An AI engine that reads your courses — and cites them.
UniMind isn’t another chatbot wrapper. It’s built to understand your courses — from the syllabus to lecture transcripts, from assignment briefs to announcements. Pick a course and ask: it answers from that course’s materials and marks its sources.
University of Sydney
Evidence first, then the answer.
Many AI study tools ask you to paste your notes into a chat box. UniMind first chunks and indexes each course’s materials, recording which module and section every passage came from. When you ask, it gathers evidence for that kind of question before the model writes its answer.
Structure over prompts
Instead of stuffing everything into a prompt window, UniMind chunks and indexes 8 types of course material, keeping each passage’s place in the module, page and section — then pulls only the passages that matter.
Retrieval over memory
Questions about your course are answered from evidence retrieved from your own materials, not the model’s general memory — and when your materials don’t cover something, it says so.
Specialization over generalization
Three task routes — pinpoint Q&A, full-course overviews and planning — each gather evidence differently, so every kind of question gets the right mix of material.
8 types of course material, one search
UniMind chunks and indexes 8 types of material from your course — syllabus, pages, assignments, quizzes, announcements, discussions, files and lecture transcripts. Before answering, it retrieves the relevant passages from across the course, so its claims map back to a source instead of a guess from general knowledge.
- Syllabus
- Pages
- Assignments
- Quizzes
- Announcements
- Discussions
- Files
- Transcripts
A pinpoint question, an overview or a plan — all from your own courses
It first works out whether you want an answer, an overview or a plan, then decides where to look and how much to read. That happens behind the scenes — you just ask.
“When’s the midterm, and what’s on it?” UniMind retrieves the relevant passages from that course, ranks them on several signals, and gives you the answer with its sources. Pro and Max search by meaning, so it finds passages even when your wording differs; other plans use keyword search.
“Give me an overview of this course.” It samples the syllabus, modules, assignments, quizzes, announcements, discussions and pages separately, then writes a course overview, module summary or weekly digest — not just the first few passages it found.
“How should I revise in the two weeks before finals?” It first pulls the deadlines out of assignments, quizzes, announcements and modules, then lays out a day-by-day plan.
Citation sources under the answer are colored by type — syllabus, assignment, slides, announcement, lecture transcript. Click one and the original opens in the app, so you can check what matters.
- Citation sources listed under every answer that uses your materials
- Down to the syllabus, assignment, page, announcement or lecture moment
- Colored by material type
- One click opens the original
Three guardrails against making things up
UniMind puts constraints on both retrieval and answering:
- 1
Relevance threshold
Results below the relevance threshold are never handed to the model as evidence — better too little than padding with unrelated text.
- 2
Cite only what was found
Every search result is numbered, and answers may only cite those numbers. A source that isn’t in the results can’t appear.
- 3
Say so when it’s not there
When your course materials don’t cover the question, UniMind tells you it found nothing instead of filling the gap with general knowledge.
By design, when your materials don’t cover a question, UniMind says it found nothing instead of inventing a plausible answer. AI can still make mistakes, so open the source for anything important.
Semantic retrieval and incremental sync
UniMind represents course content as 1536-dimensional vectors with an HNSW approximate nearest-neighbor index, so it finds relevant passages by meaning — even when your wording differs from the source. Semantic search is included in Pro and Max; other plans use keyword search.
Smart chunking and incremental sync
Materials are split at headings and paragraphs into passages of roughly 300–800 tokens, with a little sentence overlap, and each passage records its module and section. When a course changes, only the changed parts are reprocessed. Canvas courses refresh every 6 hours while the app is open, and any course can be refreshed by hand at any time.
- Embedding
- 1536-dimensional high-density vector space (OpenAI text-embedding-3-small)
- Index
- HNSW approximate nearest neighbor (m=16, ef_construction=64)
- Sync
- Incremental content-fingerprint sync (SHA-256)
- Plans
- Semantic search on Pro and Max; keyword search on other plans
- Precision
- Multi-signal fusion ranking
- Scope
- One course or all courses (all-courses mode works from an overview of each course)
Live transcription × UniMind
While you record, the live transcript is UniMind’s context — ask “what did I just miss?” any time. Transcripts linked to a course join that course’s library, so at revision time they’re searched and cited alongside your slides and syllabus.
- Ask while you listen: During a recording UniMind reads the lecture’s live text — no waiting for class to end, no manual upload.
- Part of the course library: Transcripts linked to a course join its knowledge base, so they’re searchable and citable at revision time.
- English and Chinese: English lectures are translated into Chinese as they happen, and UniMind can answer questions about the lecture in English or Chinese.
The more lectures you transcribe with UniKnow, the more of what was said in class UniMind can cite.
What’s next
What’s shipped, and the directions we’re exploring. Exploring doesn’t mean there’s a release date.
- v1.3 · Shipped
Slide and document text search
Text in the PDF, Word and PowerPoint files from your LMS is extracted and indexed, so UniMind can quote your slides directly.
- Exploring
Cross-term course context
Exploring whether UniMind could, with your permission, draw on related courses you took earlier — say, Intro to Stats when you take Advanced Statistics.
- Exploring
Study-group sharing
Exploring opt-in sharing for study groups: only the notes and transcripts you choose to share would be visible to your group’s UniMind.
- Exploring
Priority hints
Exploring hints about what to tackle first this week, based on due dates and assignment weight.
- Exploring
Exam prep, ahead of time
Exploring review materials and practice questions prepared ahead of upcoming exams.
Try UniMind today.
Download UniKnow and let UniMind attend every class with you. Free to start, no credit card required.