Production Examples

Seven standalone programs demonstrating every Medici protocol operation, in three languages across three market regimes.

Three Market Regimes

All seven examples encode market posture through a single parameter: strikeRatio. This controls how the P (call) and N (put) token value splits at inception — lower values are defensive, higher values are aggressive. There is no explicit "direction" field; the regime is expressed purely through strikeRatio, maturity duration, rebalance thresholds, and stop-loss bands.

How P/N Manipulation Works

The protocol creates both P and N tokens simultaneously from collateral when a SplitRequest executes. The value each side receives at settlement is determined by strikeRatio (strike price divided by spot price):

• Lower strikeRatio (e.g. 0.35): more value flows to N (puts) — defensive, downside protection
• Balanced strikeRatio (e.g. 0.50): equal split — neutral, treasury management
• Higher strikeRatio (e.g. 0.75): more value flows to P (calls) — aggressive, upside capture

At settlement, the DAML payout functions guarantee: P_share = strikeRatio, N_share = 1 − strikeRatio, with the invariant P + N = collateral always holding.

The Parameter SDK constrains strikeRatio to [0.01, 1.0]. Values are validated before any on-chain action.

For the full strikeRatio mechanics, SDK constraints, and DAML settlement math, see the Strategy Engine guide.

Examples

Deployment Status

Running 24×7 in medici-dev-examples

These examples run as persistent Kubernetes Deployments against the dev Canton participant, one per regime per example (9 pods total):

All deployed via FluxCD from the medici-examples repo.

Live dashboard: simulation.medici.loan — the Simulation Monitor shows bot liveness, identity summaries, intent management, and a real-time activity feed.

On-Demand (CronJob-verified daily)

These run as daily Kubernetes CronJobs that verify the full flow end-to-end, pushing metrics to a Prometheus Pushgateway for alerting:

Each CronJob pushes a verification metric on completion:

medici_example_verification{example="intent-cli", language="ts", status="success"} 1

Alertmanager fires if status="failure" for any CronJob in the last 24 hours.

Quick Start

1. Provision demo identities (one-time)

# Creates 6 Keycloak users + onboards each one via the Ledger Service.
# Idempotent — safe to re-run.
LEDGER_URL=https://ledger.dev.medici.loan \
KC_URL=https://keycloak.dev.medici.loan/auth \
KC_ADMIN_USER=admin \
KC_ADMIN_PASSWORD=<from keycloak-credentials secret> \
KC_CLIENT_SECRET=<from medici-dev-examples-secrets> \
KC_REALM=AppUser \
PACKAGE_ID='#option-index-tracker-v10' \
bash scripts/bootstrap-dev-simulation.sh

The bootstrap provisions these identities:

2. Run any example

# Each example authenticates via password grant, onboards (gets its own
# Canton party + rights), and operates as that party.

# Clone the examples repo
git clone https://github.com/jojig-dao/medici-examples
cd medici-examples

# Examples are grouped by language: go/ ts/ python/ shell/
cd go/treasury-bot
KC_USERNAME=alice_treasury KC_PASSWORD=def456 go run main.go

cd ../../ts/treasury-bot
KC_USERNAME=alice_treasury KC_PASSWORD=def456 npx tsx main.ts

cd ../../python/treasury-bot
KC_USERNAME=bob_hedge KC_PASSWORD=def456 python main.py

All examples share the same environment variable names:

Each regime variant accepts optional overrides: STRIKE_RATIO, MATURITY_DAYS, REBALANCE_THRESHOLD, STOP_LOSS_DRIFT_BPS, INTERVAL_SECONDS.

Shared Library

Each language directory provides shared plumbing so examples focus on protocol logic, not auth and HTTP boilerplate:

Each shared library handles:

• Keycloak client_credentials token acquisition with automatic refresh
• HTTP client setup with default timeouts
• Canton JSON API response normalisation (payload unwrapping, error-in-body detection)
• Structured error types with status code and operation context

Using the shared library eliminates ~60% of the auth/retry boilerplate that would otherwise be duplicated across every example. See each example's source for usage patterns.