Open source ad serving platform with machine learning powered CTR prediction and self-hosted deployment
- Stars196
- Forks32
- Open Issues1
Apache-2.0
- Python
- Makefile
- Dockerfile

About OpenAdServer
OpenAdServer is a self-hosted ad serving platform aimed at SMBs, startups, and developers who want to run their own ad network without Google-scale infrastructure. It exposes an HTTP API that handles the core ad request, ranking, and event tracking flow.
It provides eCPM-based ranking for CPM, CPC, and CPA bids, targeting by geo, device, and demographics, frequency capping, and impression, click, and conversion tracking. CTR prediction uses DeepFM and logistic regression models, with Prometheus and Grafana for monitoring.
It is written in Python with FastAPI and PyTorch and deploys via Docker Compose under the Apache 2.0 license. The project is young, and some pieces such as an admin dashboard UI and OpenRTB support are still on the roadmap.
Key features
- HTTP ad serving API on FastAPI
- eCPM ranking for CPM, CPC, and CPA
- Targeting by geo, device, and demographics
- Impression, click, and conversion tracking
- DeepFM and logistic regression CTR prediction
Details
- On GitHub since
- 2025
- Language
- Python · FastAPI · PyTorch
- License
- Apache-2.0
- Platforms
- Web · Docker
- Monitoring
- Prometheus · Grafana