Overview
The AI Assistant provides a chat-completion interface powered by an AI language model (OpenAI or compatible service). You type a prompt, the model generates a response, and the result appears in the output panel. The assistant does not automatically apply changes to the document — it produces text output that you review and act on manually.
Typical uses include drafting bookmark naming conventions, generating regex patterns for text matching, summarising document content, brainstorming TOC structures, or asking questions about PDF workflows.
Basic usage
- Open the AI Assistant from Plug-Ins > AI > AI Assistant....
- Type your prompt in the chat input field at the bottom of the dialog.
- Press Enter or click the send button to submit the prompt.
- Read the AI response in the output panel above the input field.
- Continue the conversation with follow-up prompts, or start a new topic.
Using selected PDF text
Click the Selection button to pre-fill the prompt field with text currently selected in the active PDF. This is useful for asking the AI to analyse, summarise, or transform a specific passage without re-typing it. You can then edit the pre-filled text or add instructions before submitting.
Saved prompts
Frequently used prompts can be saved for quick reuse, avoiding the need to retype complex instructions:
| Action | Description |
|---|---|
| Save Prompt | Stores the current prompt text for later reuse. |
| Load Prompt | Select a previously saved prompt from the dropdown to load it into the input field. |
| Remove Prompt | Deletes the currently selected saved prompt from the list. |
Conversation management
A conversation consists of the full exchange of prompts and responses in a session. Conversations can be persisted to disk and reopened later:
| Action | Description |
|---|---|
| Save Conversation | Saves the current conversation (all prompts and responses) to a file on disk. |
| Open Conversation | Loads a previously saved conversation, restoring the full history in the dialog. |
| Manage Conversations | Browse, review, and delete saved conversation files. |
Model parameters
Advanced parameters control how the AI model generates text. These are optional — the defaults work well for most use cases. Adjust them only if you need to tune the AI’s behaviour.
| Parameter | Effect |
|---|---|
| Temperature | Controls randomness. Lower values (e.g. 0.2) produce more deterministic, focused output. Higher values (e.g. 0.8) produce more varied, creative responses. |
| Max length | Maximum number of tokens (roughly words) the model may generate in a single response. Increase this if responses are being truncated. |
| Frequency penalty | Reduces repetition of the same words or phrases. Higher values discourage the model from repeating itself. |
| Presence penalty | Encourages the model to introduce new topics rather than staying on existing ones. Useful for brainstorming. |
| Best of | Generates multiple candidate completions internally and returns the best one. Higher values improve quality at the cost of increased API usage. |
Privacy and data
Prompts and any included document text are sent to the configured AI service over the network. Be mindful of the following:
- Avoid submitting confidential or sensitive document content unless your AI endpoint is configured for secure, private use.
- The AI service may retain prompts and responses according to its own data policies. Review your provider’s terms of service.
- Saved conversations are stored locally on disk and are not transmitted unless you explicitly share them.
Troubleshooting
| Problem | Solution |
|---|---|
| No response from the AI | Confirm that network access is available and the AI service endpoint is reachable. Check firewall and proxy settings. |
| Error about API key | Verify that your API key is correctly configured in the AI settings. Keys are typically set in the plug-in’s preferences dialog. |
| Truncated output | Increase the Max length parameter and resubmit the prompt. The model stops generating once the token limit is reached. |
| Slow or timed-out responses | Long prompts or high Best of values increase processing time. Simplify the prompt or reduce advanced parameters. |
| Unexpected or low-quality answers | Refine your prompt with more specific instructions. Lowering the temperature can make responses more precise and less random. |
Tip
Use the AI Assistant to generate regex patterns for the Regular Expression Library. Describe the text you want to match in plain language, and the AI will suggest a pattern you can save and reuse.
Warning
The AI Assistant does not modify your PDF directly. Always review AI-generated content before applying it to your document. AI responses may contain errors or hallucinations.