Struct MetricsResponse 

pub struct MetricsResponse {

    pub uptime_seconds: u64,
    pub total_requests: u64,
    pub total_errors: u64,
    pub rejected_busy: u64,
    pub rejected_unauthorized: u64,
    pub rejected_rate_limited: u64,
    pub in_flight_inferences: usize,
    pub max_concurrent_requests: usize,
    pub total_inferences: u64,
    pub total_batches: u64,
    pub average_batch_size: f64,
    pub max_batch_size: usize,
    pub average_inference_time_ms: f64,
    pub last_inference_at: Option<String>,
    pub request_latency_ms: Percentiles,
    pub inference_latency_ms: Percentiles,
    pub ready: bool,
    pub models: Vec<ModelMetrics>,
    pub llm: Vec<LlmMetrics>,
    pub nerve: Option<NerveMetrics>,
    pub timestamp: String,
}

GET /metrics: server-level counters for monitoring (originally contributed in #32)

Fields

uptime_seconds: u64

total_requests: u64

/predict requests received, including ones rejected before inference

total_errors: u64

/predict requests answered with a 4xx/5xx status

rejected_busy: u64

Requests rejected with 503 SERVER_BUSY

rejected_unauthorized: u64

Requests rejected with 401 UNAUTHORIZED for a missing or unknown API key

rejected_rate_limited: u64

Requests rejected with 429 RATE_LIMITED

in_flight_inferences: usize

Requests queued for or running inference

max_concurrent_requests: usize

--max-concurrent-requests (furnace_max_concurrent_requests in Prometheus)

total_inferences: u64

Requests that reached the model (successful or failed)

total_batches: u64

Forward passes run; with dynamic batching one pass serves several requests

average_batch_size: f64

Items per forward pass, on average

max_batch_size: usize

--max-batch-size; 1 means dynamic batching is off

average_inference_time_ms: f64

last_inference_at: Option<String>

When the last successful inference finished

request_latency_ms: Percentiles

/predict latency from arrival to response, estimated percentiles

inference_latency_ms: Percentiles

Successful inference latency, queueing included, estimated percentiles

ready: bool

Whether every model’s warm-up inference has run (see GET /readyz)

models: Vec<ModelMetrics>

The same per model; the fields above are totals over them

llm: Vec<LlmMetrics>

The LLMs (feature llm), with their own counters; absent without any

nerve: Option<NerveMetrics>

The nervous system: the served models’ pathways, and the neurons added with Server::nerve

timestamp: String

Trait Implementations

impl Debug for MetricsResponse

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for MetricsResponse

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for MetricsResponse

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.