Configuration

Curio Core's runtime configuration lives in three places:

1. CLI flags — passed to curio-core run
2. harmony_config table — set via the /setup wizard or future curio-core config set CLI
3. Environment variables — see CLI reference

CLI flags

See the CLI reference for the full list.

harmony_config rows

The state DB holds a single TOML config bundle:

[Pdp]
MarketAddress = "0x..."
WalletAddress = "0x..."
MinerID       = "f0..."

These are required identifiers for the SP. The /setup wizard at http://127.0.0.1:14994/setup writes the row; you can also edit it via:

-- Inspect
SELECT title, substr(config, 1, 300) FROM harmony_config;

-- Update (be careful; restart curio-core after)
UPDATE harmony_config SET config = '<new-toml>' WHERE title = 'base';

Changes take effect on the next curio-core run restart.

Network selection

--network calibration or --network mainnet controls:

• Which contract addresses Curio Core resolves (PDPVerifier, FilecoinPay, USDFC)
• Which DRAND chain Lantern verifies against
• Which Filecoin gateway the FEVM bridge talks to (glif's calibration vs mainnet RPC)
• Which Lantern bootstrap quorum is used

Calibration default RPC: https://api.calibration.node.glif.io/rpc/v1 Mainnet default RPC: https://api.node.glif.io/rpc/v1

Override either via --vm-bridge-rpc <url>.

Resource budget

The harmonytask engine's resource budget gates which tasks can run concurrently:

curio-core run --engine-cpu 4 --engine-ram 4GiB

Defaults are tuned for a laptop. For a small-VPS production SP, 4 CPU / 4 GiB is fine. For an SP that handles heavy concurrent client uploads, bump RAM to 8 GiB+.

If a task type's declared Cost.RAM exceeds the engine budget, the task sits queued forever and the dashboard's Tasks page shows it stuck. Curio Core logs a WARN line when this happens (see upstream PR #1245 follow-on / curio-core#24).