- Add Section 3 documenting Hermes model routing (deepseek-chat / deepseek-v4-pro) - Add Section 3.2 with 5 project-specific skills (hotel-project-map, hotel-remuneration-orchestrator, hotel-seeding-testing, hotel-rls-verification, hotel-n8n-deployment) - Add Section 3.3-3.6 covering AGENTS.md, MCP servers, cron test runner, and .hermes/ plans directory - Add Section 3.7 quick-reference table for all common commands - Update AGENTS.md to reference Hermes skills instead of inline orchestration, add worktree mode guidance - Bump AGENTS.md changelog for migration date
13 KiB
AI & Developer Workstation Guide
Variable Remuneration, Compensation, and Commissions System - Hoteles Estelar
This document outlines the environment configurations (Development vs. Production), active API keys, and recommended AI integration tools (like Model Context Protocol servers) designed to optimize this workspace for AI agents and human developers alike.
1. Environment Configurations (.env Blueprints)
The Next.js application separates environment secrets between local active development, development containers, and production servers. All credentials reside in a centralized .env file at the project root.
1.1. Dual-Stack Configuration Layout (.env)
This file configs connection parameters for both production (app-prod) and development (app-dev) containers.
# -----------------------------------------------------------------------------
# Database Setup (Fallback / Local Development)
# -----------------------------------------------------------------------------
DATABASE_URL="postgresql://special_hotel_user:***@pg.gaboggamer.online/special_hotel_dev?schema=public"
# -----------------------------------------------------------------------------
# Production Stack Environment Variables (for Docker Compose app-prod)
# -----------------------------------------------------------------------------
# Connection string pointing to the production PostgreSQL instance
DATABASE_URL_PROD="postgresql://special_hotel_user:***@pg.gaboggamer.online/special_hotel?schema=public"
# Production domain resolved by the Caddy reverse proxy
NEXT_PUBLIC_APP_URL="https://hotels.gaboggamer.online"
# Secret token used to sign NextAuth / custom session JWTs in production
NEXTAUTH_SECRET="0qqPRY4NIbCEFag33Q6EB1ea7dUQR1J6Z8h4NogrgCg="
# Production n8n calculation webhook endpoint
N8N_WEBHOOK_URL="https://n8n.gaboggamer.online/webhook/calculate-commissions"
# Token to authorize and verify n8n webhook payload signatures in production
N8N_WEBHOOK_SECRET="Ecjb2s33tHJppNBDJ/DxXEjHWKow8bNWmsQrk1sQKyQ="
# Internal container networking URL for production app mapping
APP_PROD_INTERNAL_URL="http://special-hotel-prod:3000"
# -----------------------------------------------------------------------------
# Development Stack Environment Variables (for Docker Compose app-dev)
# -----------------------------------------------------------------------------
# Connection string pointing to the development PostgreSQL instance
DATABASE_URL_DEV="postgresql://special_hotel_user:***@pg.gaboggamer.online/special_hotel_dev?schema=public"
# Dedicated connection string for the isolated test/sandbox database
TEST_DATABASE_URL="postgresql://special_hotel_user:***@pg.gaboggamer.online/special_hotel_test?schema=public"
# Local host address for development
NEXT_PUBLIC_APP_URL_DEV="http://localhost:3001"
# JWT session signing key for development
NEXTAUTH_SECRET_DEV="TdD2xx0rZYCGkYxFFB7y9Sm8L+HGyXaXInqB9lYLJsk="
# Sandbox n8n testing webhook endpoint
N8N_TEST_WEBHOOK_URL="https://n8n.gaboggamer.online/webhook-test/calculate-commissions"
# Signature token to authorize development webhook payloads
N8N_WEBHOOK_SECRET_DEV="qUHuPqPjA65psdtwQU7zgp/DVkvd1xXk2WP/vzbnEdc="
# Internal container networking URL for development app mapping
APP_DEV_INTERNAL_URL="http://special-hotel-dev:3000"
2. Optimizing the Workspace for the Antigravity Agent (agy)
To make this codebase highly friendly for the Antigravity agent (agy), you can configure the following Model Context Protocol (MCP) servers. Once registered in the agent's active configuration, agy is equipped with specialized tools to inspect database schemas, execute and debug n8n workflows, manage git versions, and run automated browser checks.
2.1. Recommended MCP Servers for agy
A. n8n MCP Server (n8n)
Exposes tools to read, execute, and write workflows directly on the n8n canvas.
- Use Case: Allows
agyto trigger calculation runs, inspect failing nodes on the canvas, check webhook logs, and modify workflows dynamically. - Harness Registration:
"n8n": { "command": "npx", "args": ["-y", "n8n-mcp"], "env": { "N8N_API_KEY": "your_n8n_api_key_here", "N8N_URL": "https://n8n.gaboggamer.online" } }
B. Prisma Postgres MCP Server (prisma-postgres)
Exposes tools allowing agy to interact with the database using type-safe schemas.
- Use Case: Allows
agyto run schema checks, dry-run validations, and automatically inspect database tables during development. - Harness Registration:
"prisma-postgres": { "command": "npx", "args": ["-y", "@prisma/mcp"], "env": { "DATABASE_URL": "postgresql://special_hotel_user:***@pg.gaboggamer.online/special_hotel_dev?schema=public" } }
C. PostgreSQL MCP Server (postgres)
Provides raw PostgreSQL connection and querying tools.
- Use Case: Enables
agyto query migration status, seed verification, and raw audit log verification directly. - Harness Registration:
"postgres": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-postgres", "postgresql://special_hotel_user:***@pg.gaboggamer.online/special_hotel_dev"] }
D. Git MCP Server (git)
Provides local Git operations tools (clone, commit, diff, log, status).
- Use Case: Allows
agyto review local branches, examine diffs of modified code files, and make structured, micro-commits during development. - Harness Registration:
"git": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-git", "/home/gabogg/Proyects/semillero-special-hotel"] }
E. Puppeteer MCP Server (puppeteer)
Exposes browser automation tools (take screenshots, click, type, fill forms).
- Use Case: Allows
agyto start a headless browser, render our Next.js pages, and verify layout responsiveness and style details against the style guide without manual developer steps. - Harness Registration:
"puppeteer": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-puppeteer"] }
3. Hermes Agent Integration
Hermes Agent is the primary AI agent used for development on this project alongside Antigravity. It runs locally with full tool access and has been configured with project-specific skills, model routing, and automated testing.
3.1. Model Routing
Hermes uses a two-tier model setup to save cost while keeping heavy tasks powerful:
| Mode | Model | Purpose |
|---|---|---|
| Main chat | deepseek-chat (fast, cheap) |
Normal conversation, file edits, simple queries |
| Delegation | deepseek-v4-pro (powerful) |
Complex debugging, code review, research, multi-step tasks |
How it works: When Hermes needs to do something complex, it spawns a subagent via delegate_task. That subagent automatically gets the pro model. Your normal chat stays fast and cheap. Configured in ~/.hermes/config.yaml:
model:
default: deepseek-chat
provider: deepseek
delegation:
model: deepseek-v4-pro
provider: deepseek
To manually switch models mid-session: /model deepseek-v4-pro or /model deepseek-chat.
3.2. Project-Specific Hermes Skills
Five Hermes skills have been created for this project. They live at ~/.hermes/skills/software-development/ and are loaded on demand with skill_view(name="skill-name").
| Skill | Load With | Purpose |
|---|---|---|
hotel-project-map |
Any session start | Full structural index: schema models, API routes, pages, deps, key libs |
hotel-remuneration-orchestrator |
Compensation/commission tasks | 3-agent orchestrator (planner → calculator → QA auditor) |
hotel-seeding-testing |
Before tests or seeding | Exact commands for dev/test DB seed, individual tests, full suite |
hotel-rls-verification |
RLS / auth work | Role isolation testing scripts, manual verification queries |
hotel-n8n-deployment |
n8n workflow work | Bootstrap scripts, workflow JSONs, verification test |
Recommended loading order for a new session:
hermes -s hotel-project-map,hotel-remuneration-orchestrator
Or inside an existing session:
/skill hotel-project-map
/skill hotel-remuneration-orchestrator
The skills contain exact commands, edge cases, and pitfalls so you (or the agent) don't rediscover them.
3.3. AGENTS.md
The file AGENTS.md at the project root is loaded automatically when Hermes works from this directory (either by launching from here or via a cron job with workdir set). It covers:
- Caveman Mode (Lite): Professional, precise, no filler
- RTK Command Proxying: All shell commands proxied through
rtkwhen available to reduce token consumption - MCP-specific guidelines: n8n silent execution, Forgejo PR integrity, Prisma schema validation
- Hermes skills reference: The 5 skills above with loading guidance
- Git Worktree Mode:
hermes -wfor parallel agent sessions without conflicts
3.4. MCP Server Configuration
The MCP servers configured for Hermes are stored in .agents/mcp_config.json. These are the same MCP servers from Section 2 but registered for Hermes. Currently configured:
- n8n →
n8n-mcppackage, pointed at n8n.gaboggamer.online - Prisma →
@prisma/mcpfor type-safe DB introspection - PostgreSQL → direct
server-postgresfor raw queries - Git →
mcp-gitpointed at the project directory - Puppeteer →
@modelcontextprotocol/server-puppeteerfor browser automation - Forgejo →
forgejo-mcpfor PR and issue management on git.gaboggamer.online
3.5. Automatic Test Runner (Cron)
A Hermes cron job runs the full test suite every 6 hours:
- Job ID:
dd941135c090 - Name:
hotel-test-runner - Schedule: Every 6 hours
- Workdir:
/home/gabogg/Proyects/semillero-special-hotel(loads AGENTS.md) - Skills loaded:
hotel-seeding-testing,hotel-project-map - Action: Seed test DB →
rtk npm run test:all→ report pass/fail
Check status:
hermes cron list
The cron job delivers results to the same conversation. If all tests pass, it summarises. If any fail, it gives error details.
3.6. .hermes/ Directory
The project contains a .hermes/ directory:
.hermes/plans/— stores Hermes plan files (used with theplanskill for executable markdown plans)
3.7. Quick Reference for Common Tasks
| Task | Command |
|---|---|
| Seed dev DB | rtk npm run db:seed |
| Seed test DB | rtk npm run db:seed-test |
| Run all tests | rtk npm run test:all |
| Run RLS tests | rtk node prisma/test-rls.js |
| Run auth+RLS tests | rtk node prisma/test-auth-rls.js |
| Run UI tests | rtk npm run test:ui |
| Run n8n tests | rtk node prisma/test-n8n-real.js |
| Bootstrap n8n workflows | rtk node scripts/n8n-bootstrap.js |
| Load Hermes project map | /skill hotel-project-map in session |
| Load orchestrator | /skill hotel-remuneration-orchestrator |
| Load seeding/testing skill | /skill hotel-seeding-testing |
| Switch to pro model | /model deepseek-v4-pro |
| Start Hermes with skills | hermes -s hotel-project-map from project dir |
| Create a plan | Create .hermes/plans/<name>.md with actionable steps |
4. Testing Procedures & Webhook Testing Flow
Testing must occur in complete isolation from production data. We achieve this by splitting execution paths using dedicated test hooks in both Next.js and n8n.
4.1. Testing Scripts Command List
Run these commands from the repository root using pnpm (or npm / yarn):
- Run Row-Level Security Tests:
Tests tenant read boundaries across admin, gerente, and colaborador roles, and verifies that update/delete actions onpnpm run test:rlsaudit_logsare blocked. - Run Auth & API RLS Integration Tests:
Runs a Next.js instance on test portpnpm run test:auth-rls3009and asserts cookies, authorization blocks, and API-level data filtration. - Run UI End-to-End Tests:
Compiles Next.js and runs automated Puppeteer scripts (pnpm run test:uitest-phase3-ui.js,test-phase4-ui.js,test-phase5-ui.js) to verify pages, styles, translations, and modals. - Run n8n Webhook Integration Tests:
Fires sales validation and settlement calculation jobs directly at the n8n webhook test endpoint.pnpm run test:n8n - Run All Tests:
pnpm run test
4.2. n8n Testing Webhook Routing
All calculations triggered by test suites route to n8n via /webhook-test path segments:
- Trigger: Test suite invokes n8n via
POST ${process.env.N8N_TEST_WEBHOOK_URL}/calculate-commissions. - n8n Path Branching:
- Within the n8n canvas, an
IFnode checks:{{ $json.headers["x-nginx-original-uri"] || $json.path }}containswebhook-test. - True: The n8n workspace connects to the database utilizing
TEST_DATABASE_URLcredentials. It pulls test sales/goals data and pushes calculations back to the Next.js dev API. - False: Connects to the main
DATABASE_URLfor production processing.
- Within the n8n canvas, an
4.3. Manual Integration Setup
To sync and seed the testing database structure:
- Spin up both databases.
- Run database schema migrations on the test database:
pnpm run db:seed-test - Run the development server or test suites.