AI-Powered Thesis and Document Summarizer

A comprehensive full-stack application designed to extract text from documents, generate AI-powered summaries in various formats, manage an approval workflow, prioritize revisions, and generate text-to-speech audio.

Architecture & Tech Stack

• Backend: FastAPI (Python), SQLAlchemy, and PostgreSQL. Handles document parsing (.pdf, .docx, .txt), summarization logic, audio generation (OpenAI TTS API), and API provisioning.
• Frontend: React 19, TypeScript, and Vite. Provides an interactive dashboard with modular components (e.g. AudioList, DocumentList, SummaryList) and modal-driven UI for handling the document flows.
• Database: PostgreSQL (Dockerized). Tracks document files, summaries with varying statuses (pending, approved, rejected), and audio references.
• Semantic search: On upload, extracted text is chunked and embedded with Ollama (embeddinggemma); GET /search?q=... ranks documents by cosine similarity and the dashboard search bar links results to their summaries. See project-docs/backend.md.
• Testing: Playwright for end-to-end (E2E) testing (tests-e2e), Pytest for backend tests (backend/tests).

Project links

• Task board
• Wiki

Release v4.0.0 (Current)

• Implement features:
  • Summary comparison view #39
  • Access relevant summaries by clicking on the original document #30
  • Various UI refinements #58

Release v3.0.0

• Implement features:
  • Add audio generation and playback functionality using OpenAI TTS API #35
  • Implement audio management UI with player and controls #34
  • Implement audio summary download option #45

Release v2.1.0

• Implement features:
  • Implemented comprehensive summary approval workflow with status-based action buttons (Approve, Reject, Regenerate, Download, Edit, Delete) #25 #26 #27
  • Added multi-format summary generation with three format options: short_text, bullet_points, and academic_tone #23 #40
  • Enhanced UI/UX with improved table styling (alternating row colors, color-coded status indicators), delete functionality for documents, and format selection during upload #29
  • Updated database schema to track approval status with three states: pending, approved, rejected #31
  • Replaced full page reloads with callback-based data refresh mechanism for better user experience

Release v2.0.0

• Implement core features for Iteration 2:
  • Documents, Summaries, Audio sections available in dashboard #29
  • Document upload functionality #16 #17 #32
  • Document auto-extraction #18
  • Document table view with a search bar and sorting options #20
  • Summarization engine #22
  • Summary editing capability #26

Installation

Backend

1. cd backend
2. python -m venv .venv
3. source .venv/Scripts/activate
4. pip install -r requirements.txt on first-time use
5. Install Docker desktop and have it running, then run the following command. NB! replace user and password with your own postgres credentials.
   docker run -d --name postgres16 -e POSTGRES_DB=summarizer -e POSTGRES_USER=postgres -e POSTGRES_PASSWORD=password -p 5432:5432 postgres:16
6. Set DATABASE_URL (required, no default), e.g. in backend/.env: DATABASE_URL=postgresql://postgres:password@localhost:5432/summarizer
7. alembic upgrade head to create/upgrade the schema (run again after pulling new migrations)
8. uvicorn app.main:app --reload pip install pydantic-settings, if any errors → runs at http://localhost:8000 / http://127.0.0.1:8000 Swagger UI is available at http://localhost:8000/docs

Backend tests (need a reachable Postgres; the test database is dropped/recreated): pip install -r requirements-dev.txt && TEST_DATABASE_URL=postgresql://postgres:password@localhost:5432/summarizer_test pytest

Frontend

1. cd frontend
2. npm install on first-time use
3. npm run dev → runs at http://localhost:5173 / http://127.0.0.1:5173