GerenciamentoDeVeiculos

mcp
Security Audit
Warn
Health Warn
  • No license — Repository has no license file
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 6 GitHub stars
Code Pass
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

Vehicle Intelligence Platform exploring Data Engineering, SENATRAN fleet analytics, Machine Learning for vehicle price prediction, FastAPI, and MCP integration for AI agents.

README.md

Vehicle Intelligence Platform

Experimental platform for vehicle data management, analysis, and applied intelligence.

The project evolved from a traditional CRUD system into an architecture integrating:

  • Backend Node.js / Express
  • React
  • PostgreSQL
  • Data Engineering
  • SENATRAN data
  • Machine Learning with TensorFlow
  • ML API with FastAPI
  • Model Evaluation
  • Model Comparison
  • MCP Server
  • MCP → ML integration

The goal is to explore an architecture in which data, analytics, and Machine Learning models can eventually be exposed as capabilities for LLM- and Agent-based systems.


Architecture

                         ┌─────────────────────┐
                         │       React         │
                         │     Frontend        │
                         └──────────┬──────────┘
                                    │
                                    ▼
                         ┌─────────────────────┐
                         │   Node.js / Express │
                         │       Backend       │
                         └──────────┬──────────┘
                                    │
                    ┌───────────────┼────────────────┐
                    │               │                │
                    ▼               ▼                ▼
             ┌────────────┐  ┌────────────┐  ┌────────────┐
             │ PostgreSQL │  │   ML API   │  │ MCP Server │
             │            │  │  FastAPI   │  │ TypeScript │
             └────────────┘  └─────┬──────┘  └─────┬──────┘
                                   │               │
                                   ▼               │
                              ┌──────────┐         │
                              │TensorFlow│         │
                              └──────────┘         │
                                                   │
                                   ┌───────────────┘
                                   │
                                   ▼
                            Future Agent / LLM

## Backend

Backend developed with:

- Node.js
- Express
- Sequelize
- PostgreSQL
- JWT
- Zod
- Helmet
- Rate Limiting
- Swagger

The backend provides the application's main API and integrates the Machine Learning service.

Machine Learning

The project includes a Machine Learning pipeline for vehicle price estimation.

Pipeline

Raw Data
   ↓
Preprocessing
   ↓
Feature Engineering
   ↓
Training
   ↓
Model Evaluation
   ↓
TensorFlow Model
   ↓
FastAPI

Features used

  • year

  • engine_cc

  • mileage_km

  • doors

  • vehicle_age

  • km_per_year

  • make

  • fuel_type

  • transmission

  • body_type

  • state

TensorFlow Model

Regression model developed with TensorFlow/Keras.

Main architecture:

Input
  ↓
Dense(64, ReLU)
  ↓
Dense(32, ReLU)
  ↓
Dense(1)

Training with:

  • Adam
  • Learning rate: 0.001
  • MSE loss
  • MAE metric
  • Early Stopping
  • Validation split

SENATRAN Data Engineering

The project uses public SENATRAN data to analyze the Brazilian vehicle fleet.

Pipeline

SENATRAN
   ↓
Ingestion
   ↓
Cleaning
   ↓
Classification
   ↓
Aggregation
   ↓
Feature Engineering
   ↓
Analytics

Categories used

  • CARRO
  • MOTO
  • PESADO
  • IMPLEMENTO
  • NAO_CLASSIFICADO
  • OUTRO

There are also pipelines for:

  • aggregation by state
  • aggregation by municipality
  • vehicle classification
  • fleet distribution
  • feature generation
  • Machine Learning integration experiments

ML API

Price prediction is provided through a FastAPI API.

Endpoint: POST /predict

{
  "make": "Toyota",
  "year": 2023,
  "engine_cc": 2000,
  "mileage_km": 25000,
  "doors": 4,
  "fuel_type": "flex",
  "transmission": "automatico",
  "body_type": "sedan",
  "state": "SP"
}

Response:

{
  "predicted_price": 102355.78,
  "model": "tensorflow"
}

MCP Server

The project includes an MCP Server developed in TypeScript.

The goal of MCP is to transform system data and capabilities into structured tools that can eventually be used by Agents.

MCP Resources

Currently:

senatran://uf/{uf}
Example:
senatran://uf/PI

MCP Tools

get_fleet_by_state

Queries the vehicle fleet of a state.
get_fleet_by_state("PI")

compare_states

Compares the fleets of two states.
compare_states("PI", "SP")

get_vehicle_distribution

Calculates the fleet's percentage distribution.
get_vehicle_distribution("PI")

predict_vehicle_price

Uses the Machine Learning model through the ML API.
predict_vehicle_price(...)

MCP Client
    ↓
MCP Server
    ↓
predict_vehicle_price
    ↓
FastAPI
    ↓
TensorFlow
    ↓
Prediction

MCP + Machine Learning

One of the project's architectural goals is to separate responsibilities:

MCP
 └── Capability exposure

Analytics
 └── Data analysis

ML
 └── Predictions

LLM / Agent
 └── Future orchestration

This way, the Machine Learning model is not directly coupled to the Agent.

The Agent may eventually decide when to use:

Vehicle Data
Fleet Analytics
ML Prediction
External Data

through the capabilities exposed by MCP.

Testing

Automated tests and validation scripts are available in:

tests/

Example: pytest

Running the Project

Backend

cd backend
npm install
npm run dev

Frontend

cd frontend
npm install
npm run dev

ML API

In the configured Python environment:

uvicorn ai.api.main:app --host 0.0.0.0 --port 8003

Health check:

curl http://localhost:8003/health

MCP Server

cd backend/mcp

npm install
npm run dev

The server uses STDIO for MCP communication.

For inspection and testing:

npx @modelcontextprotocol/inspector

Tech Stack

Área Tecnologias
Backend Node.js, Express, Sequelize, PostgreSQL, Zod, JWT
Frontend React, Vite
Data & ML Python, Pandas, Scikit-learn, TensorFlow, FastAPI, Joblib
AI Integration Model Context Protocol, TypeScript, MCP Inspector
Data Sources SENATRAN, FIPE, IBGE

This project is an architecture laboratory exploring the evolution of a traditional application into a data- and intelligence-oriented platform.

The focus is not only on price prediction.

The proposal is to build an architecture in which:

Data
 ↓
Analytics
 ↓
Machine Learning
 ↓
MCP
 ↓
LLM
 ↓
Agent

can be used in a modular and evolutionary manner.

Author: Silas Vasconcelos Cruz

Full Stack Engineer | Enterprise Modernization | DevSecOps | Applied AI
Experimental engineering project focused on Data Engineering, Machine Learning, MCP, and Agentic AI architecture.

Reviews (0)

No results found