The 'Complete Chat' task in an advanced AI task that allows you to execute raw chat completions using Open AI (either using your own Open AI api key for paid users, or Flow XO Open AI credits for all users)
The available parameters are:
System Message - here you will introduce your bot, and tell it what kinds of tasks to expect to perform. You can also put any special rules here, like 'Always answer in rhyme'
User Message - the message or actual instruction from the user. This will usually come from user input, such as a question or the message from the trigger. Keep in mind that if you use the 'Recent chat history' feature described below, you may not need to include a User message at all. This is because the most recent message from the user will already be in the chat history.
Message history - this is an advanced field where you can simulate specific parts of a conversation. Each message in the conversation should be prefixed with 'assistant' or 'user'. You will usually use this setting to provide the AI with examples to help it with its task. For example, if your AI task is to do math problems and you want the AI to answer as briefly as possible, you may include a chat history like this:
user: What is 2 plus 2?
user: What is the square root of 9
Recent chat history - Here you can choose how many messages from the recent chat history to include in the context. These will be added AFTER any history you include in the Message history configuration setting. Use this if you are including this task in a loop, and you want the bot to be able to know about previous responses. IMPORTANT: If you use the 'Recent chat history' you probably should not include a User Message, because the question from the user will already be included in the history.
AI Connection - which AI connection to use. Paid users can use their own Open AI api key for improved pricing. All users can use the Flow XO connection and pay with Flow XO AI Credits.
Model - Which Open AI model to use. Which you use depends on your needs. ChatGPT can accept about 3000 words in the context, is the fastest and the cheapest (1 Flow XO Open AI credit). ChatGPT 16k can accept much more data (10-15k words), and is the second most expensive. GPT-4 is the most powerful, accepts 5-8k words, and is the slowest of the models.
Max Tokens - the maximum tokens to generate in the output.
Temperature - a number between 0 and 2 that controls how creative (random) you want the AI to be.
Here's a sample interaction with the configuration above:
And that's it! Happy flowing!