Prompt Flags
# Connection Model Flags
In the connections menu, you'll find five connection flags (-3, -4, -c, -b, -k). These flags can be used in chat or during the GPT step of a custom command, allowing the choice between Claude and ChatGPT for executing specific steps.
You can use flags, such as "-4", "-3", and "-k", to switch between models within a single custom command. For example:
- Adding the "-3" flag to the first GPT step will generate a response using GPT-3.5.
- Adding the "-4" flag to the second GPT step will generate a response using GPT-4.
This allows for convenient switching between faster models like GPT-3.5 and the slower, but more advanced, GPT-4.
If these flags seem inconvenient, you can change them:
- Go to the connections menu.
- Select the desired connection and click on the flag icon in the top right corner.
- Type a new flag name that you find easier to remember or use.
# Page-aware mode flag
Page-aware (-p) flag will send opened page content along with your prompt.
Pressing [🔍 PAGE] button, asking "-p What is this PDF about?", and "What is this PDF about? PDF: {{page}}" all do effectively the same thing, and give the same input to the LLM.
# Web-aware mode flag
Web-aware (-w) flag will search the web for the given prompt, and send search results along with your prompt.
This is useful for getting information from the web on the recent events.
# Isolated mode flag
Using -i flag on API or CloudGPT connection will force AI to ignore the chat history, sending prompt in isolation.
Isolated flag is used to reduce token consumption, especially on multi-chunk GPT steps. Responses of the isolated requests will still appear in the chat history, and can be used or referenced with {{gpt}} parameter.
# Combining Flags
To combine two flags, write them in sequence after a single "-". For instance, to combine the page-aware and GPT-4 flags, enter '-p4' in chat.
# Links for further reading
NEXT POST
All rights reserved © HARPA AI TECHNOLOGIES LLC, 2021 — 2024
Designed and engineered in Finland 🇫🇮
