ID	String	Required
Tag	purpose	Diagnostic purpose	String	-
Field	value	Diagnostic value	Double	-
Field	stage	Diagnostic stage	String	-

5. Query Constraints

• Use UTC ISO-8601 for all time parameters, for example 2026-04-29T04:06:15Z.
• Business queries against credit_usage_daily and credit_usage_mtd_daily must include active = true.
• These measurements are credit-usage aggregates. Confirm the business rules before using them as settlement amounts.
• Do not use unfiltered COUNT(*), last(*), or dimension scans as credit-usage metrics.
• For long time ranges, prefer the hourly or daily aggregate measurements in Grafana.
• If an aggregate query does not select time, do not add ORDER BY time.
• Raw model_instance_* values are not automatically unit-converted. Confirm the metric scale before setting thresholds or displaying units.
• model_instance_statistics.status and status_last are metric values, not business status codes.
• model_instance_statistics_1h and model_instance_card_statistics_1h contain only completed UTC hour buckets.
• sample_count is the number of distinct timestamps after merging related metrics. It is not the sample count of one individual metric.
• InfluxDB point identity is determined by the measurement, all tags, and the timestamp. Retrying the same identity overwrites the existing point.

6. SQL Examples

-- Daily credit usage; active points are required
SELECT tenant_id, user_id, product_id,
       SUM(consume_credits_day) AS consume_credits_day,
       SUM(income_credits_day) AS income_credits_day
FROM credit_usage_daily
WHERE active = true
  AND time >= TIMESTAMP '2026-04-01T00:00:00Z'
  AND time <  TIMESTAMP '2026-04-02T00:00:00Z'
GROUP BY tenant_id, user_id, product_id;

-- Month-to-date credit usage; active points are required
SELECT tenant_id, user_id, product_id,
       SUM(consume_credits_mtd) AS consume_credits_mtd,
       SUM(income_credits_mtd) AS income_credits_mtd
FROM credit_usage_mtd_daily
WHERE active = true
  AND stat_month = '2026-04'
GROUP BY tenant_id, user_id, product_id;

-- Prefer aggregate measurements for long-range dashboards
SELECT stat_date, model_id,
       SUM(request_count) AS request_count,
       SUM(error_count) AS error_count
FROM model_call_daily
WHERE time >= TIMESTAMP '2026-04-01T00:00:00Z'
  AND time <  TIMESTAMP '2026-05-01T00:00:00Z'
GROUP BY stat_date, model_id
ORDER BY stat_date;

-- Limit the time range when inspecting call details
SELECT time, request_id, model_id, call_tenant_id, status, error_type, error_message
FROM model_call_event
WHERE time >= TIMESTAMP '2026-04-29T00:00:00Z'
  AND time <  TIMESTAMP '2026-04-30T00:00:00Z'
ORDER BY time DESC
LIMIT 100;

-- Check whether instance metrics exist for each deployment mode
SELECT deploy_mode,
       COUNT(*) AS point_count,
       MAX(time) AS latest_time
FROM model_instance_statistics
WHERE time >= CURRENT_TIMESTAMP - INTERVAL '24 hours'
GROUP BY deploy_mode
ORDER BY deploy_mode;

-- View recent instance metrics and model associations
SELECT time,
       model_instance_id,
       deploy_mode,
       is_cluster,
       related_model_id,
       status,
       cpu_util,
       mem_util
FROM model_instance_statistics
WHERE time >= CURRENT_TIMESTAMP - INTERVAL '24 hours'
ORDER BY time DESC
LIMIT 100;

-- View recent card-level metrics
SELECT time,
       model_instance_id,
       card_no,
       deploy_mode,
       xpu_util,
       xpu_mem_util,
       xpu_gpu_power_usage,
       xpu_gpu_temperature
FROM model_instance_card_statistics
WHERE time >= CURRENT_TIMESTAMP - INTERVAL '24 hours'
ORDER BY time DESC
LIMIT 100;

-- View completed hourly instance buckets
SELECT time,
       model_instance_id,
       deploy_mode,
       status_last,
       cpu_util_avg,
       cpu_util_max,
       mem_util_avg,
       mem_util_max,
       sample_count
FROM model_instance_statistics_1h
WHERE time >= CURRENT_TIMESTAMP - INTERVAL '24 hours'
ORDER BY time DESC
LIMIT 100;

-- View completed hourly card-level buckets
SELECT time,
       model_instance_id,
       card_no,
       deploy_mode,
       xpu_util_avg,
       xpu_util_max,
       xpu_mem_util_avg,
       xpu_mem_util_max,
       xpu_gpu_power_usage_avg,
       xpu_gpu_power_usage_max,
       xpu_gpu_temperature_avg,
       xpu_gpu_temperature_max,
       sample_count
FROM model_instance_card_statistics_1h
WHERE time >= CURRENT_TIMESTAMP - INTERVAL '24 hours'
ORDER BY time DESC
LIMIT 100;