Home.en - houyhlouis/LLMOJ GitHub Wiki
LLMOJ Wiki
įŽäŊä¸æ | English | Repository
LLMOJ is an online judge based on LibreOJ, with AI tools for problem setting. New here? Start with Installation. To find a particular setting, see the Configuration overview.
Installing and maintaining
| Page | Contents |
|---|---|
| Installation | Installation scripts, every option, and retrying a failed installation |
| Upgrade | Upgrading while keeping configuration and data; compatibility and rollback |
| Feature upgrades | The database changes of LLMOJ's features, and applying them to a custom installation |
| HTTPS | Domains, TLS certificates and external reverse proxies |
| Docker sources | Docker package sources and registry mirrors |
| Operations | Nginx, MariaDB, Redis, MinIO and systemd after installation; backups and counts |
Using LLMOJ
| Page | Contents |
|---|---|
| Contest rules | Scoring and feedback in NOI / OI, IOI, ICPC and Strict IOI contests |
| Teams | Private teams: joining by ID, reviews, groups and shared resources |
| Solutions | Solutions on problem pages, My discussions and AI solutions |
| Submissions | Submission previews, test data versions and earlier results |
| Hacks | Hacking submissions, the problem STD and hack results |
| Self Eval | Hidden pretests in NOI / OI contests and their quota |
| Fake moderation | Voiding cheating submissions while keeping the evidence |
| Registration approval | Registration settings and delegated reviews |
| Two-factor sign-in | Authenticator apps, passkeys, recovery codes and confirming sensitive changes |
The member list (/u) opens from View all in the Top Users panel on the home page, for users with the ViewUsers privilege.
Judging
| Page | Contents |
|---|---|
| Distributed judging | Architecture, the judge queue, draining and revoking judge machines |
| Remote judges | Setting up a separate judge machine and checking it |
| Judge configuration | Judge settings, the sandbox and its self-test, capacity and resizing |
AI
| Page | Contents |
|---|---|
| AI | Each user's models and search services, permissions, AI jobs, and the site's AI pause and limits |
| AI environment | AI directories, the local sandbox connection, keys and environment variables |
Reference
| Page | Contents |
|---|---|
| Configuration overview | Where each setting lives |
| Backend configuration | Backend settings: accounts, security, database, mail, storage, contests and display |
| Frontend configuration | Building the frontend locally; optional CDN settings |
| Environment | Environment variables for running, installing and building; maintenance scripts |
| Publishing | Applying updates with GitHub Desktop, publishing this wiki, and releases |
A full installation (--role all) runs the website, storage and a local judge, and can use remote judge machines too. A website-only installation (--role web) builds no local judging environment; remote judges handle submissions, but AI validation and test data generation need the local environment of a full installation.
LLMOJ is tested on Ubuntu 24.04 amd64. Its own code is under the MIT license; upstream components keep their licenses.