Dashboard Get started

Prometheus metrics

Scrape TokenID cost and token counters into your existing Grafana / Prometheus stack.

Endpoint

GET/metrics

Returns Prometheus text exposition format scoped to the authenticated organization. Authentication is the same Bearer token used elsewhere — pass it in your Prometheus scrape config.

What's exposed

Metric Type Labels
llm_tokens_total counter provider, model, org_id
llm_cost_usd_total counter provider, model, org_id
llm_requests_total counter provider, model, org_id
llm_cost_usd_rolling_24h gauge org_id

llm_tokens_total is the sum of input_tokens + output_tokens across all api_call events for the org. llm_cost_usd_rolling_24h is a 24-hour rolling window so it's well-suited as an alerting input.

Example scrape config

scrape_configs:
  - job_name: tokenid
    metrics_path: /metrics
    scheme: https
    static_configs:
      - targets: ['token-api.audit.id']
    authorization:
      type: Bearer
      credentials: 'eyJhbGciOi…'   # your TokenID API token
    scrape_interval: 60s

Endpoint details

GET/metrics
Property Value
Auth Authorization: Bearer td_live_xxxx — same token as the rest of the API
Scope Per-organization — the response only contains metrics for the calling org
Content type text/plain; version=0.0.4; charset=utf-8 (Prometheus text exposition)
Suggested scrape interval 60s (computed on demand from event-log aggregates)

Metric reference

Metric Type Labels Meaning
llm_requests_total counter provider, model, org_id Cumulative number of api_call events
llm_tokens_total counter provider, model, org_id Cumulative input_tokens + output_tokens
llm_cost_usd_total counter provider, model, org_id Cumulative cost in USD
llm_cost_usd_rolling_24h gauge org_id Sum of cost over the trailing 24 hours

All labels are emitted lowercase. Unknown values (an event without a model) are dropped from the per-(provider, model) counters — they never show up as model="unknown" blowing up your cardinality.

Sample prometheus.yml

scrape_configs:
  - job_name: tokenid
    metrics_path: /metrics
    scheme: https
    static_configs:
      - targets: ['token.audit.id']
    bearer_token: td_live_xxxx
    scrape_interval: 60s

For Kubernetes-hosted Prometheus, store the token in a Secret and reference it via bearer_token_file:

    bearer_token_file: /etc/prometheus/secrets/tokenid/api-key

Alerting on burn rate

groups:
  - name: tokenid
    rules:
      - alert: TokenIDDailyBurnHigh
        expr: llm_cost_usd_rolling_24h > 500
        for: 5m
        annotations:
          summary: "TokenID 24h spend exceeded $500"

This complements the in-product Spend Alerts feature — Prometheus alerting is the right channel for ops teams who already run a Grafana / Alertmanager stack and want LLM cost in the same surface.