KlarText

KlarText turns dense German or English text into easy-to-understand language and can read it aloud. It’s built for people who get overwhelmed by long sentences, legal/bureaucratic phrasing, or technical language.

Important: KlarText produces “easy language” / plain-language simplifications. It is not certified “Leichte Sprache” and does not guarantee legal/medical accuracy.

🎯 Current Status

Phase 0: ✅ Complete (January 10, 2026)

Component	Status	Details
API Core	✅ Working	/v1/simplify endpoint with Groq LLM
PDF Extraction	✅ Working	/v1/ingest/pdf with PyMuPDF
Gradio Demo	✅ Working	Direct LLM & Via API modes
Prompt Templates	✅ Working	German & English from prompts/templates/
Quality Scoring	✅ Working	LIX, sentence length, readability
Swagger Docs	✅ Available	http://localhost:8000/docs
Batch Endpoint	📝 Defined	Logic pending
TTS	📝 Defined	Logic pending
URL Extraction	📝 Defined	Logic pending
Frontend (Next.js)	⏳ Planned	Will replace Gradio

Quick Start:

# First time setup:
cd services/api
cp env.example .env
# Edit .env and add your GROQ_API_KEY

# Terminal 1: API
cd services/api && uvicorn app.main:app --reload --port 8000

# Terminal 2: Demo
cd demo && python app.py

# Open: http://localhost:7860

See docs/phase_0_testing_guide.md for detailed setup instructions.

What it does (MVP)

• Paste text → get an easy-to-read version (DE/EN)
• Upload a PDF → extract text → simplify it
• Optional Text-to-Speech (TTS) for the simplified text
• No login required (MVP)

Stretch (optional):

• Paste a URL → extract main article text → simplify
• Chrome extension: simplify selected text or the current page

Architecture

flowchart LR
  U[User] -->|Paste text<br/>Upload PDF<br/>Enter URL| WEB[Web UI<br/>Next.js]
  WEB -->|REST| API[API<br/>FastAPI]
  API --> INGEST[Ingestion]
  INGEST --> CHUNK[Chunking]
  CHUNK --> LLM[LLM Adapter]
  LLM --> SIMP[Simplified Text]
  SIMP -->|optional| TTS[TTS Provider]
  API --> WEB
  TTS --> API

Loading

Components:

• Ingestion — Extract text from PDF, pasted text, or URL
• Chunking — Split long texts, ensure no hallucinations
• LLM Adapter — Configurable provider (OpenAI, Azure, Groq, Google)
• TTS Provider — Text-to-speech for audio output

Tech stack (recommended)

• Frontend: Next.js (React), TypeScript, Tailwind (or vanilla CSS), accessible UI patterns
• Backend: Python + FastAPI + Uvicorn
• LLM: configurable provider via an adapter layer
• PDF extraction: PyMuPDF or pdfplumber
• URL extraction (stretch): trafilatura or readability-lxml
• Rate limiting / caching (optional): Redis
• Storage (optional): Postgres (metrics/logging) + local file storage in dev

API overview

Base path: /v1

• POST /v1/simplify

  • Input: { text, target_lang: "de"|"en", level: "very_easy"|"easy"|"medium" }
  • Output: { simplified_text, warnings?: string[] }

• POST /v1/ingest/pdf

  • Input: multipart/form-data with file
  • Output: { extracted_text, pages, warnings?: string[] }

• POST /v1/ingest/url (stretch)

  • Input: { url }
  • Output: { extracted_text, title?, warnings?: string[] }

• POST /v1/tts

  • Input: { text, lang: "de"|"en" }
  • Output: { audio_url | audio_base64 }

• GET /healthz

Swagger / OpenAPI docs

Swagger UI: http://localhost:8000/docs (when API is running)
ReDoc: http://localhost:8000/redoc
OpenAPI JSON: http://localhost:8000/openapi.json

Quick start (local dev)

1) Prereqs

• Docker + Docker Compose

2) Configure env

Copy the example env file:

cp .env.example .env

Then set your LLM/TTS provider keys (see comments inside .env.example).

3) Run

docker compose up --build

4) Open

• Web: http://localhost:3000
• API: http://localhost:8000
• API docs: http://localhost:8000/docs

Repo layout

• apps/web – Web UI
• services/api – API service
• apps/extension – Chrome extension (optional track)
• prompts – prompt templates + eval fixtures
• docs – architecture notes + screenshots

Design reference

Lovable prototype: https://lovable.dev/projects/7d82e6e7-3919-4389-9b79-30543806c5e0

License

Usage & Licensing Notice (Non-Commercial) This project, including all code, models, datasets, documentation, and related materials, is provided for personal, educational, and non-commercial research use only.

Allowed Personal use Academic and non-commercial research Forking and modifying for non-commercial purposes Contributing via pull requests Not Allowed Any commercial use Selling or monetizing the software, models, datasets, or outputs Using the project in products or services that generate revenue Use within commercial organizations without permission Attribution If you share or publish derivative work, you must provide credit to:

Based on work from the KlarText Team (2025).

Commercial Licensing To request commercial usage rights, contact a repository administrator.