Polito Notes

Polito Notes Pipeline

Convert a PDF into two markdown files: Italian (notes.md) and English (notes-en.md), placed in the correct course folder.

Repository Structure

~/Documenti/github/polito/
├── first-year/
│   ├── first-semester//notes/
│   └── second-semester//notes/
├── second-year/
│   └── first-semester//notes/

Existing courses (check with ls — new ones may exist):

Input

The user provides: 1. A PDF file (via file send or local path) 2. The course name (in Italian or English) 3. Optionally: lecture number and/or date

Output

Two files in /notes//:

Folder Naming

N-titolo-in-kebab-case — N is the lecture number (provided by user or inferred from content/prior folders).

Template

Both files follow this exact structure:

# [Title]
> Course: Course Name  
> Lecture: N  
> Date: YYYY-MM-DD  
> Source: original-filename.pdf
Overview
[2-3 sentences: what this lecture covers and how it connects to previous topics]
Content
[Section 1 — Name]
[Fluid narrative prose. NOT bullet-list dumps. Explain each concept with
connecting logic, context, and motivation before diving into details.]
[Formulas in LaTeX inline/block, each symbol explained verbally right after.]
[When a comparison, taxonomy, or parameter set appears → table]
[When a process, pipeline, architecture, or relationship between concepts
appears → Mermaid diagram]
[Section 2 — Name]
...
Key Concepts
| Concept | Definition | Formula / Note |
|---------|-----------|----------------|
| ...     | ...       | ...            |
[Only at the end, as quick-reference. Does NOT replace the full explanations.]
Connections
[Links to other lectures or courses when relevant.]

Rules

1. Zero information loss — every definition, formula, example, use case, caveat, and technical detail from the PDF must appear in the output. Nothing gets skipped. 2. Narrative flow — write in continuous prose, not mechanical lists. Bullet points only when genuinely natural (e.g., listing properties). Prefer "The key idea behind X is..." over "• X is...". 3. Visual reconstruction — every diagram, schema, or figure in the PDF becomes a Mermaid diagram when possible (flowcharts, sequences, hierarchies). For complex visual layouts (matrices, heatmaps, scatter plots, mathematical plots), use a descriptive paragraph explaining what the figure shows instead of forcing an inaccurate Mermaid diagram. 4. Formulas — LaTeX $$block$$ or $inline$. After each formula, explain every symbol in words. 5. Terminology — first occurrence of a technical term: bold. Standard English terms stay in English even in the Italian version (with Italian explanation on first use). 6. Update vs create — if the PDF covers topics already in an existing notes file, integrate the new content into the existing file. Don't duplicate. 7. Both files must be structurally identical — same sections, same tables, same diagrams. Only language differs. 8. Backup — before overwriting an existing notes file, copy it to .bak (e.g., notes.md.bak, notes-en.md.bak). 9. Sensitive data warning — PDFs may contain sensitive or personal data. All content from the PDF will be preserved verbatim in the generated markdown notes. Ensure notes are stored securely and not shared publicly.

Workflow

1. Extract text from PDF (use pdftotext or python script) 2. Identify the course and resolve the target folder 3. Check existing folders to determine lecture number 4. Generate notes.md (Italian) following the template 5. Generate notes-en.md (English) — same structure, natural academic English 6. Write both files to the target folder 7. Confirm to the user with: course, lecture number, folder path, and a brief summary of what was covered

RAG Integration

After generating notes, the new markdown files will be automatically picked up by the local-rag skill during the next indexing run (daily cron at 9 AM). No manual action needed.

To search across all notes immediately:

~/.local/share/local-rag/venv/bin/python ~/.openclaw/workspace/skills/lookupmark-local-rag/scripts/query.py "attention mechanism in transformers" --top-n 5

Example Output Structure

polito/second-year/first-semester/large-language-models/notes/
├── 01-introduction-to-llms/
│   ├── notes.md          # Italian
│   └── notes-en.md       # English
├── 02-transformer-architecture/
│   ├── notes.md
│   └── notes-en.md
└── ...

1. Zero information loss — every definition, formula, example, use case, caveat, and technical detail from the PDF must appear in the output. Nothing gets skipped. 2. Narrative flow — write in continuous prose, not mechanical lists. Bullet points only when genuinely natural (e.g., listing properties). Prefer "The key idea behind X is..." over "• X is...". 3. Visual reconstruction — every diagram, schema, or figure in the PDF becomes a Mermaid diagram when possible (flowcharts, sequences, hierarchies). For complex visual layouts (matrices, heatmaps, scatter plots, mathematical plots), use a descriptive paragraph explaining what the figure shows instead of forcing an inaccurate Mermaid diagram. 4. Formulas — LaTeX $$block$$ or $inline$. After each formula, explain every symbol in words. 5. Terminology — first occurrence of a technical term: bold. Standard English terms stay in English even in the Italian version (with Italian explanation on first use). 6. Update vs create — if the PDF covers topics already in an existing notes file, integrate the new content into the existing file. Don't duplicate. 7. Both files must be structurally identical — same sections, same tables, same diagrams. Only language differs. 8. Backup — before overwriting an existing notes file, copy it to .bak (e.g., notes.md.bak, notes-en.md.bak). 9. Sensitive data warning — PDFs may contain sensitive or personal data. All content from the PDF will be preserved verbatim in the generated markdown notes. Ensure notes are stored securely and not shared publicly.