DocsQuick Start

Quick Start

Get your first agent running in under 60 seconds.

Prerequisites

  • Python 3.10 or higher
  • 6GB disk space (for the AI model)
  • 8GB RAM minimum (16GB recommended)
  • macOS, Linux, or Windows (WSL)
1

Install tp-ops

Install the TernaryPhysics CLI using pip:

Terminal
$ pip install tp-ops

This installs the tp-ops command-line tool.

2

SSH to your resource

Connect to the server, VM, or machine where you want to drop an agent:

Terminal
$ ssh my-server.example.com

Agents run on the resource they monitor. This is what makes them so effective — they have direct access to logs, metrics, and system state.

3

Drop an agent

Run tp-ops drop on the resource. The agent auto-detects what type of resource it's running on:

Terminal
$ tp-ops drop

Detecting resource type...
✓ Detected: Kubernetes cluster (kubectl configured)
✓ Agent type: k8s-agent

Downloading TernaryPhysics-7B model...
████████████████████████████████ 100% (4.2GB)

Checking network requirements...
✓ Mesh discovery port - open
✓ Mesh messaging port - open

Agent dropped successfully!
Name: prod-cluster
Type: k8s-agent
Status: active

First run: The TernaryPhysics-7B model (~4GB) downloads once per resource. Subsequent runs start instantly.

4

Talk to your agent

Start a conversation with your agent using natural language:

Interactive session
$ tp-ops ask prod-cluster

prod-cluster > What's the cluster health?

Checking cluster health...

Cluster: prod-cluster
Nodes: 6/6 healthy
Pods: 142 running, 3 pending, 0 failed
CPU: 45% average utilization
Memory: 62% average utilization

No critical issues detected.

prod-cluster > Why are 3 pods pending?

Investigating pending pods...

3 pods pending due to resource constraints:
├─ payment-api-7d9f8 (needs 2Gi memory, node has 1.5Gi)
├─ worker-batch-3k2j (needs GPU, no GPU nodes available)
└─ cache-warm-x92m (node selector: zone=us-east-1a, no nodes match)

Recommendation: Scale node pool or adjust resource requests.
Show detailed fix options? [yes/no]

What's next?