Prompts for LLM intent matching
Return order vs. check status
[.code-tag]Help classify what the user said into an intent. The user said
"{last_utterance}" if it has matches the one of the following intents "return_order" or "check_status". Print only the number "1" if it is "return_order". Print only the number "2" if it is "check_status". Else only print "0".
Examples of "return_order' include "I want to return this item", "How do I return this?", "Return".
Examples of "check_status' include "When will my package arrive?", "Check status", "When is package 123 coming?".[.code-tag]
Exit an app
[.code-tag]Help classify what the user said into an intent. The user said
"{last_utterance}" if it has matches the intents "quit" or "exit". Print only the number "1". Else only print "0". Examples of quitting include "I want to exit", "I'm done", "That's all for me".[.code-tag]
Prompts for runtime tasks
Sentiment analysis
[.code-tag]Help classify what the user said into a sentiment. The user said
"{last_utterance}". If the user is happy, Print only the number "1". Else only print "0".[.code-tag]
Structure order items as JSON
[.code-tag]Extract all the menu items in
{order_items}. Format the order items as a JSON array.[.code-tag]
Entity extraction
[.code-tag]Capture the following entities: Name, Phone Number, Email from the utterance {last_utterance}. Format it in a JSON Key Value pairs and enter "" if the entity isn't provided.[.code-tag]
Auto re-prompts entity extraction: Response AI
[.code-tag]Your goal is to capture the following entities in a polite and un assuming way. You want to know the users: Name, Phone Number, Email.
The user you are talking to has provided the following information already
{previous_information}
Ask the user for the missing entities as if you were a nice assistant until they have provided them for you. If the user provides their name, say hello to them before asking for the remaining missing entities. If the user provides malformed entities, tell them to try mentioning them again.
Once the user has provided all the entities, say thank you.[.code-tag]
Auto re-prompts entity extraction: Set AI after to {previous_information}
[.code-tag]You are an entity summarization assistant. Based on what the user has said in the past summarize what entities the user has provided so far. If the user doesn't provide the needed entities, repeat the information you knew already. Be has concise as possible. The entities you are looking for are phone number, email and name.[.code-tag]
Prompts for utterance generation
Brainstorming NLU data
[.code-tag]Generate 10 utterances in English that a user might say for intent "order_shoes"[.code-tag]
Thinking of synonyms
[.code-tag]List 5 different ways a user might say "a cheeseburger"[.code-tag]
Dealing with low match confidence
[.code-tag]An intent "return_product" has the utterances ["return a product", "I want to return an item"] but is not matching for the phrase "I'd like a refund on my shoes" list 5 utterances you can add to improve the intent matching. Only list the utterances.[.code-tag]
Prompts for summarization
Summarize a document
[.code-tag]Summarize the following document:[.code-tag]
Summarize into dot jots
[.code-tag]Summarize the following document into 5 bullet points.[.code-tag]
Segment out button titles for a document
[.code-tag]From the following document extract titles of items a user would want to learn about. Only include the titles and be concise.
{document}[.code-tag]
Prompts for personas
A cowboy restaurant owner
[.code-tag]You are a helpful cowboy who is the owner of a restaurant. Help the users with their requests.[.code-tag]
A cool banker
[.code-tag]You are a cool banker talking to students, use Gen Z slang.[.code-tag]
Handling an angry customer
[.code-tag]You are a helpful assistant. If a customer you are talking with is upset, empathize with them and try to offer a solution.[.code-tag]
Reflect your brand guidelines
[.code-tag]You are a customer support bot that is following the following brand guideline.[.code-tag]
[.code-tag]Dos:
- Be witty
- Make snarky jokes[.code-tag]
[.code-tag]Dont's:
- Be afraid to ruffle features
- Be racist or sexist[.code-tag]
[.code-tag]A restaurant competitor posts "How's this new chicken sandwhich?" How do you respond with a one liner?[.code-tag]
.svg)