Globot

Open-source decision-support workspace for supply-chain disruption review.

Globot helps operators inspect disruption signals, compare route options, review compliance documents, and keep a human accountable for every consequential recommendation. It is built as a transparent research and demonstration system, not as an autonomous execution engine.

Default demos use mock and synthetic data. Outputs must be independently reviewed before they inform operational, compliance, or financial decisions.

Why this project matters

Shipping and trade disruptions affect food, energy, medical supply, and manufacturing lead times. Existing tooling is often either a static dashboard or a black-box model. Globot sits in between:

• Inspectable multi-agent activity — market, logistics, compliance, and hedging perspectives in one review flow
• Human-in-the-loop by design — recommendations cannot be treated as actions without explicit approval
• Reproducible open demo — local fixtures and documented scenarios so contributors can verify behaviour
• Pluggable integrations — mock sources today; production adapters tomorrow (AIS, news, insurance, market data)

The repository is an ecosystem building block for responsible agentic logistics research: a shared reference for HITL controls, provenance-friendly summaries, and maintenance-friendly FastAPI + React structure.

Table of contents

• Current status
• Core capabilities
• Quick start
• Configuration
• Architecture
• Roadmap
• Contributing
• License

Current status

Globot is actively maintained as a public research / demonstration project.

Area	Status
Stars / forks	Tracked on GitHub; community PRs welcome
License	MIT
CI	Backend syntax + frontend production build on main and PRs
Default data	Mock / synthetic — not live trading, sanctions, or execution
Production readiness	Not production-ready without deployers adding auth, secrets, data licenses, and review controls

Core capabilities

1. Scenario-driven demo — reproducible disruption walkthrough with route selection and approval steps
2. Multi-perspective review — market signals, route risk, compliance document experiments, hedging calculations
3. Map-centric visualisation — Deck.gl-based logistics map for ports, routes, and risk highlighting
4. Human approval gate — Approve / details / override path before a recommendation is treated as an action
5. Optional model providers — Ollama by default; OpenAI and other providers remain opt-in via server-side env vars

Quick start

Prerequisites

• Python 3.11
• Node.js 20+
• pnpm (lockfile: frontend/pnpm-lock.yaml)

Backend

cd backend
python -m venv .venv
# Windows PowerShell
.\.venv\Scripts\Activate.ps1
# macOS/Linux
# source .venv/bin/activate

pip install -r requirements.txt
cp .env.example .env   # Windows: copy .env.example .env
python start_server.py

API: http://localhost:8000

Frontend

cd frontend
pnpm install --frozen-lockfile
cp .env.example .env   # Windows: copy .env.example .env
pnpm dev

UI: http://localhost:5173

Demo path

1. Open http://localhost:5173/pay
2. Choose Watch Demo → route selection (/port)
3. Confirm a route (default Shanghai → Rotterdam) → Start Simulation (/demo)
4. Follow the activity feed and use Approve & Execute only after human review

Configuration

Copy example env files before local edits. Never commit API keys, Clerk credentials, personal allowlists, uploads, or production data. Defaults contain no credentials and enable no administrator.

File	Purpose
backend/.env.example	API, optional LLM providers, Clerk issuer, allowlist
frontend/.env.example	Vite public Clerk key stub and API base URL

backend/data/ holds demonstration assets. Treat all outputs as illustrative unless you replace those sources with licensed, verified data and add operational controls.

Architecture

React / Vite frontend (review UI + map)
              |
              v
FastAPI backend
   |        |         |
 demo   risk/route   documents & hedging experiments
   |
 mock scenarios + local knowledge assets

Longer-term maintainer automation and opt-in OpenAI usage: docs/OPENAI_API_PLAN.md.

Project health

Signal	Detail
License	OSI-approved MIT
Community files	CONTRIBUTING, SECURITY, CODE_OF_CONDUCT, GOVERNANCE, SUPPORT
Automation	GitHub Actions CI + Dependabot
Maintainer	Primary maintainer @Vector897 with write access
Codex plan	docs/OPENAI_API_PLAN.md

Roadmap

See docs/ROADMAP.md. Near-term priorities:

• Repeatable API and scenario tests in CI
• Explicit, opt-in provider integrations with evaluation fixtures
• Provenance and review records for every UI recommendation
• Maintainer automation for issue triage, PR review summaries, release notes, and dependency hygiene

Contributing

Please read CONTRIBUTING.md before opening an issue or PR.

Doc	Purpose
SECURITY.md	Vulnerability reporting
CODE_OF_CONDUCT.md	Community standards
GOVERNANCE.md	Maintainer decision-making
SUPPORT.md	How to get help
CHANGELOG.md	Notable changes
AGENTS.md	Guidance for automated coding agents

Local checks before review:

python -m compileall -q backend
cd frontend && pnpm build

License

Globot is released under the MIT License.

Primary maintainer: Vector897