🚀 LegendStack Agentic FastAPI Template

Ship production-ready AI agents in hours, not months.

📚 Documentation · 🎓 Tutorials · 🔧 Guides · 💬 Discord

---

⚡ Time to First Agent

git clone https://github.com/LegendStack/agentic-fastapi-template
cd agentic-fastapi-template && docker compose up

That's it. Your agent is running at http://localhost:8000/docs with:

• 🤖 Demo Agent showcasing all features (no API keys needed)
• 📚 RAG pipeline with vector search
• 🛡️ Safety guardrails (PII masking, moderation)
• 💬 Real-time streaming (SSE/WebSocket)

Tip

New to LegendStack? Start with the Interactive Tutorial — it walks you through every feature in 90 minutes.

---

🎯 Why LegendStack?

Building AI Agents From Scratch	With LegendStack
Set up FastAPI, auth, database, caching...	✅ Batteries included
Build RAG pipeline from scratch	✅ Production-ready RAG with reranking
Figure out LangGraph patterns	✅ 11 modular node examples
Add safety guardrails	✅ PII masking, moderation built-in
Implement caching, rate limiting	✅ Semantic cache, tiered limits
Debug memory, threading issues	✅ Entity-aware cross-thread memory
Build monitoring, cost tracking	✅ OpenTelemetry + cost tracking
Weeks/Months	Hours

---

🏗️ What's Included

🔧 Core Platform ⚡ FastAPI + SQLAlchemy 2.0 (fully async) 🔐 JWT Auth + Microsoft Entra ID (Azure AD) 👮 Tiered Rate Limiting (Free/Standard/Premium) 🧊 Redis Caching + Background Jobs (ARQ) 🐳 Docker Compose (dev/staging/production)	🤖 Agentic AI Framework 🧠 LangGraph Orchestration with state management 📚 RAG with Redis, Azure AI Search, or pgvector 🔗 Graph-RAG with Neo4j knowledge graphs 🛡️ Safety Guardrails (PII, moderation) ⚡ Semantic Caching (reduce LLM costs 70%+)
🔌 Enterprise Integrations 📝 Jira, Confluence, SharePoint, OneDrive 🔍 Azure OpenAI, Cohere, Cross-Encoders 📊 OpenTelemetry (Datadog, Honeycomb) 🔐 Azure Key Vault, Managed Identity 🛡️ HashiCorp Vault (AppRole Auth)	🎓 Developer Experience 🖥️ LegendStack Studio (Streamlit dashboard) 🛠️ CLI for scaffolding agents/connectors 📓 Jupyter Tutorial (8 chapters) 📋 Production Checklist (28 items)

---

📚 Learning Resources

Resource	Format	Time	Best For
Quick Start	Docs	5 min	Get running fast
Jupyter Tutorial	Notebook	90 min	Hands-on learning
Streamlit Walkthrough	Web App	30 min	Visual learners
Integration Guide	Docs	15 min	Connect real services
Customization Cookbook	Docs	—	Add your logic
Production Checklist	Checklist	—	Pre-deploy review

---

🚀 Quick Start Examples

🤖 Chat with the Demo Agent

from app.agents.demo import LegendDemoAgent

agent = LegendDemoAgent()  # Uses mocks (no API keys needed)
result = await agent.chat("What is LegendStack?")

print(result["response"])
print(result["features_used"])  # Shows which features were invoked

☁️ Connect Azure OpenAI

from app.core.integration_config import configure_integrations, get_azure_openai_client

configure_integrations()  # Auto-configures from .env
client = get_azure_openai_client()
response = await client.chat([{"role": "user", "content": "Hello!"}], deployment="gpt-4o")

🔍 RAG with Reranking

from app.agents import RerankingService, CrossEncoderReranker

docs = await vector_store.search("How to deploy?", k=20)
reranker = RerankingService(CrossEncoderReranker())
top_docs = await reranker.rerank("How to deploy?", docs, top_k=5)

🛡️ Resilient LLM Calls

from app.agents import ResilientClient

client = ResilientClient(name="openai")  # Retry + circuit breaker
result = await client.execute(call_llm, prompt)  # Auto-retries on failure

🛠️ CLI: Generate New Agent

uv run python src/app/cli/main.py create-agent CustomerSupport
uv run python src/app/cli/main.py create-connector Slack

🖥️ Launch Studio Dashboard

streamlit run studio/main.py

---

⏱️ Time to Production

Milestone	Time
Clone & run with mocks	5 min
Connect real Azure OpenAI	30 min
Customize for your use case	2-4 hours
Production-ready	1-2 days

---

🏃 Setup Options

Interactive setup:

./setup.py  # Choose: local / staging / production

Environment	Config	Best For
local	Uvicorn + auto-reload	Development
staging	Gunicorn + workers	Load testing
production	NGINX + Gunicorn	Deployment

Then:

docker compose up

📖 Full setup guide: Getting Started

---

🗺️ Project Structure

src/app/
├── agents/           # 🤖 Agentic AI framework
│   ├── demo/         #    Demo agent with all features
│   ├── nodes/        #    Reusable LangGraph nodes
│   ├── connectors/   #    Enterprise integrations
│   └── guardrails/   #    Safety & moderation
├── api/v1/           # 🔌 REST endpoints
├── core/             # ⚙️ Config, DB, auth
└── cli/              # 🛠️ Developer tools

docs/
├── getting-started/  # 📚 Quick start
├── tutorials/        # 🎓 Interactive learning
├── guides/           # 🔧 Integration & cookbook
└── user-guide/       # 📖 Detailed reference

notebooks/
└── tutorial.ipynb    # 📓 8-chapter Jupyter tutorial

studio/
└── tutorial_app.py   # 🖥️ Streamlit walkthrough

---

🧪 Testing

# Run all tests (161 tests)
uv run pytest tests/ -v

# Run specific test file
uv run pytest tests/test_demo_agent.py -v

---

🔐 HashiCorp Vault Configuration

To enable Vault integration (AppRole Auth):

VAULT_ENABLED=true
VAULT_URL=http://localhost:8200
VAULT_ROLE_ID=<your-role-id>
VAULT_SECRET_ID=<your-secret-id>
VAULT_SECRET_PATH=secret/data/my-app

---

🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

• 🐛 Report bugs
• 💡 Request features
• 📖 Improve docs

---

📜 License

MIT — use it for anything.

---

_{Built with ❤️ by the LegendStack team}