One repo. Nine production-ready AI modules. REST API + React frontend + one-command Docker.
Built for engineers who ship — not decks.
AgentX is a modular full-stack AI system that packages nine recurring high-value engineering tasks into independently deliverable, demo-ready modules. Every module ships behind the same clean REST API + React demo UI + Docker deployment, so you can go from "here's what I can do" to "here's it running" in under five minutes.
🎯 Built around real project work: text classification, NLP phase detection, probability calibration, job scheduling, path optimization, computational geometry, ETL/OCR pipelines, financial modeling, and a full-stack API layer.
graph TB
subgraph Client["Frontend"]
UI["React Demo UI<br/>module playground"]
end
subgraph API["API Layer"]
GW["FastAPI Gateway<br/>auth - rate-limit - docs"]
end
subgraph Core["Nine Modules"]
M1["TextClassifier"]
M2["PhaseDetect"]
M3["Calibrator"]
M4["Scheduler"]
M5["PathOpt"]
M6["GeoEngine"]
M7["ETL/OCR"]
M8["FinModel"]
M9["REST API"]
end
subgraph Infra["Infra"]
DB[("SQLite / Postgres")]
DOCS["OpenAPI Docs"]
end
UI --> GW
GW --> Core
Core --> DB
GW --> DOCS
style Client fill:#e1f5ff,stroke:#0099ff
style API fill:#fff4e1,stroke:#ff9900
style Core fill:#e8ffe8,stroke:#00bb00
style Infra fill:#ffe1e1,stroke:#dd0000
| # | Module | What it does | Tech stack |
|---|---|---|---|
| 1 | Text Classifier | Multi-class EN/ZH text classification | TF-IDF - LogReg - SVM - scikit-learn |
| 2 | Phase Detector | Sentiment + topic + entity extraction | Lexicons - TF-IDF - KMeans - regex |
| 3 | Calibrator | Probability calibration (Platt / Isotonic) + reliability diagrams | scikit-learn - Brier - log-loss |
| 4 | Scheduler | Job scheduling, WSPT priorities, parallel machines | Kahn sort - priority queues |
| 5 | Path Optimizer | Shortest path / routing optimization | NetworkX - heuristic search |
| 6 | Geo Engine | Computational geometry - spatial queries | shapely - SciPy |
| 7 | ETL / OCR | Multi-source ingestion, cleaning, quality gates, OCR | pandas - SQLAlchemy - OCR |
| 8 | Financial Model | Time-series forecasting, risk metrics, portfolio analytics | statsmodels - pandas - numpy |
| 9 | REST API | Full OpenAPI layer - every module as an endpoint | FastAPI - Pydantic - Docker |
- 9 independent modules - demo one, quote one, deliver one
- One-command Docker -
docker-compose upand the whole stack is live - Tested - every module has unit tests + example inputs
- OpenAPI docs - interactive Swagger UI out of the box
- Modular by design - add your own module by dropping a folder
# Clone
git clone https://github.com/LeonxLJX/AgentX.git
cd AgentX
# Run with Docker (recommended)
docker-compose up --build
# Or run locally
pip install -r requirements.txt
uvicorn agentx.main:app --reloadThen open:
| What | URL |
|---|---|
| Demo UI | http://localhost:8000 |
| API Docs (Swagger) | http://localhost:8000/docs |
| Health check | http://localhost:8000/health |
AgentX/
├── agentx/ # core package
│ ├── modules/ # the 9 modules, one folder each
│ ├── api/ # FastAPI routes + schemas
│ └── core/ # config, logging, persistence
├── examples/ # sample inputs + outputs
├── tests/ # unit tests
├── docs/ # architecture notes
├── tools/ # scripts & utilities
└── docker-compose.yml
| Layer | Tools |
|---|---|
| Language | Python 3.10+ |
| API | FastAPI - Pydantic - Uvicorn |
| ML / AI | scikit-learn - pandas - numpy - SciPy - statsmodels |
| Data | SQLite - SQLAlchemy - pandas |
| Infra | Docker - docker-compose |
| Quality | pytest - ruff - mypy |
This project is a demonstration of end-to-end AI engineering: from data ingestion and model selection, through API design and testing, to containerized deployment and a working demo UI.
Built by Xin Liu (Leon) - AI & full-stack engineer. Master's in Finance (WorldQuant). I build LLM agents, RAG systems, and data pipelines that ship.
GitHub: @LeonxLJX