docs

// the cli is the surface; here's the cookbook

[ install ]

curl -fsSL https://c0mpute.com/install.sh | sh

Drops three binaries into ~/.c0mpute/bin: c0mpute, coinpay, infernet. Self-upgrades on its own schedule.

Variants:

# c0mpute only (no coinpay, no infernet)
curl -fsSL https://c0mpute.com/install.sh | sh -s -- --minimal

# Worker box: also installs Docker + FFmpeg readiness checks
curl -fsSL https://c0mpute.com/install.sh | sh -s -- --worker

# Reinstall over an existing install
curl -fsSL https://c0mpute.com/install.sh | sh -s -- --force

[ identity & wallet ]

Every actor on the network has a CoinPay DID. Create one before submitting jobs or registering a worker.

c0mpute coinpay did create                  # creates did:coinpay:user:<id>
c0mpute coinpay did create --role worker    # creates a worker DID
c0mpute coinpay did status                  # show active DID + key fingerprint
c0mpute coinpay wallet status               # linked addresses + balances

[ run a worker ]

Register, then start. --gpu opts into transcode + inference roles; storage / gateway roles run on any box.

c0mpute worker register
c0mpute worker start \
  --roles storage,transcode,gateway,verifier \
  --storage 500GB \
  --gpu

# Inspect status / restart / stop
c0mpute worker status
c0mpute worker stop

[ submit a transcode job ]

The transcode plugin handles FFmpeg workloads — H.264 / HEVC / AV1 with hardware acceleration where available. Validation is done via ffprobe + (optionally) VMAF.

# 1080p H.264 four-rendition HLS bundle
c0mpute transcode submit input.mov \
  --preset hls \
  --max-price 1.25

# AV1 4K transcode capped at $5
c0mpute transcode submit input.mp4 \
  --preset video-4k \
  --max-price 5.00

# List available presets
c0mpute transcode preset list

Job manifest schema (sent to workers, signed by your DID):

{
  "version": "0.1",
  "network": "c0mpute",
  "type": "ffmpeg.transcode",
  "buyer": "did:coinpay:buyer:abc",
  "input": {
    "uri": "https://your-storage/input.mov",
    "sha256": "sha256:..."
  },
  "runtime": {
    "image": "ghcr.io/c0mpute/ffmpeg-runner@sha256:...",
    "command": ["transcode", "--preset", "hls"]
  },
  "output": {
    "format": "hls",
    "requirements": {
      "videoCodec": "h264",
      "audioCodec": "aac",
      "maxWidth": 1920,
      "maxHeight": 1080
    }
  },
  "payment": { "escrow": "coinpay", "maxPriceUsd": 1.25 },
  "validation": {
    "mode": "ffprobe",
    "checks": ["duration", "codec", "resolution", "bitrate"]
  }
}

[ run AI inference ]

The infernet plugin runs LLM and other ML inference on c0mpute workers' GPUs. Validation uses output-schema checks + spot-check duplicate execution.

# Simple batch with default model
c0mpute infernet run prompts.jsonl \
  --model qwen \
  --max-price 5.00

# Pin a specific model + runtime image hash for reproducible runs
c0mpute infernet run prompts.jsonl \
  --model llama-3.1-8b \
  --max-price 0.25

# List models the network advertises
c0mpute infernet models list

# Benchmark a model against the network's workers
c0mpute infernet benchmark --model qwen

[ track jobs ]

c0mpute job status <job-id>          # one-shot status
c0mpute job logs <job-id> --follow   # tail logs
c0mpute job cancel <job-id>          # cancel queued/running
c0mpute tui                          # interactive worker / job dashboard

[ inspect trust / reputation ]

Every worker, validator, buyer, and org has badges anchored to their DID. No opaque trust score; surface the metrics.

c0mpute coinpay reputation inspect did:coinpay:worker:def456

Sample output:

did: did:coinpay:worker:def456
  431 completed jobs
  98.7% validation success
  $500 staked
  KYC verified (optional)
  H100 attested (optional)
  No slashing events in 90 days

[ install third-party plugins ]

# Install from a plugin's signed install.sh
c0mpute plugin install https://example.com/my-plugin/install.sh

# Install by id from the c0mpute marketplace
c0mpute plugin install my-plugin

# List installed
c0mpute plugin list

# Disable / re-enable / remove
c0mpute plugin disable my-plugin
c0mpute plugin enable my-plugin
c0mpute plugin uninstall my-plugin

Browse what's available at /plugins.

[ self-check ]

c0mpute doctor walks the full stack — binary presence, FFmpeg, Docker, GPU drivers, disk, clock drift, coordinator reach, and the known-issues feed.

c0mpute doctor                         # full report
c0mpute doctor --fix                   # auto-apply known remediations
c0mpute doctor --report                # send anonymized telemetry

[ design proposals ]

Architecture decisions live in the dips/ directory. Highlights:

  • DIP-0005 — c0mpute rebrand, three-CLI architecture
  • DIP-0006 — module manifest + dispatch model
  • DIP-0007 — CoinPay DID as the identity layer
  • DIP-0008 — CLI-first UI strategy
  • DIP-0010 — libp2p bootstrap seed nodes
  • DIP-0011 — no central backend

→ getting-started · plugins · contact