DeepTutor on Railway
Open-source AI tutor that answers from your own documents, with a protected login, on Railway.
DeepTutor is an open-source AI tutor from the Data Intelligence Lab at the University of Hong Kong. You upload textbooks and notes, ask questions and get answers with citations, generate quizzes, and follow guided learning paths.
It suits students, teachers and small learning groups who want a tutor grounded in their own material rather than a general chatbot. It needs a chat model provider and an embedding model, which you supply with your own API key.
DeepTutor is released under the Apache-2.0 licence.
What DeepTutor can do
- Question answering over uploaded documents with citations
- Quiz generation
- Guided learning paths
- Knowledge bases and notes
- Multiple user accounts managed by an admin
- Works with your own LLM provider key
What people use it for
- A personal study assistant built from your textbooks and notes
- A class or study group tutor with separate learner accounts
- Generating practice quizzes from course material
- Self-hosting a tutor so documents stay on your server
Videos
Running DeepTutor on Railway
On Railway, the DeepTutor template deploys the official image with the web app on the public URL and the API on loopback, a volume for settings, accounts, knowledge bases and notes, and Redis with append-only persistence. An admin account is created with a generated password and sign-up is closed. It runs as one replica because the data sits on one volume. Railway bills the CPU and memory used; model usage is billed by your LLM provider.
DeepTutor templates on Railway
Tracklayer templates are listed first. Other DeepTutor templates follow, ordered by total deployments on Railway. How this works.
Deployment numbers as of 2026-11-20.
Questions about DeepTutor
How do I log in?
Open your domain and sign in as admin with DEEPTUTOR_ADMIN_PASSWORD from the service Variables tab. Sign-up is closed; the admin creates learner accounts at /admin/users.
What do I need to provide?
A chat model provider key and an embedding model. Set LLM_API_KEY at deploy (defaults are openai and gpt-5-mini) or add a provider in Settings after login. For other providers, configure the embedding settings too.
Can it run code?
Not by default. DeepTutor's exec tool normally runs in a separate sandbox container, which Railway cannot host. Enabling DEEPTUTOR_ALLOW_CODE_EXEC runs code in the app container without isolation from its data and keys, so enable it only if you accept that.
How does it scale?
As one replica. Raise DEEPTUTOR_WORKERS for more parallel work; the template estimates about 0.5 GB RAM per worker but notes this is unmeasured. With one worker you can remove Redis.
How do I upgrade?
Back up the volume, change the image tag and digest in the Dockerfile and redeploy.
Similar apps
Tracklayer maintains the templates marked as such; DeepTutor is a trademark of its owners and this site is not affiliated with them.
