How corentin-tin/langgraph-agent-templates works
Project type and purpose
• This is a full-stack microservices template repo for two AI-agent services (React Agent, RAG Agent) plus shared libraries, optimized for Docker Compose and Kubernetes deployment.
• The agents are FastAPI servers implementing LangGraph graphs (ReAct and RAG patterns) and rely on external LLMs, tool APIs, and vector stores.
Main components
a. Agents (Microservices)
• React Agent
• FastAPI server (server.py)
• Core graph (graph.py), state & utils (prompts, nodes, tools)
• External tool integrations: web search (Tavily), LLM providers
• RAG Agent
• FastAPI server (server.py)
• Multi-graph modules (indexing, retrieval, planning)
• Vector store connectors (Elasticsearch, Pinecone, MongoDB, FAISS), LLM providers
b. Shared Package
• packages/shared
• Logging, settings, types used by both agents
c. External Dependencies
• LLM Providers (OpenAI, Anthropic, Fireworks)
• Tool APIs (Tavily)
• Vector Stores (Elasticsearch, Pinecone, MongoDB, FAISS)
• Optional: Redis (caching)
d. Deployment Infrastructure
• Docker Compose
• react-agent, rag-agent, optional elasticsearch, redis services
• Kubernetes
• Namespace, Deployments, Services, Ingress for each agent
• CI/CD (GitHub Actions workflows)
e. Client / User Interfaces
• LangGraph Studio (web UI connecting to agents)
• Direct API clients (curl / Postman)
Relationships & Data Flows
• Clients → Ingress → Service (React/RAG Agent)
• Agents → Shared library (internal call)
• Agents → External LLMs & Tool APIs (HTTP/SDK calls)
• RAG Agent → Vector Stores (REST or SDK)
• Optional: Agents ↔ Redis cache
• CI/CD → Docker build & push → Kubernetes registry → Deploy
• Docker Compose orchestrates local multi-container setup
Architectural patterns & principles
• Microservices: each agent runs as an independent container/service
• Modular code: shared utilities extracted to a separate package
• Graph pattern: core logic expressed as LangGraph graphs (nodes, state machines)
• Infrastructure as code: docker-compose.yml and k8s manifests
• CI/CD pipeline automates testing, linting, deployment
Final checklist
• Grouping boundaries for Docker Compose and Kubernetes environments