INIT(app): initial setup

- Initialize project structure - Add base application files
2025-12-23 22:31:45 +09:00
commit 346b0c79ef
16 changed files with 1006 additions and 0 deletions
--- a/.gitignore
+++ b/.gitignore
@@ -0,0 +1,58 @@
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+.Python
+venv/
+env/
+ENV/
+.venv
+pip-log.txt
+pip-delete-this-directory.txt
+.pytest_cache/
+.coverage
+htmlcov/
+dist/
+build/
+*.egg-info/
+
+# Environment variables
+.env
+.env.local
+.env.*.local
+*.env
+
+# Chainlit
+.chainlit/
+.files/
+
+# IDE
+.vscode/
+.idea/
+*.swp
+*.swo
+*~
+.DS_Store
+
+# Logs
+*.log
+logs/
+
+# Docker
+docker-compose.override.yml
+
+# Kubernetes secrets (local only)
+*-secret.yaml
+!deploy/vault/*.yaml
+
+# Node modules (if any frontend added)
+node_modules/
+package-lock.json
+yarn.lock
+
+# Temporary files
+tmp/
+temp/
+*.tmp
+
--- a/README.md
+++ b/README.md
@@ -0,0 +1,261 @@
+# MAS (Multi-Agent System)
+
+MAS is a unified UI and orchestration layer for multiple AI agents (similar to ChatGPT, Claude, Gemini), running on your own Kubernetes cluster.
+
+## 🎯 Architecture
+
+### Agents
+- **Claude Code (Orchestrator)**: overall coordinator & DevOps expert  
+- **Qwen Backend**: backend engineer (FastAPI, Node.js)  
+- **Qwen Frontend**: frontend engineer (Next.js, React)  
+- **Qwen SRE**: monitoring & reliability engineer  
+
+### Tech stack
+- **Backend**: LangGraph + LangChain + FastAPI  
+- **UI**: Chainlit (chat-style UI)  
+- **Database**: PostgreSQL (CNPG)  
+- **Cache**: Redis  
+- **LLMs**: Claude API + **Groq Llama 3.x** (OpenAI-compatible API)  
+- **Deploy**: Kubernetes + ArgoCD  
+
+---
+
+## 🚀 Local development
+
+### 1. Run with Docker Compose
+
+```bash
+cd deploy/docker
+
+# Copy or create .env and fill in your API keys
+# (ANTHROPIC_API_KEY, GROQ_API_KEY, etc.)
+
+# Start the full stack
+docker compose up -d
+
+# Tail logs
+docker compose logs -f mas
+```
+
+Open: `http://localhost:8000`
+
+### 2. Run backend directly (Python)
+
+```bash
+cd services/backend
+
+# Create venv
+python -m venv venv
+source venv/bin/activate  # Windows: venv\Scripts\activate
+
+# Install dependencies
+pip install -r requirements.txt
+
+# Environment variables
+cp .env.example .env
+# Edit .env and set your API keys
+
+# Run Chainlit app
+chainlit run chainlit_app.py
+```
+
+---
+
+## ☸️ Kubernetes deployment
+
+### 1. Create namespace and secrets
+
+```bash
+kubectl create namespace mas
+
+kubectl create secret generic mas-api-keys \
+  --from-literal=anthropic-api-key=YOUR_CLAUDE_KEY \
+  --from-literal=openai-api-key=YOUR_OPENAI_KEY \
+  --from-literal=google-api-key=YOUR_GEMINI_KEY \
+  -n mas
+```
+
+### 2. Deploy via ArgoCD
+
+```bash
+# Create ArgoCD Application
+kubectl apply -f deploy/argocd/mas.yaml
+
+# Sync and check status
+argocd app sync mas
+argocd app get mas
+```
+
+### 3. Deploy from your server (example)
+
+```bash
+# SSH into your k3s master
+ssh oracle-master
+
+# Apply ArgoCD Application
+sudo kubectl apply -f /path/to/deploy/argocd/mas.yaml
+
+# Check status
+sudo kubectl get pods -n mas
+sudo kubectl logs -f deployment/mas -n mas
+```
+
+Ingress example (if configured): `https://mas.mayne.vcn`
+
+---
+
+## 🎨 UI customization
+
+### Chainlit theme & behavior
+
+You can customize the UI via `services/backend/.chainlit`:
+
+```toml
+[UI]
+name = "MAS"
+show_readme_as_default = true
+default_collapse_content = true
+```
+
+### Agent prompts
+
+System prompts for each agent live in `services/backend/agents.py`.  
+You can tune:
+- how the **Orchestrator** routes tasks  
+- coding style of backend/frontend agents  
+- SRE troubleshooting behavior  
+
+---
+
+## 📊 Observability
+
+### Prometheus ServiceMonitor (example)
+
+```yaml
+apiVersion: monitoring.coreos.com/v1
+kind: ServiceMonitor
+metadata:
+  name: mas
+  namespace: mas
+spec:
+  selector:
+    matchLabels:
+      app: mas
+  endpoints:
+  - port: http
+    path: /metrics
+```
+
+### Grafana dashboards
+
+Recommended panels:
+- LangGraph workflow metrics  
+- Per-agent latency & error rate  
+- Token usage and cost estimates  
+- Backend API latency & 5xx rate  
+
+---
+
+## 🔧 Advanced features
+
+### 1. MCP (Model Context Protocol) with Claude
+
+Using Claude Code as Orchestrator, MAS can access:
+- Filesystem (read/write project files)  
+- Git (status, commit, push, PR)  
+- SSH (run remote commands on your servers)  
+- PostgreSQL (schema inspection, migrations, queries)  
+- Kubernetes (kubectl via MCP tool)  
+
+This allows fully automated workflows like:
+- “Create a new service, add deployment manifests, and deploy to k3s.”  
+- “Debug failing pods and propose a fix, then open a PR.”  
+
+### 2. Multi-agent collaboration (LangGraph)
+
+Typical workflow:
+
+```text
+User request
+  ↓
+Claude Orchestrator
+  ↓ decides which agent(s) to call
+Backend Dev → Frontend Dev → SRE
+  ↓
+Claude Orchestrator (review & summary)
+  ↓
+Final answer to user
+```
+
+Examples:
+- Full‑stack feature (API + UI + monitoring)  
+- Infra rollout (Harbor, Tekton, CNPG, MetalLB) with validation  
+
+---
+
+## 📝 Usage examples
+
+### Backend API request
+
+```text
+User: "Create a signup API with FastAPI.
+       Use PostgreSQL and JWT tokens."
+
+🎼 Orchestrator:
+  → routes to Qwen Backend
+
+⚙️ Qwen Backend:
+  → generates FastAPI router, Pydantic models, DB schema, JWT logic
+
+🎼 Orchestrator:
+  → reviews, suggests improvements, and outputs final code snippet & file layout
+```
+
+### Frontend component request
+
+```text
+User: "Build a responsive dashboard chart component using Recharts."
+
+🎼 Orchestrator:
+  → routes to Qwen Frontend
+
+🎨 Qwen Frontend:
+  → generates a Next.js/React component with TypeScript and responsive styles
+
+🎼 Orchestrator:
+  → explains how to integrate it into your existing app
+```
+
+### Infra / SRE request
+
+```text
+User: "Prometheus is firing high memory alerts for the PostgreSQL pod.
+       Help me stabilize it."
+
+🎼 Orchestrator:
+  → routes to Qwen SRE
+
+📊 Qwen SRE:
+  → analyzes metrics & logs (conceptually),
+    proposes tuning (Postgres config, indexes, pooler),
+    and suggests alert threshold adjustments.
+```
+
+---
+
+## 🤝 Contributing
+
+Contributions are welcome:
+- New agents (e.g., data engineer, security engineer)  
+- New tools (Harbor, Tekton, CNPG, MetalLB integrations)  
+- Better prompts and workflows  
+- Docs and examples  
+
+Feel free to open issues or PRs in your Git repository.
+
+---
+
+## 📄 License
+
+MIT
+
--- a/deploy/argocd/mas.yaml
+++ b/deploy/argocd/mas.yaml
@@ -0,0 +1,21 @@
+apiVersion: argoproj.io/v1alpha1
+kind: Application
+metadata:
+  name: mas
+  namespace: argocd
+spec:
+  project: default
+  source:
+    repoURL: https://gitea0213.kro.kr/bluemayne/mas.git
+    targetRevision: HEAD
+    path: deploy/k8s
+  destination:
+    server: https://kubernetes.default.svc
+    namespace: mas
+  syncPolicy:
+    automated:
+      prune: true
+      selfHeal: true
+    syncOptions:
+    - CreateNamespace=true
+
--- a/deploy/docker/Dockerfile
+++ b/deploy/docker/Dockerfile
@@ -0,0 +1,24 @@
+FROM python:3.11-slim
+
+WORKDIR /app
+
+# 시스템 의존성 설치
+RUN apt-get update && apt-get install -y \
+    build-essential \
+    curl \
+    git \
+    && rm -rf /var/lib/apt/lists/*
+
+# Python 의존성 설치
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# 애플리케이션 코드 복사
+COPY . .
+
+# Chainlit 포트
+EXPOSE 8000
+
+# Chainlit 실행
+CMD ["chainlit", "run", "chainlit_app.py", "--host", "0.0.0.0", "--port", "8000"]
+
--- a/deploy/docker/docker-compose.yml
+++ b/deploy/docker/docker-compose.yml
@@ -0,0 +1,73 @@
+version: '3.8'
+
+services:
+  mas:
+    build: ../../services/backend
+    container_name: mas
+    ports:
+      - "8000:8000"
+    environment:
+      - ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY}
+      # Groq API (OpenAI-compatible)
+      - GROQ_API_KEY=${GROQ_API_KEY}
+      - GROQ_API_BASE=${GROQ_API_BASE:-https://api.groq.com/openai/v1}
+      # (optional) keep other providers
+      - OPENAI_API_KEY=${OPENAI_API_KEY}
+      - GOOGLE_API_KEY=${GOOGLE_API_KEY}
+      - DATABASE_URL=postgresql+asyncpg://mas:mas@postgres:5432/mas
+      - REDIS_URL=redis://redis:6379/0
+    depends_on:
+      - redis
+      - postgres
+      - ollama
+    volumes:
+      - ../../services/backend:/app
+    networks:
+      - mas-network
+
+  # Ollama (로컬 Qwen 모델)
+  ollama:
+    image: ollama/ollama:latest
+    container_name: mas-ollama
+    ports:
+      - "11434:11434"
+    volumes:
+      - ollama-data:/root/.ollama
+    networks:
+      - mas-network
+
+  # PostgreSQL
+  postgres:
+    image: postgres:16-alpine
+    container_name: mas-postgres
+    environment:
+      POSTGRES_DB: mas
+      POSTGRES_USER: mas
+      POSTGRES_PASSWORD: mas
+    ports:
+      - "5432:5432"
+    volumes:
+      - postgres-data:/var/lib/postgresql/data
+    networks:
+      - mas-network
+
+  # Redis
+  redis:
+    image: redis:7-alpine
+    container_name: mas-redis
+    ports:
+      - "6379:6379"
+    volumes:
+      - redis-data:/data
+    networks:
+      - mas-network
+
+volumes:
+  ollama-data:
+  postgres-data:
+  redis-data:
+
+networks:
+  mas-network:
+    driver: bridge
+
--- a/deploy/k8s/deployment.yaml
+++ b/deploy/k8s/deployment.yaml
@@ -0,0 +1,73 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: mas
+  namespace: mas
+  labels:
+    app: mas
+spec:
+  replicas: 2
+  selector:
+    matchLabels:
+      app: mas
+  template:
+    metadata:
+      labels:
+        app: mas
+    spec:
+      containers:
+      - name: mas
+        image: harbor.mayne.vcn/mas/platform:latest
+        ports:
+        - containerPort: 8000
+          name: http
+        env:
+        - name: ANTHROPIC_API_KEY
+          valueFrom:
+            secretKeyRef:
+              name: mas-api-keys
+              key: anthropic-api-key
+        - name: OPENAI_API_KEY
+          valueFrom:
+            secretKeyRef:
+              name: mas-api-keys
+              key: openai-api-key
+        - name: GOOGLE_API_KEY
+          valueFrom:
+            secretKeyRef:
+              name: mas-api-keys
+              key: google-api-key
+        - name: GROQ_API_KEY
+          valueFrom:
+            secretKeyRef:
+              name: mas-api-keys
+              key: groq-api-key
+        - name: GROQ_API_BASE
+          value: "https://api.groq.com/openai/v1"
+        - name: DATABASE_URL
+          valueFrom:
+            secretKeyRef:
+              name: mas-postgres
+              key: database-url
+        - name: REDIS_URL
+          value: "redis://redis:6379/0"
+        resources:
+          requests:
+            memory: "512Mi"
+            cpu: "500m"
+          limits:
+            memory: "2Gi"
+            cpu: "2000m"
+        livenessProbe:
+          httpGet:
+            path: /health
+            port: 8000
+          initialDelaySeconds: 30
+          periodSeconds: 10
+        readinessProbe:
+          httpGet:
+            path: /health
+            port: 8000
+          initialDelaySeconds: 10
+          periodSeconds: 5
+
--- a/deploy/k8s/ingress.yaml
+++ b/deploy/k8s/ingress.yaml
@@ -0,0 +1,29 @@
+apiVersion: networking.k8s.io/v1
+kind: Ingress
+metadata:
+  name: mas
+  namespace: mas
+  annotations:
+    cert-manager.io/cluster-issuer: letsencrypt-prod
+    nginx.ingress.kubernetes.io/ssl-redirect: "true"
+    nginx.ingress.kubernetes.io/websocket-services: "mas"
+    nginx.ingress.kubernetes.io/proxy-read-timeout: "3600"
+    nginx.ingress.kubernetes.io/proxy-send-timeout: "3600"
+spec:
+  ingressClassName: nginx
+  tls:
+  - hosts:
+    - mas.mayne.vcn
+    secretName: mas-tls
+  rules:
+  - host: mas.mayne.vcn
+    http:
+      paths:
+      - path: /
+        pathType: Prefix
+        backend:
+          service:
+            name: mas
+            port:
+              number: 8000
+
--- a/deploy/k8s/kustomization.yaml
+++ b/deploy/k8s/kustomization.yaml
@@ -0,0 +1,13 @@
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+
+namespace: mas
+
+resources:
+  - namespace.yaml
+  - ../vault/mas-api-keys.yaml
+  - ../vault/mas-postgres.yaml
+  - deployment.yaml
+  - service.yaml
+  - ingress.yaml
+
--- a/deploy/k8s/namespace.yaml
+++ b/deploy/k8s/namespace.yaml
@@ -0,0 +1,7 @@
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: mas
+  labels:
+    name: mas
+
--- a/deploy/k8s/service.yaml
+++ b/deploy/k8s/service.yaml
@@ -0,0 +1,17 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: mas
+  namespace: mas
+  labels:
+    app: mas
+spec:
+  type: ClusterIP
+  ports:
+  - port: 8000
+    targetPort: 8000
+    protocol: TCP
+    name: http
+  selector:
+    app: mas
+
--- a/deploy/vault/mas-api-keys.yaml
+++ b/deploy/vault/mas-api-keys.yaml
@@ -0,0 +1,31 @@
+apiVersion: external-secrets.io/v1beta1
+kind: ExternalSecret
+metadata:
+  name: mas-api-keys
+  namespace: mas
+spec:
+  refreshInterval: 1h
+  secretStoreRef:
+    kind: ClusterSecretStore
+    name: vault-backend
+  target:
+    name: mas-api-keys
+    creationPolicy: Owner
+  data:
+    - secretKey: anthropic-api-key
+      remoteRef:
+        key: mas/api-keys
+        property: ANTHROPIC_API_KEY
+    - secretKey: groq-api-key
+      remoteRef:
+        key: mas/api-keys
+        property: GROQ_API_KEY
+    - secretKey: openai-api-key
+      remoteRef:
+        key: mas/api-keys
+        property: OPENAI_API_KEY
+    - secretKey: google-api-key
+      remoteRef:
+        key: mas/api-keys
+        property: GOOGLE_API_KEY
+
--- a/deploy/vault/mas-postgres.yaml
+++ b/deploy/vault/mas-postgres.yaml
@@ -0,0 +1,27 @@
+apiVersion: external-secrets.io/v1beta1
+kind: ExternalSecret
+metadata:
+  name: mas-postgres
+  namespace: mas
+spec:
+  refreshInterval: 1h
+  secretStoreRef:
+    kind: ClusterSecretStore
+    name: vault-backend
+  target:
+    name: mas-postgres
+    creationPolicy: Owner
+  data:
+    - secretKey: database-url
+      remoteRef:
+        key: mas/postgres
+        property: DATABASE_URL
+    - secretKey: username
+      remoteRef:
+        key: mas/postgres
+        property: USERNAME
+    - secretKey: password
+      remoteRef:
+        key: mas/postgres
+        property: PASSWORD
+
--- a/services/backend/.chainlit
+++ b/services/backend/.chainlit
@@ -0,0 +1,17 @@
+[project]
+enable_telemetry = false
+user_env = []
+session_timeout = 3600
+cache = false
+
+[features]
+prompt_playground = true
+unsafe_allow_html = true
+latex = true
+
+[UI]
+name = "MAS Platform"
+default_collapse_content = true
+default_expand_messages = false
+hide_cot = false
+
--- a/services/backend/agents.py
+++ b/services/backend/agents.py
@@ -0,0 +1,240 @@
+"""
+MAS (Multi-Agent System) 에이전트 정의
+"""
+from typing import Annotated, Literal, TypedDict
+from langchain_anthropic import ChatAnthropic
+from langchain_openai import ChatOpenAI
+from langgraph.graph import StateGraph, END
+from langgraph.prebuilt import ToolNode
+from langchain_core.messages import HumanMessage, SystemMessage
+import os
+
+
+class AgentState(TypedDict):
+    """에이전트 간 공유되는 상태"""
+    messages: list
+    current_agent: str
+    task_type: str
+    result: dict
+
+
+# ===== 1. Claude Code - Orchestrator =====
+claude_orchestrator = ChatAnthropic(
+    model="claude-3-5-sonnet-20241022",
+    api_key=os.getenv("ANTHROPIC_API_KEY"),
+    temperature=0
+)
+
+ORCHESTRATOR_PROMPT = """당신은 MAS의 총괄 조율자이자 DevOps 전문가입니다.
+
+**역할**:
+- 사용자 요청을 분석하여 적절한 에이전트에게 작업 할당
+- Kubernetes, ArgoCD, Helm, Kustomize 관리
+- CI/CD 파이프라인 구성
+- 최종 코드 리뷰 및 승인
+
+**사용 가능한 에이전트**:
+1. backend_developer: FastAPI, Node.js 백엔드 개발
+2. frontend_developer: Next.js, React 프론트엔드 개발
+3. sre_specialist: 모니터링, 성능 최적화, 보안
+
+요청을 분석하고 어떤 에이전트가 처리해야 할지 결정하세요.
+"""
+
+
+# ===== 2. Groq #1 - Backend Developer =====
+# Groq OpenAI-compatible endpoint
+GROQ_API_BASE = os.getenv("GROQ_API_BASE", "https://api.groq.com/openai/v1")
+GROQ_API_KEY = os.getenv("GROQ_API_KEY", "")
+
+groq_backend = ChatOpenAI(
+    model=os.getenv("GROQ_BACKEND_MODEL", "llama-3.3-70b-specdec"),
+    base_url=GROQ_API_BASE,
+    api_key=GROQ_API_KEY,
+    temperature=0.7,
+)
+
+BACKEND_PROMPT = """당신은 백엔드 개발 전문가입니다.
+
+**역할**:
+- FastAPI, Node.js 백엔드 개발
+- REST API 설계 및 구현
+- 데이터베이스 쿼리 최적화
+- 비즈니스 로직 구현
+
+요청된 백엔드 작업을 수행하고 코드를 생성하세요.
+"""
+
+
+# ===== 3. Groq #2 - Frontend Developer =====
+groq_frontend = ChatOpenAI(
+    model=os.getenv("GROQ_FRONTEND_MODEL", "llama-3.1-8b-instant"),
+    base_url=GROQ_API_BASE,
+    api_key=GROQ_API_KEY,
+    temperature=0.7,
+)
+
+FRONTEND_PROMPT = """당신은 프론트엔드 개발 전문가입니다.
+
+**역할**:
+- Next.js, React 컴포넌트 개발
+- UI/UX 구현
+- 상태 관리
+- 반응형 디자인
+
+요청된 프론트엔드 작업을 수행하고 코드를 생성하세요.
+"""
+
+
+# ===== 4. Groq #3 - SRE Specialist =====
+groq_sre = ChatOpenAI(
+    model=os.getenv("GROQ_SRE_MODEL", "llama-3.1-8b-instant"),
+    base_url=GROQ_API_BASE,
+    api_key=GROQ_API_KEY,
+    temperature=0.3,
+)
+
+SRE_PROMPT = """당신은 SRE(Site Reliability Engineer) 전문가입니다.
+
+**역할**:
+- 시스템 모니터링 (Prometheus, Grafana, Loki)
+- 로그 분석 및 알람 설정
+- 성능 튜닝
+- 보안 취약점 점검
+
+요청된 SRE 작업을 수행하고 솔루션을 제시하세요.
+"""
+
+
+def orchestrator_node(state: AgentState) -> AgentState:
+    """Claude Code - 작업 분석 및 할당"""
+    messages = state["messages"]
+    
+    response = claude_orchestrator.invoke([
+        SystemMessage(content=ORCHESTRATOR_PROMPT),
+        HumanMessage(content=messages[-1]["content"])
+    ])
+    
+    # 작업 타입 결정
+    content = response.content.lower()
+    if "backend" in content or "api" in content or "fastapi" in content:
+        next_agent = "backend_developer"
+    elif "frontend" in content or "ui" in content or "react" in content:
+        next_agent = "frontend_developer"
+    elif "monitoring" in content or "performance" in content or "sre" in content:
+        next_agent = "sre_specialist"
+    else:
+        next_agent = "orchestrator"  # 자신이 직접 처리
+    
+    state["messages"].append({
+        "role": "orchestrator",
+        "content": response.content
+    })
+    state["current_agent"] = next_agent
+    
+    return state
+
+
+def backend_node(state: AgentState) -> AgentState:
+    """Groq #1 - 백엔드 개발"""
+    messages = state["messages"]
+    
+    response = groq_backend.invoke([
+        SystemMessage(content=BACKEND_PROMPT),
+        HumanMessage(content=messages[-1]["content"])
+    ])
+    
+    state["messages"].append({
+        "role": "backend_developer",
+        "content": response.content
+    })
+    state["current_agent"] = "orchestrator"  # 결과를 오케스트레이터에게 반환
+    
+    return state
+
+
+def frontend_node(state: AgentState) -> AgentState:
+    """Groq #2 - 프론트엔드 개발"""
+    messages = state["messages"]
+    
+    response = groq_frontend.invoke([
+        SystemMessage(content=FRONTEND_PROMPT),
+        HumanMessage(content=messages[-1]["content"])
+    ])
+    
+    state["messages"].append({
+        "role": "frontend_developer",
+        "content": response.content
+    })
+    state["current_agent"] = "orchestrator"
+    
+    return state
+
+
+def sre_node(state: AgentState) -> AgentState:
+    """Groq #3 - SRE 작업"""
+    messages = state["messages"]
+    
+    response = groq_sre.invoke([
+        SystemMessage(content=SRE_PROMPT),
+        HumanMessage(content=messages[-1]["content"])
+    ])
+    
+    state["messages"].append({
+        "role": "sre_specialist",
+        "content": response.content
+    })
+    state["current_agent"] = "orchestrator"
+    
+    return state
+
+
+def router(state: AgentState) -> Literal["backend_developer", "frontend_developer", "sre_specialist", "end"]:
+    """다음 에이전트 라우팅"""
+    current = state.get("current_agent", "orchestrator")
+    
+    if current == "backend_developer":
+        return "backend_developer"
+    elif current == "frontend_developer":
+        return "frontend_developer"
+    elif current == "sre_specialist":
+        return "sre_specialist"
+    else:
+        return "end"
+
+
+# ===== LangGraph 워크플로우 구성 =====
+def create_mas_graph():
+    """MAS 워크플로우 그래프 생성"""
+    workflow = StateGraph(AgentState)
+    
+    # 노드 추가
+    workflow.add_node("orchestrator", orchestrator_node)
+    workflow.add_node("backend_developer", backend_node)
+    workflow.add_node("frontend_developer", frontend_node)
+    workflow.add_node("sre_specialist", sre_node)
+    
+    # 엣지 정의
+    workflow.set_entry_point("orchestrator")
+    workflow.add_conditional_edges(
+        "orchestrator",
+        router,
+        {
+            "backend_developer": "backend_developer",
+            "frontend_developer": "frontend_developer",
+            "sre_specialist": "sre_specialist",
+            "end": END
+        }
+    )
+    
+    # 각 에이전트는 작업 후 orchestrator로 복귀
+    workflow.add_edge("backend_developer", "orchestrator")
+    workflow.add_edge("frontend_developer", "orchestrator")
+    workflow.add_edge("sre_specialist", "orchestrator")
+    
+    return workflow.compile()
+
+
+# 그래프 인스턴스 생성
+mas_graph = create_mas_graph()
+
--- a/services/backend/chainlit_app.py
+++ b/services/backend/chainlit_app.py
@@ -0,0 +1,85 @@
+"""
+Chainlit UI for MAS Platform
+"""
+import chainlit as cl
+from agents import mas_graph, AgentState
+import os
+from dotenv import load_dotenv
+
+load_dotenv()
+
+
+@cl.on_chat_start
+async def start():
+    """채팅 시작 시"""
+    await cl.Message(
+        content="🤖 **Multi-Agent System**에 오신 것을 환영합니다!\n\n"
+                "저는 다음 전문가 팀과 함께 작업합니다:\n\n"
+                "- 🎼 **Claude Code**: 총괄 조율자 & DevOps 전문가\n"
+                "- ⚙️ **Qwen Backend**: 백엔드 개발자\n"
+                "- 🎨 **Qwen Frontend**: 프론트엔드 개발자\n"
+                "- 📊 **Qwen SRE**: 모니터링 & 성능 전문가\n\n"
+                "무엇을 도와드릴까요?"
+    ).send()
+
+
+@cl.on_message
+async def main(message: cl.Message):
+    """메시지 수신 시"""
+    
+    # 초기 상태
+    initial_state: AgentState = {
+        "messages": [{"role": "user", "content": message.content}],
+        "current_agent": "orchestrator",
+        "task_type": "",
+        "result": {}
+    }
+    
+    # 응답 메시지 생성
+    response_msg = cl.Message(content="")
+    await response_msg.send()
+    
+    # MAS 그래프 실행
+    async for event in mas_graph.astream(initial_state):
+        for node_name, state in event.items():
+            if node_name != "__end__":
+                last_message = state["messages"][-1]
+                agent_name = last_message["role"]
+                agent_content = last_message["content"]
+                
+                # 에이전트별 아이콘
+                agent_icons = {
+                    "orchestrator": "🎼",
+                    "backend_developer": "⚙️",
+                    "frontend_developer": "🎨",
+                    "sre_specialist": "📊"
+                }
+                
+                icon = agent_icons.get(agent_name, "🤖")
+                
+                # 스트리밍 업데이트
+                response_msg.content += f"\n\n{icon} **{agent_name}**:\n{agent_content}"
+                await response_msg.update()
+    
+    # 최종 업데이트
+    await response_msg.update()
+
+
+@cl.on_settings_update
+async def setup_agent(settings):
+    """설정 업데이트"""
+    print(f"Settings updated: {settings}")
+
+
+# 사이드바 설정
+@cl.author_rename
+def rename(orig_author: str):
+    """에이전트 이름 매핑"""
+    rename_dict = {
+        "orchestrator": "Claude Code (Orchestrator)",
+        "backend_developer": "Qwen Backend Dev",
+        "frontend_developer": "Qwen Frontend Dev",
+        "sre_specialist": "Qwen SRE"
+    }
+    return rename_dict.get(orig_author, orig_author)
+
--- a/services/backend/requirements.txt
+++ b/services/backend/requirements.txt
@@ -0,0 +1,30 @@
+# LangGraph & LangChain
+langgraph==0.2.53
+langchain==0.3.13
+langchain-anthropic==0.3.0
+langchain-openai==0.2.14
+langchain-google-genai==2.0.8
+
+# Chainlit (UI)
+chainlit==1.3.1
+
+# API Framework
+fastapi==0.115.6
+uvicorn[standard]==0.34.0
+pydantic==2.10.5
+pydantic-settings==2.7.0
+
+# Database
+sqlalchemy==2.0.36
+asyncpg==0.30.0
+psycopg2-binary==2.9.10
+
+# MCP (Model Context Protocol)
+mcp==1.1.2
+
+# Utilities
+python-dotenv==1.0.1
+redis==5.2.1
+aioredis==2.0.1
+httpx==0.28.1
+