It is necessary to automate work in the browser specifically in the GPT chat, not through the API. I create stories using a Brief
I have a Brief (master prompt) and a file with topics for stories. I have outlined each step I take in the GPT chat, and I will attach a video if possible to show how everything happens. The main goal is batch processing capability: I input the required prompts and 10-50 topics for stories into text files, then I receive 10-50 ready stories in the end.
Example of my routine
https://drive.google.com/file/d/11pg0C80_cT9jnPoxS6Q2mwQIEn6gnaBg/view?usp=sharing
1) The work begins with creating a new chat and the first request for the BRIEF, the master prompt (one text file - one request - editable)
2) Then GPT asks to specify the topic for the story and other details: number of characters, style, genre, and more. Topics will be prepared in advance and should be taken sequentially from a file that can be edited. Other data are taken from a separate text file, which can also be edited — and all this should be one request: topic and details (number of characters, style, etc.) (both are 2 text files - 1 request - editable)
3) After that, GPT creates a story outline according to all data and asks for confirmation to continue: WRITE THE FIRST PART, this part should also be taken from a text file, as additional data needs to be entered here and edited. (one text file - one request - editable)
4) Then GPT writes the first section, and after that, after each subsequent section, you need to type CONTINUE. This request should also be taken from a text file for editing purposes. (one text file - one request - editable)
5) The story is saved, then the software proceeds to create the next story, creating a new branch in GPT and starting over again.
Details
1) The Brief creates 5-10 parts of the story plan, resulting in 5 to 10 requests specifically for writing the story. Thanks to the Brief, GPT creates an optimal number of requests (parts of the plan) for each story. I don’t know how to stop the software at the right moment; the number of parts can vary in each story. As an option, setting requests specifically for the plan, even if the plan has 5 requests, the software will make 10 requests and save everything received. In this case, the plan will be saved at the beginning of the story, and I will manually delete the unnecessary parts based on the number of plan parts.
2) Branches in GPT should not be deleted; a new story means a new branch.
3) Even if something goes wrong, perhaps due to a bug or other issue, a new branch is created, and the software attempts to recreate this story from scratch, starting with the BRIEF.
4) Each story request must be separated by 5-7 blank lines, possibly numbered.
Please provide your price. It is essential that I receive software with open source code for editing. If there are any changes in GPT, the software could become unusable, as you understand. Any suggestions for improvement are welcome.
