Linux

Prerequisites

Python 3.11+ and git — the one-liners below clone each repo and need both present on PATH; they’ll fail with a clear message naming whichever is missing.

Development setup

Gateway

curl -sSL https://raw.githubusercontent.com/bantex01/VectorStep-Gateway/main/install-gateway.sh | bash

Then edit ~/.vectorstep-gateway/config.yaml with your LLM provider keys, add a first agent under agents/ (see Creating agents), export your provider key, and start it:

cd ~/.vectorstep-gateway
export ANTHROPIC_API_KEY=sk-ant-...
source .venv/bin/activate && python -m gateway.main

# Find your operator token (auto-generated on first run)
cat ~/.vectorstep-gateway/identity/device-auth.json
# Copy the 'operator' token — you'll need it for VectorStep's config

What the one-liner does, or set it up by hand

# 1. Install dependencies
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# 2. Copy and edit the config template
cp samples/config.yaml.example config.yaml
# Edit config.yaml — set your LLM provider keys and MCP servers

# 3. Create your agents directory
mkdir -p agents/my-agent
# Add agent.yaml and soul.md — see Creating Agents below

# 4. Set environment variables for any ${VAR_NAME} placeholders in config.yaml
export ANTHROPIC_API_KEY=sk-ant-...

# 5. Start the gateway (host/port come from config.yaml's `server:` section)
python -m gateway.main

# 6. Find your operator token (auto-generated on first run)
cat ~/.vectorstep-gateway/identity/device-auth.json
# Copy the 'operator' token — you'll need it for VectorStep's config

Both config.yaml and agents/ are gitignored — they contain personal credentials and environment-specific agent definitions. Use samples/config.yaml.example as your starting point.

VectorStep service

curl -sSL https://raw.githubusercontent.com/bantex01/VectorStep/main/install-service.sh | bash

Then start it and try a test webhook:

cd ~/.vectorstep/service
source .venv/bin/activate && uvicorn src.main:app --reload --port 8000

# Test webhook (alertmanager)
curl -X POST "http://localhost:8000/webhook?source=alertmanager" \
  -H "Content-Type: application/json" \
  -d @tests/fixtures/alertmanager_critical.json

# Test webhook (generic source)
curl -X POST "http://localhost:8000/webhook?source=generic" \
  -H "Content-Type: application/json" \
  -d @tests/fixtures/generic_new_order.json

What the one-liner does, or set it up by hand

cd service
python -m venv .venv
source .venv/bin/activate
pip install -r ../requirements.txt   # requirements.txt lives at the repo root

# Run service
uvicorn src.main:app --reload --port 8000

Running the test suite, including the Postgres test lane, is covered on Testing.

SQLite vs. Postgres

The ORM layer (SQLAlchemy async) is dialect-agnostic — switching backends is a database.url change only, no code changes. SQLite (zero infrastructure) is right for this local setup; Postgres is for production.

Note

For the full database setup, migration mechanism, and dedup-race hardening details, see Deployment.

Production (systemd)

For a plain VM or instance without a container runtime, both repos ship systemd units and a step-by-step install guide under deploy/systemd/ (install.md, *.service, env.example) — code in /opt/vectorstep/<service>, config in /etc/vectorstep/<service>/, state in /var/lib/vectorstep/<service>/, logs in /var/log/vectorstep/<service>/. Both units support systemctl reload (pipelines/steps for VectorStep, agents for the Gateway) without dropping the process; a code, dependency, or database change needs systemctl restart.

Where next

• Deployment — the full config.yaml reference and database/migration mechanics.
• Docker / Kubernetes — containerized alternatives to both paths above.