A chrome extension which allows you to start a chat with GPT directly in a webpage with a click by highlighting relevant text and asking a question.
- Start subthreads with GPT (the namesake): Start multiple follow-up threads with GPT, focusing on different highlighted text while incorporating surrounding context.
- Summarize documents, focusing on a subtext (purple highlight) and taking surrounding text as context (gray highlight, which can be optionally displayed via configuration)
- Get additional context on terminology and concepts in articles
Some of the chat features include:
- Start a chat on (almost) any webpage
- Supports Open AI chat and reasoning models
- Uses highlighted text and surrounding context to ground the question
- Supports multiple concurrent chats on a page which persist as long as the page is loaded
- Customizable chat box placement and sizing
- Markdown and LaTeX formatting support
Additionally, your API key is stored locally on the Chrome browser and not shared.
- Download the Chrome extension
- Add your OpenAI API key in the extension settings
- Highlight text on any webpage
- Start chatting with GPT!
The original motivation for this project (and the reason for the name) was that current LLM chat apps don't natively support "threads" which would facilitate in-context follow-ups without jumping around chat history (like with quotes). With conversation threads, chats could be more similar a branching tree of exploring follow-ups rather than one straight-shot line to the answer. These can be particularly useful for complicated topics where one long GPT response prompts many follow-up questions. Given this, the original intention was to use this extension as a temporary bandaid in ChatGPT in hopes one day threads would be implemented. In practice, this can be a low friction way to start chats with GPT using context from news sites, documentation pages, or any site of your choice.
The extension can be customized through the extension popup menu:
- Model Version: Choose from various GPT models (e.g. gpt-4o, o3-mini)
- Surrounding Context: Control how many surrounding words around your highlight are included in the context (total, split roughly equal above and below).
- Max Response Words: Approx maximum words in the model response.
- Use Previous Box Placement: If enabled, keep customized chat chat box locations if chats are dragged. Applies to current tab.
- Use Previous Chat Box Size: If enabled, keep customized chat box sizes if chats are resized. Applies to current tab.
- Show Context highlights: Enable highlights to visualize the surrounding context being sent to the model
- Extension Toggle: Enable/disable the extension