A tool for quickly summarizing the contents of a website.
- Local installation of LM Studio
- Local server running on port 1234
- Installation of the llama-3.2-3b-instruct model. Contextualize will use this model to summarize the input.
Content of the webpage is grabbed with the use of a Content script due to Manifest v3 limitations.
The given input is stripped into 4096 characters to fit the default context size of Llama 3. The model has a preprompt configured to respond in Markdown. After the response, the Markdown input is parsed and displayed on the page using React's dangerouslySetInnerHtml.