Filecoin calibration stability console
Live signals, tipset stream, upgrade readiness.
calix.reiers.io

---

Calix is a real-time stability console for the Filecoin calibration network. The audience is the people running calibration at scale: storage providers, protocol engineers, and the network upgrade team. The job is "is the network healthy right now, and what's about to go wrong before the next upgrade?".

Calix is part of the wider calibration stability project at reiers.io. Sister projects:

• Plumbline — calibration faucet for tFIL + USDFC
• The Doctor calibration cluster

What's on the dashboard

Status header

A sticky bar across the top.

• Operational pill — OPERATIONAL / WATCH / DEGRADED / UPGRADE PENDING, computed live from chain head age and proximity to the next network upgrade.
• Live epoch, head age, network version, and current blocks-per-epoch cadence.

Status banner

First panel on the page. Spells out the current operational level in plain English with a reason. For example, during the run-up to Fire Horse: "Fire Horse upgrade approaching. Network version 28 activates in 1d 20h."

Vital signs (8 KPIs)

Eight cards in a grid, each with an inline 60-epoch sparkline. Each card flips colour (green / amber / red) as the underlying value moves through its thresholds.

• Blocks per epoch — calibration steady state is around 2.0
• Null round % — count of empty tipsets in the window
• Average win count — drand winning tickets per epoch
• Base fee — in atto, with K/M scaling for spikes
• Network QAP — quality-adjusted power
• Active miners — those above min-power
• Total pledge — locked FIL across the network
• Initial pledge per 32 GiB CC — computed locally via the actor v17 formula

Tipset stream

Sixty vertical bars covering the last 30 minutes of chain history.

• Green when 3+ blocks landed (healthy)
• Amber for 1–2 (degraded)
• Red for 0 (null round) Hover any bar for epoch · block count · timestamp.

Upgrade readiness

Two cards: the current network version (nv28 Fire Horse, activated epoch 3,694,534) and a live countdown to the next upgrade: nv29 Solstice, epoch 4,109,133, 2026-09-28T12:59:30Z (window 12:59-13:59 UTC). Driven by one schedule table (calibUpgrades in api/main.go, UPGRADES in web/index.html) that flips automatically at activation. /api/v1/upgrade returns current, next and the full upgrades list.

Network power chart

Twenty-four-hour QAP trend, drawn as a native SVG line + area. Redraws on window resize so the geometry stays in real pixels.

Leaderboards

• Top miners by power — raw, QAP, 24-hour delta, blocks mined, and on-chain tags (Curio Storage Inc., etc.)
• Rich list — top accounts by balance, with actor type and percent of supply

Ecosystem

• FEVM — daily contract creates, invocations, gas
• Plumbline — calibration faucet card
• Open source — link to this repository

Tools

Bidirectional epoch ↔ time converter. Calibration genesis 1667326380, 30-second epochs.

Architecture

calix.reiers.io   (Cloudflare → nginx)
  ├── /          → static dashboard (web/index.html)
  └── /api/v1/*  → calix-api (Go, 127.0.0.1:8765)
                    ├── reads f04 (power) + f02 (reward) actor state
                    ├── samples last 60 tipsets via ChainGetTipSetByHeight
                    ├── computes IP via the actor v17 formula
                    ├── classifies status from head age and upgrade window
                    └── proxies + caches a handful of filfox endpoints

A single Go binary (~7 MB, no cgo, no third-party deps) plus one static HTML file (~50 KB, no framework, no bundler). Every endpoint is cached server-side with stale-on-error fallback so the dashboard never goes blank when an upstream rate-limits.

Endpoints

GET /api/v1/health           Health + cache freshness
GET /api/v1/version          Calix build info
GET /api/v1/status           Operational status with reason
GET /api/v1/signals          Vital signs (KPIs + 60-epoch sparkline series)
GET /api/v1/tipsets/recent   Last 60 tipsets with block count + win sum
GET /api/v1/upgrade          Next upgrade source of truth
GET /api/v1/miners/top       Top miners by power
GET /api/v1/rich-list        Rich list
GET /api/v1/power-history    Power history (?duration=24h | 7d | 30d)
GET /api/v1/fevm-stats       FEVM daily statistics
GET /api/v1/faucet           Plumbline faucet metadata

Run locally

# API
cd api && CALIX_ADDR=:8765 go run .

# Dashboard
cd web && python3 -m http.server 8766
# open http://127.0.0.1:8766

Deploy

./deploy/deploy.sh

Builds linux/amd64, rsyncs to the host, sets up the systemd unit, symlinks the nginx server block, fixes file modes for www-data, and reloads.

Configurable via env: CALIX_ADDR, CALIX_LOTUS_RPC, CALIX_FILFOX_API, CALIX_CORS, CALIX_FAUCET_URL.

What's coming

• Per-miner upgrade readiness — the chain doesn't expose Lotus version directly, but for Storage Providers running Curio we can correlate via the cluster webrpc. A "X of N miners are on Curio v1.28" tile is the single most useful Fire Horse readiness signal.
• Chain-weight / fork visualisation — calibration occasionally forks; weight divergence over time is a high-signal stability metric.
• Cluster lens behind auth — sealing tasks, recent failures, machines online. Sourced from a Curio cluster's webrpc through a tunnel.
• Alerts — a /api/v1/alerts surface so other tooling can poll Calix; later, push to Telegram / matrix / webhook.

License

MIT, see LICENSE.

Chain finality

Samples Filecoin.ChainGetTipSetFinalityStatus (Lotus v2 API, CALIX_LOTUS_RPC_V2, defaults to the v1 URL with /rpc/v2) every 30s. Shows the EC probabilistic finality depth (FRC-0089, 2^-30 reorg bound; -1 = bound not met), the F3 finalized depth, the depth the node actually treats as final and which mechanism sets it, plus a 2h history. When the status call returns no F3 tipset, Calix retries Filecoin.F3GetLatestCertificate over v1, because Glif's load-balanced backends intermittently fail to reach F3 and would otherwise report false F3 outages. Served at /api/v1/finality; exported as calix_finality_* in /metrics. Suggested by rvagg.