Search lecture files with AI
Index authorised course files locally, retrieve relevant evidence and check answers against precise source references.
Search lecture files with AI when an answer is scattered across slides, documents and notes. The classcorpus skill builds persistent course memory and retrieves relevant passages for your agent. References let you return to the underlying file and inspect the evidence instead of accepting a fluent answer without a traceable basis.
It supports PDF, PowerPoint, Word, text and selected transcript formats. Visual inspection and OCR are optional: extracted words alone cannot explain every diagram. Generated records stay in local user storage, and the package has no built-in hosted model service. Your chosen agent may still send selected text or images to its provider, so local indexing does not make the entire conversation offline.
What it does
- Search within a selected course
- Retrieve evidence with source references
- Compare explanations across files
- Identify slides needing visual inspection
- Build study guides from retrieved material
- Track local flashcard review progress
Install and use
- Install the skill and ClassCorpus package for a file-capable agent with Python3.11 or newer. Optional OCR, PDF and embedding features need their documented dependencies.
- Authorise a course folder and ask the agent to check its environment before indexing.
- Ask: “Find this concept in this course and answer with source locations; mark anything unsupported.”
- Inspect the citations and request visual review or local OCR when relevant evidence is missing.
FAQ
Source
Made by Jackson Wu. Original skill: classcorpus, published on GitHub under the Apache-2.0 licence.
View on GitHub