lecture-to-hw

What problem does it solve?

Automates the end-to-end workflow of finishing course homework by scanning a messy course folder, extracting questions from many file formats, matching them to taught lecture methods and code, and producing concise, student-style Markdown answers that are ready to submit.

Core Features & Use Cases

• Multi-format homework ingestion: Reads common inputs including PDF, DOCX/DOC, Markdown/TXT, HTML, notebooks, images (with OCR/vision), and archives, and flags uncertain reads instead of inventing content.
• Lecture/code grounding: Searches lecture slides and classroom demo code using question keywords, prioritizes course terminology/formulas, and reuses runnable code/results when available.
• Style imitation from history: Learns formatting conventions (headings, formula/table/image styles, length/detail) from existing 作业/hw*_solution/*.md without copying the content.
• Single or parallel agent workflow: Acts as a controller to decompose tasks and optionally run up to 4 bounded subagents, then unifies edits to a single final Markdown.
• TA-style review and confidence reporting: Checks for missed grading points, formula/numeric/logic errors, method mismatch with lectures, and “AI smell,” then outputs a confidence level (high/medium/low) plus manual review suggestions.

Quick Start

Use lecture-to-hw to complete your current homework and output a submission-ready Markdown solution.