LLM Router — Live Dashboard

Routing at a glance · waiting for data

First-hop success–Fallback depth → p95 response time–95% of measured calls finish within this time → Request guardrail headroom–Configured local daily guardrails → Provider / key issues–Observed failures and account probes →

Waiting for the first metrics update.

Router

–

port –

Requests today

–

– req/min now

Success rate today

–

– errors

Cloud spend today

–

budget –

Latency today — end-to-end response time

–

average end-to-end

Status codes today

Throughput (req/min, last ~10 min)

Local vs Cloud today

Local – Cloud –

– local · – cloud calls

Token throughput today

–

tokens per second

Fallback depth today

–

won on the first hop

Cost today — call history vs budget

Runtime & feed health

Dashboard feed pipeline–

Sources today — what actually served

Captcha — tile-pick lane and paid solvers

Cooldowns — hops resting after 429 / 5xx

Capacity — local GPU nodes

GPU detail — RTX nodes (main + fallback)

Latency by pool / task / caller / node / modelwhich provider is actually fast

Per-task breakdown today–

Docker containers — MAIN onlyloading…

Cloud share today — which pool served

–

Free usage left — local daily guardrails

–

Provider keys — upstream account health

–

Cloud pools — this run, plus wins todayfree fallback chain

Search providers or models Show Sort by

Ready / unused means configured, with no observed attempt this run. Wins confirm a response. Daily caps below are local guardrails; vendor token and rate limits may apply sooner.

#	Pool	Model	Attempts (run)	Wins (run)	Wins (today)	Win share (run)	Success (run)	429/4xx	5xx	Local daily cap	Serving tasks	Status

Cloud calls per hourtoday · restart-safe

Callers today — who is using the router

Caller	Requests	Local / Cloud	Success	Tasks

Failures today — non-2xx calls–