Optimized Prompts
π¨βπΌ Our users are complaining that when they use the
suggest_tags prompt, it's
taking a while to run because it has to run a few tools to get the data it
needs.Instead, we can just include the data in the prompt!
So replace the places where we instruct the LLM to run tools to acquire the data
it needs with references to the data itself. Effectively, we'll have a prompt
with multiple messages:
- Instructions (please suggest tags)
- The entry data
- The existing tags
You can use embedded resources for the data. For example:
const prompt = {
messages: [
{
role: 'user',
content: {
type: 'text',
text: '{Instructions}',
},
},
{
role: 'user',
content: {
type: 'resource',
resource: {
uri: '{uri}',
mimeType: 'application/json',
text: JSON.stringify(data),
},
},
},
// etc...
],
}