FlexGPT is like ChatGPT in the sense that it is best used in combination with good prompts. To fully exploit FlexGPT’s features, it’s helpful to understand how FlexGPT works. This page contains a quick guide, and a longer explanation.
If you’ve read this guide and are still stuck, feel free to use the support form or email [email protected] with your specific question and I’ll help you out as soon as possible.
🚀 Quick Tips
- do not leave search or memory enabled when you don’t need them
- it will cost more credit
- it could lead to irrelevant answers and hallucination
- for example, if you upload a book you have written to FlexGPT, and leave web search enabled when asking about it, FlexGPT will search the web and find irrelevant answer because your book doesn’t yet exist on the internet
- avoid using GPT-4
- GPT-4 is 45x more expensive than GPT-3.5
- GPT-3.5 can accomplish most tasks when using memory, web search, or a combination of both
- switch models/settings during the conversation
- if you only need the more expensive GPT-4 for one or two messages in the conversation, switch back to GPT-3.5 to save on the rest of the conversation!
- same goes for memory and web search functionalities, as these functions cost credit and token credit
🧠 Memory Guide
Okay, so you want to add something to FlexGPT’s memory and get a relevant answer from it.
- Upload the item to the memory. Let’s say we upload the entire text of ‘The Great Gatsby’.
- Enable memory using the checkbox in the left sidebar.
- Create a chat and send a message. You do not need to tell FlexGPT to use a particular piece of memory - it already has it. You just need to ask the question.
- For example, ask ‘how does Fitzgerald portray the American Dream in The Great Gatsby?’.
- FlexGPT will use the most relevant sections of the text in its response to give a relevant answer.
You do not need to:
- tell FlexGPT to use a document
- tell FlexGPT to use its memory
- tell FlexGPT to