Here's a technical video that covers the code and functionality of a basic AI researcher. Full source code is available at the link below.

It takes a user's research topic, which can include formatting instructions, and performs research that's a bit deeper (using AI generated Google searches for more sources) than ChatGPT or Copilot Chat when they access the internet.

It does this by taking the user's request and generating Google search phrases designed to return pages related to the topic. Then it searches Google using a SERP Search Service API. Next the Python requests library is used to pull down page sources. Finally, the sources are combined into a unified LLM API call to fulfill the research request.

Thid video covers:

• Generating Google search phrases from a topic.
• Calling a SERP Search Result service to get related Google searches.
• Using Python's request library to pull down page source.
• Summarization of web contents tailored to the original request.
• Detailed Context Window discussion.
• Comparison's to ChatGPT and Copilot chat.

Links

• Code
• SERP: Google Search API - ValueSERP
• Prompting so that AI doesn't sound like AI.

The next video will address the context length limts and improve the quality of the output.