Skip to content

TextTool

TextTool is a web-based application for LLM-supported text editing. It bundles common editing and analysis tasks into a set of curated tools with predefined prompts and complements them with custom instructions, an editing history, and multi-format export. Processing is performed via an OpenAI-compatible interface against an externally provided LLM endpoint.

At a glance

  • Correct, rephrase, summarise, or translate texts at the click of a button, each with task-specific instructions.
  • Rework administrative texts so they become understandable without prior knowledge of higher-education administration.
  • Convert texts to gender-inclusive language or to plain language ("Leichte Sprache").
  • Derive content analyses, balanced discussions of an idea, or task plans from a source text.
  • Refine results iteratively in multiple steps using "Apply" without losing the editing trail.
  • Export the result and the source text simultaneously in four formats (.txt, .md, .html, .docx).

Highlights

In contrast to a direct prompt to an LLM or a plain chat interface, TextTool combines curated instructions, tool-specific generation parameters, and a stateful editing model. The result is more consistent and more reproducible output for recurring text tasks.

  • Curated tools with fixed prompts — Each task (correction, gender-inclusive rewriting, plain administrative language, translation, and others) is backed by an elaborated prompt that specifies role, rules, and output format. Stylistic drift between sessions is reduced.
  • Direct and fast editing — Tools are triggered through a compact button bar; no prompt needs to be authored manually. Source text and result are placed side by side for immediate comparison.
  • Tool-specific generation parameters — Sensitive tools such as "Plain administrative language" use low temperature and restricted top-p to suppress hallucinations and unsolicited additions. More creative tools use different settings.
  • Optional modifier instruction — A free-text modifier can be applied across all tools (e.g. "in a more casual tone", "with examples") and persists until the field is cleared.
  • Custom command — For tasks outside the curated set, a free instruction field is available; the shared system prompt ensures a consistent output style.
  • Iterative editing with history — Results can be promoted to the new source text via "Apply"; recent editing steps remain available in a session-bound history and can be restored.
  • Connection to an LLM endpoint — Communication with an internally provided inference service through the OpenAI-compatible Chat Completions interface. Endpoint and model name are configurable and interchangeable.
  • Session-isolated multi-user operation — Editing state and exported files are kept per browser session in memory; concurrent users do not interfere with each other.
  • Multi-format export in a single step — Source text, result, or both are made available simultaneously as .txt, .md, .html, and .docx; generation is performed entirely in memory.