VideoQ documentation
VideoQ automatically answers student questions from lecture videos and links each answer to the scene that supports it. Course creators and instructors organize videos into courses and share them; students ask questions and jump to the cited scenes.
Start here to install VideoQ or make your first change. Choose an environment below; you do not need to read every design diagram before getting started.
Choose how to run VideoQ
| Goal | Start here |
|---|---|
| Evaluate on-prem on one server first | Single-server installation |
| Deploy to an existing Kubernetes cluster | On-prem and Helm guide |
| Run locally to change the application | Development environment |
Start an on-prem evaluation on one server. Test installation, video processing, Q&A, updates and recovery, then use capacity measurements to decide whether to separate services. Keep backups on another host during the evaluation too.
Understand the AI
Start with How AI builds an answer to follow one question through search, evidence, an answer, and a playable citation. It also explains what the AI reads and how much conversation context it receives.
Then explore transcription and scene search and Q&A prompts and answer persistence. The AI behavior menu groups these explanations together.
AI methods and references lists original papers, official model sources, implementation mappings, and reusable BibTeX citations.
The current pipeline uses independent questions, tool-based retrieval, and structured answers. Answer text and citations stream during generation; completed course answers are saved after validation. No post-answer AI scoring runs. The former Study mode and PLOG learning graph have been removed. The AI overview reflects the implementation checked on 2026-10-31.
New to the project?
Follow these steps to connect the user experience with the implementation.
| Step | Page | Ready to move on when… |
|---|---|---|
| 1 | Run the development environment | You can log in locally |
| 2 | Add a video and ask questions | You can follow an answer's citation to a scene in the video |
| 3 | Find your way around the code | You know where the UI, API, and video processing live |
| 4 | Make your first change | You can submit a small change with verification results for review |
For documentation-only changes, start with Update the documentation.
What to know first
- A video is one learning resource; a course groups videos for questions and sharing. See the data model
- React handles the UI, Hono handles the API, and Python handles time-consuming video processing. See the architecture
- Transcription and search preparation happen in sequence. Not every feature is ready immediately after upload. Understand the states
Find the guide for your task
| What you want to do | Guide |
|---|---|
| Connect Google Classroom | Classroom add-on integration |
| Connect Moodle or another LMS | LTI integration |
| Change screens, copy, or forms | Change the frontend |
| Add data or operations to the API | Change the API |
| Change tables or columns | Change the database |
| Change transcription or indexing | Change asynchronous video processing |
| Verify a change | Tests and verification commands |
| Troubleshoot startup or video processing | Troubleshooting |
| Deploy to an existing Kubernetes cluster | On-prem installation |
| Keep LLM, embeddings and Whisper on site | Local AI inference |
Explore the details
Use Design reference in the sidebar when you need a diagram or specification. Look up database names in the data dictionary and abbreviations in the glossary.