AI support can be used to check, edit and create texts as well as to create images and graphics.
In principle, all OpenAI-like interfaces (API) can be used. WritingTool particularly supports local solutions for AI, such as localAI (Installation instructions can be found here: https://writingtool.org/index.php/en/help/installation-of-a-local-ai).
Different servers and models can be entered for both cases. The input options are selected using the following two tabs:
- Text
- Images
- Text to speech (since WT 1.1)
The following applies to all cases:
- Use AI Support: If this is selected, the support is activated and the following information must be filled in.
Note: This must be a valid access; otherwise an error message will be triggered when accessing- AI Server URL: Complete URL for the Text, Image, or TTS API
- For the Text API, version 1 of the chat/completion interface from OpenAI is supported (the URL must end with chat/completions or chat. Any unknown URL endings will be interpreted as chat/completion). For example:
- Standard localAI: http://localhost:8080/v1/chat/completions/
- Standard Ollama: http://localhost:11434/api/chat/
(If the local AI is not running on the same machine, localhost is replaced with the IP address or server name on the local network.) - Commercial APIs can also be used if they comply with the OpenAI standard. However, these are usually fee-based, and validating large documents can be expensive.
- The Ollama Generate API is also supported as a text API (the URL must end with "generate"):
- Standard Ollama: http://localhost:11434/api/generate/
- For the Text API, version 1 of the chat/completion interface from OpenAI is supported (the URL must end with chat/completions or chat. Any unknown URL endings will be interpreted as chat/completion). For example:
- AI Server URL: Complete URL for the Text, Image, or TTS API
-
-
- As an image API, only the localAI API is currently supported:
- Example: http://localhost:8080/v1/images/generations/
- As a TTS API, only the localAI API is currently supported:
- Example: http://localhost:8080/tts/
- As an image API, only the localAI API is currently supported:
-
-
- AI Model: An AI model supported by the server (e.g. gpt-4)
- AI API Key: A valid API key if required.
Note: Such a key is often not used for local applications. Nevertheless, any non-empty character string must be entered here (from 26.4 it may also be empty).
For text, the following can also be entered:
- Enable Automatic error detection: In addition to the grammar check by LanguageTool, an automatic check is also carried out by the AI.Remember: With a paid server, a large volume of data can accumulate when checking extensive texts.
- Show only grammatical and orthographic hints: This is the default setting. At this stage of development, most hints can be trusted.
- Also show suggestions for small stylistic changes: Suggestions for stylistic changes are shown if they affect individual words.
- Also show extensive stylistic changes: This setting is still very experimental and leads to numerous, often undesirable, suggestions for changes in many models. It is therefore not recommended at the current stage of development.
- Show suggestion for stylistic rewriting of the paragraph in the sidebar: The AI will request a stylistic rewriting of the paragraph and display it in the sidebar. (Note: This will double the number of requests to the AI server.)
- AI detected grammar error: Here you can specify the color and style for grammatical errors (see also Grammar rules). The default is a blue wave.
- AI detected stylistic hints: Here you can specify the color and style for stylistic hints (see also Style rules). The default is a green wave.
