Simple prompt
Simple prompt
An example of a simple prompt to jarvis looks like this:
curl --location 'https://api.paymentsos.com/hackathon-ai/prompt' \
--header 'private-key: replace-me' \
--header 'app-id: replace-me' \
--header 'Content-Type: application/json' \
--data '{
"question": "Can you please tell me what is the most pouplar programing language, please provide short answer",
"model": {
"stop_sequences": [],
"name": "anthropic.claude-v2",
"max_tokens": 1000,
"temperature": 1,
"top_k": 250,
"top_p": 0.999
}
}'
and this is the answer you will get:
{
"query": "Can you please tell me what is the most pouplar programing language, please provide short answer",
"answer": " Python is currently one of the most popular programming languages."
}
Now lets go over all the params we had in this http request
question - This is the prompt that will be sent to the FM
stop_sequences - A set of chars that will stop the response from the FM like \n for new line
* name - This is the model id from the - Bedrock supported models
max_tokens - The maximum tokens that will be in the answer - (1 token equal to 3/4 words)
temperature - Read here
top_k - Read here
top_p - Read here
Playing with the request body: different models and different params like temperature, top_k and top_p will give you different answers, play with them and find the right tool for the job.
As an example, lets ask FM from AI21 the same question
curl --location 'https://api.paymentsos.com/hackathon-ai/prompt' \
--header 'private-key: replace-me' \
--header 'app-id: replace-me' \
--header 'Content-Type: application/json' \
--data '{
"question": "Can you please tell me what is the most pouplar programing language, please provide short answer",
"model": {
"stop_sequences": [],
"name": "ai21.j2-mid",
"max_tokens": 1000,
"temperature": 1,
"top_k": 250,
"top_p": 0.999
}
}'
and this is the answer you will get:
{
"query": "Can you please tell me what is the most pouplar programing language, please provide short answer",
"answer": " According to recent surveys, JavaScript is currently the most popular programming language among developers."
}
As you can see different models trained and tuned on different sets of data can return various responses.