🌿 Orbisage

Orbisage is a friendly, practical starter template for exploring how you can build a multi-agent AI router using LangChain and LangGraph.

---

Orbisage lets you experiment with how an AI bot can greet you, guide you around a building, tell you the latest news, share a fun fact, or even make you laugh — all routed through a simple, visual graph. It’s small enough to learn in an afternoon but flexible enough to grow into real-world workflows.

---

✨ How It Works

Instead of tangled if/else statements, Orbisage uses a directed graph:

• Nodes: Each agent is a Python function.
• Edges: Conditional transitions that inspect your input.
• State: Keeps your conversation context as you move through nodes.

---

📦 Quick Start

Requirements

• Python 3.9+
• (Optional) OpenAI API Key for real LLM calls

Install & Setup

git clone git@github.com:semework/OrbiSage.git
cd OrbiSage

python -m venv .venv && source .venv/bin/activate

pip install -e .[dev]

cp .env.example .env  # Add OPENAI_API_KEY if you want

Verify

python -c "from orbisage_router.graph import build_router_graph; print(build_router_graph())"

---

📝 Notebook Demo

jupyter notebook orbisage_full_demo.ipynb

• Visualise the graph (Mermaid).
• Inline Gradio chat.
• External chat served on 0.0.0.0:7860 for Docker/K8s.

---

🌐 Web UI

python web/app.py

Open http://localhost:7860

---

🚢 Deployment

Docker

docker build -t orbisage:latest .
docker run -e OPENAI_API_KEY=$OPENAI_API_KEY -p 7860:7860 orbisage:latest

Kubernetes

apiVersion: apps/v1
kind: Deployment
metadata: { name: orbisage }
spec:
  replicas: 2
  selector: { matchLabels: { app: orbisage } }
  template:
    metadata: { labels: { app: orbisage } }
    spec:
      containers:
      - name: orbisage
        image: your-dockerhub/orbisage:latest
        env:
        - name: OPENAI_API_KEY
          valueFrom:
            secretKeyRef: { name: openai, key: api_key }
        ports: [ { containerPort: 7860 } ]
---
apiVersion: v1
kind: Service
metadata: { name: orbisage-svc }
spec:
  selector: { app: orbisage }
  type: LoadBalancer
  ports: [ { port: 80, targetPort: 7860 } ]

kubectl apply -f k8s/orbisage.yaml

---

🔍 Example Queries

Query	Agent	Response
“Hello!”	greet	“Hello! How can I help you today?”
“How do I get to the gym?”	Navigator	“Office → Elevator → Gym.”
“What’s in AI news today?”	News	“LLM-based agents are trending.”
“Tell me something about honey.”	Discovery	“Honey from ancient Egypt is edible!”
“Give me a programming joke.”	Joke	“Why do programmers prefer dark mode?”

---

🧩 Orbisage

Orbisage is basically your own friendly playground for trying out multi-agent ideas — you can start really small, test out one thing at a time, and see how they work together. It’s simple enough that you don’t feel stuck reading a giant manual, but it’s flexible enough to build into something real if you want to. Think of it as a laid-back lab assistant that never gets tired of your experiments — whether you want to navigate an imaginary building, share a science fact, or just tinker with different chat responses, you can do all of that here.”

---

🧩 Extending Orbisage

• Add new tools or plug in real APIs.
• Add nodes and update the graph.
• Use persistent memory like Redis or a vector store.
• Monitor with LangSmith.

---

🤝 Contributing & License

Pull requests and forks are welcome — add tests, lint with ruff.

Licensed under MIT License.

---

Happy routing! — The Orbisage Team