Skip to content

quickstart.md

Latest commit

 

History

History
64 lines (47 loc) · 1.83 KB

quickstart.md

File metadata and controls

64 lines (47 loc) · 1.83 KB
Raw

Quick start

Requirements

  • Python 3.11 or newer
  • curl for the one-command installer
  • An optional OpenAI-compatible model server for AI nodes

Install

curl -fsSL https://raw.githubusercontent.com/ActivLayer/activlayer/community-edition/install.sh | sh
activlayer --version

The installer uses an isolated environment and makes activlayer available from any directory. If it adds ~/.local/bin to your shell profile, open a new terminal once after installation.

Initialize

activlayer init \
  --organization "Example Organization" \
  --owner owner@example.com

Initialization automatically discovers Ollama, vLLM, and llama.cpp on their standard local ports and on non-default ports declared by running vLLM or llama.cpp processes. It configures every provider that reports a model and selects the first one. If none is running, continue with manual configuration below. Use --no-detect-llm to skip discovery.

Configure a model

activlayer llm add local --type ollama --model qwen3:8b
activlayer llm list
activlayer llm select <displayed-provider-name>
activlayer llm test

AI nodes require a model provider. Trigger, rule, approval, HTTP, decision, and output nodes run without one.

Provision and run

activlayer agent provision examples/request_review.json --publish
activlayer agent graph request-review --published

activlayer run start request-review \
  --input '{"request":"Review the weekly report"}' \
  --permission requests.approve \
  --actor owner@example.com

The run processes the trigger and AI node, then pauses durably at the approval node.

activlayer run list
activlayer run show <run-id> --events
activlayer run approve <run-id> --actor owner@example.com --reason "Checked"

Continue with the CLI guide and Agent JSON reference.