Examples¶

Real, runnable programs that demonstrate JarvisCore in production-grade scenarios. Each example is a complete Python script (or multi-script cluster) in the examples/ directory of the repo: clone it, install the required extras, and run.

Prerequisites¶

The TypeSafe Jev decision demo needs no Redis or generative model. The distributed examples require Redis:

docker compose -f docker-compose.infra.yml up -d  # Redis + optional infra
cp .env.example .env                               # then edit and set API keys
pip install -e ".[redis,prometheus]"

All Examples¶

Example	Profile	Infra	What it teaches
TypeSafe Jev Decisions	`CustomAgent`	TypeSafe API key	Native Choice, Score, and Noul decisions through Mesh injection
Financial Pipeline	`AutoAgent`	Redis	Workflow DAGs, crash recovery, blob storage
Research Network	`AutoAgent`	Redis + P2P	Multi-node SWIM clusters, capability-based dispatch
Support Swarm	`CustomAgent`	Redis + P2P	Mailbox routing, Nexus OSS OAuth, auth injection
Content Pipeline	`CustomAgent`	Redis	Cross-step memory, LTM persistence, pure Python logic
Investment Committee	`AutoAgent` + `CustomAgent`	Redis	Mixed profiles, complex fan-in DAG, multi-agent deliberation

Choosing the Right Starting Point¶

New to JarvisCore?Building distributed systems?Need P2P messaging?Want the full picture?

Start with the TypeSafe Jev decision demo for one agent and no infrastructure, or Financial Pipeline for a durable workflow DAG.

Jump to Research Network. It shows how to run a multi-node cluster where each process contributes different capabilities and the mesh dispatches work automatically.

Support Swarm demonstrates keyword-based routing via the mailbox system, SWIM peer discovery, and the Nexus OSS auth flow for external API calls.

Investment Committee is the most advanced example: it mixes AutoAgent and CustomAgent profiles in a single mesh with a real multi-role deliberation pipeline.

How the Mesh auto-detects infrastructure¶

from jarviscore import Mesh

# The Mesh requires no mode= argument.
# It probes available infrastructure at start() and enables features accordingly.

mesh = Mesh()                        # Minimal: workflow engine only
mesh = Mesh(config={"redis_url": "redis://localhost:6379/0"})  # + Redis persistence
mesh = Mesh(config={"p2p_enabled": True})   # + SWIM discovery + ZMQ messaging

await mesh.start()

# Inspect what was detected
mesh.has_capability("redis")         # True when Redis is reachable
mesh.has_capability("peer_swim")     # True when P2P stack is active
mesh.has_capability("blob")          # True: LocalBlobStorage always available
mesh.has_capability("workflow")      # True: WorkflowEngine always enabled

[!TIP] You can also set REDIS_URL and P2P_ENABLED=true in your .env file: the Mesh reads these automatically, so you don't need to pass them in the config dict.