Quickstart
Deploy Baker Street to your Kubernetes cluster in minutes.
Quickstart
This guide gets Baker Street running in your Kubernetes cluster. By the end, you will have a working AI agent with memory, background job processing, and a web UI.
Prerequisites
Before you begin, make sure you have:
- A Kubernetes cluster (Docker Desktop, k3s, minikube, EKS, GKE, or AKS)
- kubectl configured and pointing at your cluster
- Docker for building images (or access to a container registry)
- An Anthropic API key for Claude
- A Voyage AI API key for memory embeddings
- Node.js 20+ and pnpm (for building from source)
Clone the Repository
git clone https://github.com/The-Baker-Street-Project/baker-street.git
cd baker-street
Deploy with the Interactive Script
Baker Street ships with an interactive deployment script that walks you through the entire setup:
scripts/deploy-all.sh
The script handles:
- Prerequisite checks -- verifies kubectl, Docker, and cluster connectivity
- Secret configuration -- prompts for API keys and creates scoped Kubernetes secrets
- TypeScript compilation -- builds all packages and services
- Docker image builds -- builds container images for Brain, Worker, Gateway, and UI
- Kubernetes deployment -- applies manifests via Kustomize
Windows users can use the PowerShell equivalent: scripts/deploy-all.ps1.
Verify the Deployment
Once the script completes, check that all pods are running:
kubectl get pods -n baker-street
You should see pods for the Brain, Worker(s), Gateway, NATS, Qdrant, and the UI. All should show Running status.
NAME READY STATUS RESTARTS
baker-brain-xxx 1/1 Running 0
baker-worker-xxx 1/1 Running 0
baker-gateway-xxx 1/1 Running 0
baker-ui-xxx 1/1 Running 0
nats-0 1/1 Running 0
qdrant-0 1/1 Running 0
Access the Web UI
Port-forward the UI service to your local machine:
kubectl port-forward -n baker-street svc/baker-ui 8080:80
Open http://localhost:8080 in your browser. You should see the Baker Street chat interface.
Your First Conversation
Type a message in the chat input and press Enter. The Brain will:
- Search vector memory for any relevant context (empty on first run)
- Send your message to Claude with the available tool set
- Stream the response back in real time
Try something like:
Hello! My name is Alex and I work on Kubernetes infrastructure.
The agent will respond and store facts about you in memory. In future conversations, it will remember who you are and what you work on.
What's Next?
- Deploy to Kubernetes with Helm for production-ready configuration
- Build an Extension to add custom tools
- Configure Memory to understand how the agent remembers