LO Lucas Oliveira
Repository Portfolio ↗
ATLAS / BUILD 0.2 In development

Architecture & Tech Lead AI System

From production signals to explainable technical decisions.

A human-in-the-loop engineering copilot for architecture analysis, troubleshooting and code intelligence. ATLAS turns sanitized context into traceable hypotheses, system boundaries and prioritized action plans.

Explore architecture ↓ View delivery roadmap
HUMAN APPROVALREAD-ONLY BY DEFAULTEVIDENCE FIRSTAI OUTSIDE THE CORE

01 / PRODUCT SCOPE

A decision system for the Tech Lead workflow.

The first release focuses on architecture interviews and production troubleshooting. It advises, cites assumptions and proposes reversible actions; it never changes production automatically.

01

Architecture

Capture context and constraints, model boundaries and explain technology trade-offs.

SYSTEM DESIGN · ADR · RISKS
02

Troubleshooting

Correlate symptoms, telemetry and runbooks into ranked, testable hypotheses.

METRICS · LOGS · TRACES
03

Code intelligence

Inspect approved repositories to identify coupling, obsolescence and evolution opportunities.

GITHUB · QUALITY · UPGRADES
04

Action planning

Return P0–P2 recommendations with evidence, limitations and human approval gates.

RUNBOOK · PRIORITY · AUDIT

02 / SYSTEM ARCHITECTURE

Decoupled intelligence.
Controlled operational access.

The Core API owns policy, sessions and auditability. Model providers, retrieval and tools remain isolated behind the Intelligence Service.

Implemented or baseline ready Planned boundary → Controlled request → Persistence / telemetry / retrieval

03 / DELIVERY ROADMAP

Progress backed by repository evidence.

Eight product milestones track the path from demonstrative prototype to a production-oriented, observable and auditable engineering assistant.

01
COMPLETE

Scope, guardrails and polyrepo boundaries

Product principles, repositories and responsibility boundaries documented.

02
COMPLETE

Interactive Architecture Command Center

Scenario intake, component map, recommendations and troubleshooting runbook with frontend tests and CI.

03
COMPLETE

Core API foundation

Java 21, Spring Boot 4.1.1, PostgreSQL/Flyway, JWT, Modulith, OpenAPI, Testcontainers and telemetry baseline.

04
IN PROGRESS

Diagnostic domain and frontend integration

Persist real sessions and replace the demonstrative frontend diagnosis with versioned API contracts.

05
NEXT

Model Gateway and structured analysis

Provider-independent model routing, structured outputs, confidence and policy enforcement.

06
PLANNED

RAG and repository intelligence

Approved knowledge, pgvector retrieval, citations and read-only repository analysis.

07
PLANNED

MCP and troubleshooting evidence

Restricted adapters for engineering tools, logs, metrics and traces with sensitive-data controls.

08
PLANNED

Observability, evaluations and hardening

Quality evaluations, dashboards, alerts, resilience tests, security review and deployment automation.

04 / ENGINEERING DECISIONS

Production thinking before AI autonomy.

ADR / 01

AI outside the transactional core

Business requests never depend on an LLM. ATLAS is an advisory path, so model latency or failure cannot break a production transaction.

ISOLATION · AVAILABILITY · COST
ADR / 02

Human approval before mutation

Rollback, scaling, deployment and configuration changes require explicit approval and an audit trail.

SAFETY · ACCOUNTABILITY · BLAST RADIUS
ADR / 03

Modules before more services

RAG and MCP begin inside the Intelligence Service. They split only when scaling, security or release cadence provides measurable value.

BOUNDARIES · COMPLEXITY · EVOLUTION

PROJECT STATUS

The foundation is running.
The intelligence layer is next.

Follow the implementation through the public repository ecosystem. Each phase will add code, tests, architecture decisions and measurable evidence.

Explore ATLAS on GitHub ↗