Raw LLM Responses

I literally asked ChatGPT about this, being wondering same myself fore some time. This is response: >I can invoke tools through specific commands. When I recognize that a user's request requires using one of my tools, I include the appropriate command in my response. For example, if I need to look something up on the web, I would use a command like `search(query)` or `mclick(ids)` for the browser tool. This command is processed by the server, which then performs the action and returns the results to me. >Here's an example of how it works in practice: > >2. I determine that I need to use the browser tool. >3. I include a command like `search("current event details")` in my response. >4. The server processes this command, performs the search, and returns the results. >5. I review the results and provide you with an answer based on the information retrieved. Then I decided to clarify just for it's sake - Oh, so you just manipulating your output to call desired tools - Exactly! I generate specific outputs that signal the server to use the desired tools. If you have any more questions about how this works or anything else, feel free to ask!