GPTLocalhost v1.0.47 makes AI for Word even more powerful with new Cross-Doc features, allowing you to work across multiple Word documents more efficiently. This release also includes bug fixes, performance improvements, and enhancements to the local Word Add-in experience.
- Add Select Your Model function to support seamless switching between local and cloud AI inference
-
Select Your Model lets you choose an AI model running either locally or in the cloud. At the bottom of the Add-in, click the model selector to choose your preferred inference provider.
-
For localhost, simply enter the port number to connect to your locally hosted AI model. If your device has a powerful GPU, you can select localhost to run inference locally for greater privacy and lower latency.
-
For a remote host, provide your API key to access the service. If your device has limited GPU resources, the interface makes it easy to switch between cloud providers such as Gemini, Mistral, or NVIDIA NIM, using either free-tier or paid inference services.
-
- Add Cross-Doc Generation function for generating content based on another open Word document
- Cross-Doc Generation lets you create text based on another open Word document. Simply select a peer document as the source and define the input range. In your current document, choose the output location and write a prompt, then click the Generate button. The Add-in retrieves content from the source document, combines it with your prompt, sends it to your AI model for inference, and inserts the generated result at your selected location.
- Add Cross-Doc Dialogue function for two documents to chat with each other automatically
- Cross-Doc Dialogue lets two documents automatically converse through multi-turn AI generation. Select a source document, a target document, and write prompts and AI models for each. Set the desired number of turns and watch as responses flow back and forth — the source document’s output becomes the target’s input, which then feeds back to the source, creating a continuous conversational loop that builds on both documents’ evolving content.