Skip to main content
The tfy-llm-gateway provides detailed, Prometheus-compatible metrics to monitor the health, performance, and cost of your LLM applications with Grafana.
You will be able to export metrics only if the gateway is hosted on your end.

​
Setup

The gateway exposes a /metrics endpoint that can be scraped by your Prometheus instance. This is the standard way to collect metrics. Alternatively, if your setup uses an OpenTelemetry Collector, you can configure the gateway to push metrics directly. To do this, set the following environment variables for the tfy-llm-gateway service:
  • ENABLE_OTEL_METRICS: Set to "true".
  • OTEL_EXPORTER_OTLP_METRICS_ENDPOINT: The endpoint of your OTEL metrics exporter.
  • OTEL_EXPORTER_OTLP_METRICS_HEADERS: (Optional) Headers for authentication.

​
Example Push Configuration

ENABLE_OTEL_METRICS: 'true'
OTEL_EXPORTER_OTLP_METRICS_ENDPOINT: https://<prometheus-host>/api/v1/otlp/v1/metrics
OTEL_EXPORTER_OTLP_METRICS_HEADERS: 'Authorization=Bearer <your-token>'
LLM_GATEWAY_METADATA_LOGGING_KEYS: '["customer_id", "request_type"]'

​
Labels

Labels provide dimensions for filtering and aggregating metrics.
LabelDescription
model_nameThe name of the model used for the request (e.g., gpt-4o).
tenant_nameThe name of the tenant associated with the request.
usernameThe user associated with the request.
tool_nameThe name of the tool called by an agent (only for agent metrics).
llm_gateway_metadata_*Custom labels generated from LLM_GATEWAY_METADATA_LOGGING_KEYS. For example, customer_id becomes llm_gateway_metadata_customer_id.

​
Gateway Metrics

These metrics provide an overview of the gateway’s performance and usage.

​
Token Usage and Cost

Metric NameTypeDescriptionLabels
llm_gateway_input_tokensCounterThe number of input tokens processed.model_name, tenant_name, username, llm_gateway_metadata_*
llm_gateway_output_tokensCounterThe number of output tokens generated.model_name, tenant_name, username, llm_gateway_metadata_*
llm_gateway_request_total_costCounterThe estimated cost of the tokens used.model_name, tenant_name, username, llm_gateway_metadata_*

​
Latency

Metric NameTypeDescriptionLabels
llm_gateway_request_processing_msHistogramThe total time taken to process a request.model_name, tenant_name, username, llm_gateway_metadata_*
llm_gateway_first_token_latency_msHistogramThe time to receive the first token from the model.model_name, tenant_name, username, llm_gateway_metadata_*
llm_gateway_inter_token_latency_msHistogramThe average time between subsequent tokens in a stream.model_name, tenant_name, username, llm_gateway_metadata_*

​
Errors and Failures

Metric NameTypeDescriptionLabels
llm_gateway_request_model_inference_failureCounterThe number of failed model inference requests.model_name, tenant_name, username, llm_gateway_metadata_*
llm_gateway_config_parsing_failuresCounterThe number of configuration parsing errors.model_name, tenant_name, username, llm_gateway_metadata_*

​
Configuration Metrics

Metric NameTypeDescriptionLabels
llm_gateway_rate_limit_requests_totalCounterTotal number of requests that hit rate limits.model_name, tenant_name, username, ruleId, llm_gateway_metadata_*
llm_gateway_load_balanced_requests_totalCounterTotal number of requests that were load balanced.model_name, tenant_name, username, ruleId, llm_gateway_metadata_*
llm_gateway_fallback_requests_totalCounterTotal number of requests that were served by fallback.model_name, tenant_name, username, ruleId, llm_gateway_metadata_*
llm_gateway_budget_requests_totalCounterTotal number of requests that hit budget limits.model_name, tenant_name, username, ruleId, llm_gateway_metadata_*
llm_gateway_guardrails_requests_totalCounterTotal number of requests that hit guardrails.model_name, tenant_name, username, ruleId, llm_gateway_metadata_*

​
Agent-Specific Metrics

These metrics are for monitoring agent-based interactions.
Metric NameTypeDescriptionLabels
llm_gateway_agent_request_duration_msHistogramThe total duration of an agent request.model_name, status, tenant_name, username
llm_gateway_agent_request_tool_calls_totalHistogramThe number of tool calls in a single agent request.model_name, tenant_name, username
llm_gateway_agent_request_iteration_limit_reached_totalCounterThe number of requests hitting the iteration limit.model_name, iteration_limit, tenant_name, username
llm_gateway_agent_llm_latency_msHistogramThe latency of LLM calls within an agent.model_name, iteration_number, status, tenant_name, username
llm_gateway_agent_tool_calls_totalCounterThe total number of tool calls made by agents.tool_name, integration_fqn, tenant_name, username
llm_gateway_agent_tool_latency_msHistogramThe execution time of each tool call.tool_name, integration_fqn, status, tenant_name, username
llm_gateway_agent_mcp_connect_latency_msHistogramThe time taken to connect to an MCP server and fetch tools.integration_fqn (or server_url), tenant_name, username

​
Grafana Integration

Our pre-built Grafana dashboard helps you monitor the LLM Gateway. It is organized into several views so you can analyze gateway activity from different perspectives.
Note
You can find the Grafana Dashboard JSON at the following link:
https://github.com/truefoundry/infra-charts/blob/main/charts/tfy-grafana/dashboards/llm-gateway-metrics.json

​
Dashboard Views

Our pre-built Grafana dashboard is organized into several views to help you analyze gateway activity from different perspectives:
  • Model View: Groups metrics by model_name to compare model performance.
  • User View: Groups metrics by username to monitor usage patterns.
  • Config View: Groups metrics by ruleId to show the impact of gateway configurations.
  • MCP Invocation Metrics: Contains all agent-related metrics, including MCP server performance and tool call latency.
Grafana dashboard showing model performance metrics including token usage, cost, and latency Grafana dashboard showing agent and MCP server metrics including tool calls and latency

​
Importing the Dashboard

  1. Copy the JSON dashboard definition below.
  2. In your Grafana instance, navigate to Dashboards > Import.
  3. Paste the JSON into the Import via panel json text area.
  4. Click Load.
  5. On the next screen, select your Prometheus data source.
  6. Click Import.

​
Customizing the Dashboard

The pre-built dashboard includes filters for model_name, tenant_name, and username. If you use custom metadata labels (via LLM_GATEWAY_METADATA_LOGGING_KEYS), you can add them as filters to your dashboard for more granular analysis.

​
Add a Dashboard Variable

For example, to filter by a custom metadata key like customer_id, add a new variable to your dashboard settings:
  1. Go to Dashboard settings > Variables.
  2. Click New variable.
  3. Configure the variable as follows:
    • Name: customer_id
    • Type: Query
    • Label: Customer ID
    • Data source: Your Prometheus source
    • Query: label_values(llm_gateway_input_tokens, llm_gateway_metadata_customer_id)
    • Multi-value: Enabled
    • Include All option: Enabled

​
Use the Variable in Queries

Update your panel queries to use the new variable. For example, to filter input tokens by customer_id: 
sum(rate(llm_gateway_input_tokens{model_name=~"$model_name", tenant_name=~"$tenant_name", llm_gateway_metadata_customer_id=~"$customer_id"}[5m]))