Infrastructure Documentation

This documentation covers the infrastructure components of the TFG-Chatbot platform, including Docker orchestration, monitoring, logging, alerting, and CI/CD pipelines.

Overview

The TFG-Chatbot uses a containerized microservices architecture with comprehensive observability:

Category Components
Orchestration Docker Compose
Databases MongoDB, Qdrant
Monitoring Prometheus, Grafana
Logging Loki, Promtail
Alerting Alertmanager
LLM Observability Phoenix (Arize AI)
CI/CD GitHub Actions

Architecture Diagram 

graph TB
    subgraph "Application Layer"
        FE[Frontend :3000]
        BE[Backend :8000]
        CB[Chatbot :8080]
        RAG[RAG Service :8081]
    end
    
    subgraph "AI Layer"
        VLLM[vLLM :8001]
        Ollama[Ollama :11435]
    end
    
    subgraph "Data Layer"
        MongoDB[(MongoDB :27017)]
        Qdrant[(Qdrant :6333)]
        ME[Mongo Express :8082]
    end
    
    subgraph "Observability Stack"
        Phoenix[Phoenix :6006]
        Prometheus[Prometheus :9093]
        Grafana[Grafana :3001]
        Loki[Loki :3100]
        Promtail[Promtail]
        Alert[Alertmanager :9094]
    end
    
    FE --> BE
    BE --> CB
    CB --> RAG
    CB --> VLLM
    RAG --> Qdrant
    RAG --> Ollama
    BE --> MongoDB
    CB --> MongoDB
    
    CB --> Phoenix
    BE --> Prometheus
    CB --> Prometheus
    RAG --> Prometheus
    Promtail --> Loki
    Loki --> Grafana
    Prometheus --> Grafana
    Prometheus --> Alert

Services Overview

Application Services

Service Container Port Description
Frontend tfg-frontend 3000 React SPA with nginx
Backend tfg-gateway 8000 FastAPI gateway
Chatbot tfg-chatbot 8080 LangGraph agent
RAG Service rag_service 8081 Document search

AI Services

Service Container Port Description
vLLM vllm-service 8001 GPU-accelerated LLM inference
Ollama ollama-service 11435 Embedding generation

Data Services

Service Container Port Description
MongoDB mongo 27017 Document database
Mongo Express mongo-express 8082 MongoDB web UI
Qdrant qdrant-service 6333/6334 Vector database

Observability Services

Service Container Port Description
Prometheus prometheus 9093 Metrics collection
Grafana grafana 3001 Visualization dashboards
Loki loki 3100 Log aggregation
Promtail promtail 9080 Log collection
Alertmanager alertmanager 9094 Alert routing
Phoenix phoenix 6006 LLM observability

Quick Start

Start All Services 

# Start everything
docker compose up -d

# View logs
docker compose logs -f

# Check status
docker compose ps

Start Minimal Stack 

# Start only core services (no vLLM, no monitoring)
docker compose up -d mongo qdrant ollama rag_service chatbot backend frontend

Start Monitoring Stack 

# Start observability services
docker compose up -d prometheus grafana loki promtail alertmanager phoenix

Environment Configuration

Create a .env file from the example:

cp .env.example .env

Key variables:

Variable Description Default
LLM_PROVIDER LLM backend (vllm/gemini/mistral) mistral
SECRET_KEY JWT signing key -
MONGO_ROOT_PASSWORD MongoDB password example
GRAFANA_ADMIN_PASSWORD Grafana admin password admin
PHOENIX_ENABLED Enable LLM tracing true

Resource Requirements

Minimum (Development)

  • CPU: 4 cores
  • RAM: 8 GB
  • Disk: 20 GB
  • CPU: 8+ cores
  • RAM: 32 GB
  • GPU: NVIDIA with 8+ GB VRAM
  • Disk: 100 GB SSD

Volumes

Persistent data is stored in Docker volumes:

Volume Service Purpose
mongodb_data MongoDB Database files
qdrant_storage Qdrant Vector indices
ollama_models Ollama Downloaded models
rag_documents RAG Service Uploaded documents
vllm_cache vLLM Model cache
phoenix_data Phoenix Trace data
prometheus_data Prometheus Metrics
loki_data Loki Logs
grafana_data Grafana Dashboards

Documentation Index

This documentation is part of the TFG Chatbot project - a dual degree thesis (Computer Science + Mathematics) at the University of Granada.