Syllabus Parser

https://syllabus-parser-alpha.vercel.app/

Upload a syllabus PDF to automatically extract course information for students success

Tech Stack

• Frontend: React + Vite (TypeScript), utility CSS classes, Fetch API
• Backend: FastAPI (Python), Uvicorn, Pydantic
• AI: OpenAI Chat Completions (structured JSON extraction) with heuristic fallback
• PDF: Text extraction via a PDF reader utility
• Export: Client-side ICS (iCalendar) and CSV generation

What It Does (Features)

• PDF upload: Sends the syllabus to the backend for parsing
• Summary: Concise 2–4 sentence overview
• Events:
  • Extracts academic deadlines with ISO dates (YYYY-MM-DD)
  • Displays in a friendly format (“September 23rd, 2025”)
  • Export:
    • .ics for Google/Apple Calendar (all-day events)
    • .csv for Notion (Title, Date)
• Evaluations:
  • Extracts assessment components and their weights
  • Normalizes weights so total ≈ 100%
  • Renders a table with a computed total row and subtle warning if not exactly 100

How It Works (Flow)

1. User uploads a PDF in the frontend.
2. Backend extracts text (limits pages/characters for speed).
3. If OpenAI API key is set:
   • Sends text with strict JSON instructions to the model.
   • Parses summary, events, and evaluations from JSON.
4. If the model fails or no key:
   • Heuristic fallback scans lines for dates, exam/assignment keywords, and percent weights.
5. Evaluations are validated and normalized to sum to 100.
6. Frontend renders sections and offers export buttons for calendars/Notion.

Key Files (Frontend)

• src/App.tsx
  • Uploads PDFs, calls backend, displays Summary, Events, Evaluations
  • Wires export buttons (ICS/CSV)
• src/components/EventDate.tsx
  • Formats ISO dates to “Month DayOrdinal, Year”
• src/components/EvaluationsTable.tsx
  • Renders assessment/weight table with total
• src/utils/types.ts
  • EventItem, EvaluationItem, ParseResult types
• src/utils/exportCalendar.ts
  • Builds ICS and CSV on the client
  • ICS details:
    • All-day events: DTSTART;VALUE=DATE:YYYYMMDD
    • DTEND is next day (exclusive)
    • Unique UID per event
  • CSV details:
    • Columns: Title, Date (ISO)

Key Files (Backend)

• app/main.py
  • FastAPI app setup, middleware (e.g., CORS), route wiring
• app/services/syllabus_parser.py
  • System prompt, OpenAI call, heuristics, validation, normalization
• app/services/pdf_reader.py
  • Extracts text from PDFs (page/size limits)
• app/models.py
  • Pydantic models: Event, Assessment, ParseResult
• app/config.py
  • Loads OPENAI_API_KEY and other settings

Setup

Backend (Python)

• macOS Terminal:
  • cd backend
  • python3 -m venv .venv
  • source .venv/bin/activate
  • pip install -r requirements.txt
  • uvicorn app.main:app --reload

Frontend (Node)

• cd frontend
• npm install
• npm run dev

Usage

1. Start backend and frontend.
2. Open the frontend URL (localhost:5173).
3. Upload a syllabus PDF.
4. Review Summary, Events, and Evaluation Breakdown.
5. Export:
   • Download .ics to import into Google/Apple Calendar
   • Download .csv to import into a Notion database (map Date column)

Notes & Guarantees

• Event dates must be valid ISO; invalid dates are dropped.
• Evaluation weights are normalized to sum to ≈ 100 (largest item adjusted to fix rounding residue).
• ICS export creates all-day events for best compatibility.
• CSV export targets Notion import with minimal mapping.