Skip to content

README.md

Latest commit

 

History

History
175 lines (148 loc) · 6.04 KB

README.md

File metadata and controls

175 lines (148 loc) · 6.04 KB
Raw

EstateWise Docker Stack

This directory contains production-grade Docker assets for the EstateWise monorepo.

flowchart LR
  User((User)) -->|HTTP| Nginx[Nginx :80]
  Nginx --> Frontend[Next.js :3000]
  Nginx --> Backend[API :3001]
  Backend --> Mongo[(MongoDB)]
  Backend --> Pinecone[(Pinecone)]
  Backend --> Neo4j[(Neo4j)]
  Agentic[Agentic AI :4318] --> MCP[MCP Server]
  MCP --> Backend
  MCP --> Frontend
Loading

What's Included

  • compose.prod.yml: Full-stack production compose for Docker (frontend, backend, nginx, mongo; optional neo4j).
  • podman-compose.prod.yml: Equivalent full-stack compose for Podman (same services, no version key).
  • backend.Dockerfile: Multi-stage backend image with healthcheck.
  • frontend.Dockerfile: Multi-stage Next.js image with healthcheck.
  • agentic-ai.Dockerfile: Multi-stage Agentic AI + MCP image with healthcheck.
  • nginx/: Reverse proxy that routes /api + /trpc to the backend and all other traffic to the frontend.

Prerequisites

  • Docker + Docker Compose (v2+), or Podman (4.1+) with podman-compose or the podman compose plugin.
  • A root .env file based on /.env.example.

Quick Start

  1. Create and edit your root .env:

    • Copy /.env.example to /.env and fill in real values.
    • At minimum, set MONGO_URI, JWT_SECRET, GOOGLE_AI_API_KEY, PINECONE_API_KEY, PINECONE_INDEX.

  2. Build and run:

    # Docker
docker compose -f docker/compose.prod.yml --env-file .env up --build -d

# Podman
podman compose -f docker/podman-compose.prod.yml --env-file .env up --build -d

  3. Access:

    • App: http://localhost
    • API docs: http://localhost/api-docs
    • Metrics: http://localhost/metrics
sequenceDiagram
  participant Browser
  participant Nginx
  participant Frontend
  participant Backend
  participant Mongo
  Browser->>Nginx: GET /
  Nginx->>Frontend: Proxy to Next.js
  Browser->>Nginx: POST /api/chat
  Nginx->>Backend: Proxy to API
  Backend->>Mongo: Query/Write
  Backend-->>Browser: JSON response
Loading

Ports & Endpoints

Service Port Internal Healthcheck
Nginx 80 80 N/A
Frontend N/A 3000 GET /
Backend N/A 3001 GET /metrics
Agentic AI 4318 (optional) 4318 GET /health
Mongo N/A 27017 mongosh ping
Neo4j 7474/7687 (optional) 7474/7687 N/A

Environment Configuration

Minimum variables for core stack (in .env):

  • MONGO_URI, JWT_SECRET, GOOGLE_AI_API_KEY, PINECONE_API_KEY, PINECONE_INDEX

Optional variables:

  • OPENAI_API_KEY, NEO4J_*, GOOGLE_AI_MODEL, OPENAI_MODEL
  • NEXT_PUBLIC_API_BASE_URL (frontend build-time)
  • API_BASE_URL, FRONTEND_BASE_URL (MCP defaults for agentic)
  • NGINX_PORT, AGENTIC_PORT, NEO4J_HTTP_PORT, NEO4J_BOLT_PORT
flowchart TB
  subgraph Profiles
    Core[default] --> Mongo
    Core --> Backend
    Core --> Frontend
    Core --> Nginx
    Core -. optional .-> Agentic
    Core -. optional .-> Neo4j
  end
Loading

Neo4j (optional)

Enable the Neo4j container via profile and update the .env values:

# Docker
docker compose -f docker/compose.prod.yml --profile graph --env-file .env up -d

# Podman
podman compose -f docker/podman-compose.prod.yml --profile graph --env-file .env up -d

Agentic AI (optional)

Enable the Agentic AI HTTP server via profile:

# Docker
docker compose -f docker/compose.prod.yml --profile agentic --env-file .env up -d

# Podman
podman compose -f docker/podman-compose.prod.yml --profile agentic --env-file .env up -d

Then call POST /run on http://localhost:4318 with { goal, runtime?, rounds?, threadId? }. Ensure your .env provides at least one of GOOGLE_AI_API_KEY or OPENAI_API_KEY, plus any Pinecone/Neo4j settings you plan to use in tools.

sequenceDiagram
  participant Client
  participant Agentic
  participant MCP
  participant Backend
  Client->>Agentic: POST /run {goal, runtime}
  Agentic->>MCP: Tool calls
  MCP->>Backend: API requests
  Backend-->>MCP: JSON data
  MCP-->>Agentic: Tool results
  Agentic-->>Client: Report/summary
Loading

Notes

  • mongo runs locally by default and backs the backend via mongodb://mongo:27017/estatewise.
  • If you use an external MongoDB, set MONGO_URI accordingly and remove/disable the mongo service.
  • The frontend currently hardcodes the backend URL in frontend/lib/api.ts and frontend/pages/chat.tsx. For a fully self-contained Docker deployment, update those to read from process.env.NEXT_PUBLIC_API_BASE_URL (or use relative /api paths) so NEXT_PUBLIC_API_BASE_URL=http://localhost can be injected at build time.

Operational Guidance

  • Healthchecks are baked into backend, frontend, and agentic images.
  • Use docker compose logs -f backend (or podman compose logs -f backend) to inspect API startup and Mongo connectivity.
  • For production TLS, terminate HTTPS in front of Nginx (load balancer or managed ingress).
  • Scale backend/agentic with replicas behind Nginx if needed (external orchestrator required).

Stopping

# Docker
docker compose -f docker/compose.prod.yml down

# Podman
podman compose -f docker/podman-compose.prod.yml down

Cleanup

To remove all containers, networks, and volumes created by the compose file:

# Docker
docker compose -f docker/compose.prod.yml down -v

# Podman
podman compose -f docker/podman-compose.prod.yml down -v

Podman Notes

  • Podman 4.1+ is required for BuildKit-style --mount=type=cache directives in the Dockerfiles.
  • Podman reads Dockerfile natively — no Containerfile symlinks are needed.
  • .containerignore files are included alongside .dockerignore for Podman tooling that checks for them.
  • If you hit permission issues with rootless Podman, add :Z to volume mounts for SELinux relabeling (e.g., ./logs:/var/log/mcp:Z).
  • podman-compose (PyPI) and the built-in podman compose plugin both work. The plugin ships with Podman 4.7+.

For more advanced setups (e.g., Kubernetes, cloud deployments), refer to the respective documentation in the repository.