What is a prompt?

Prompts are a specific set of inputs provided by you, the user, that guide LLMs on Amazon Bedrock to generate an appropriate response or output for a given task or instruction.

User Prompt:
Who invented the airplane?

When queried by this prompt, Titan provides an output:

Output:
The Wright brothers, Orville and Wilbur Wright are widely credited 
with inventing and manufacturing the world's first successful airplane.

(Source of prompt: AWS, model used: Amazon Titan Text)

Components of a prompt

A single prompt includes several components, such as the task or instruction you want the LLMs to perform, the context of the task (for example, a description of the relevant domain), demonstration examples, and the input text that you want LLMs on Amazon Bedrock to use in its response. Depending on your use case, the availability of the data, and the task, your prompt should combine one or more of these components.

Consider this example prompt asking Titan to summarize a review:

User Prompt:
The following is text from a restaurant review:

“I finally got to check out Alessandro’s Brilliant Pizza and it is now 
one of my favorite restaurants in Seattle. The dining room has a beautiful view 
over the Puget Sound but it was surprisingly not crowded. I ordered 
the fried castelvetrano olives, a spicy Neapolitan-style pizza 
and a gnocchi dish. The olives were absolutely decadent, and the pizza came 
with a smoked mozzarella, which was delicious. The gnocchi was fresh and wonderful. 
The waitstaff were attentive, and overall the experience was lovely. 
I hope to return soon." 

Summarize the above restaurant review in one sentence.

(Source of prompt: AWS)

Based on this prompt, Titan responds with a succinct one-line summary of the restaurant review. The review mentions key facts and conveys the main points, as desired.

Output: 
Alessandro's Brilliant Pizza is a fantastic restaurant in Seattle 
with a beautiful view over Puget Sound, decadent and delicious food, 
and excellent service.

(Model used: Amazon Titan Text)

The instruction Summarize the above restaurant review in one sentence and the review text I finally got to check out ... were both necessary for this type of output. Without either one, the model would not have enough information to produce a sensible summary. The instruction tells the LLM what to do, and the text is the input on which the LLM operates. The context (The following is text from a restaurant review) provides additional information and keywords that guide the model to use the input when formulating its output.

In the example below, the text Context: Climate change threatens people with increased flooding ... is the input which the LLM can use to perform the task of answering the question Question: What organization calls climate change the greatest threat to global health in the 21st century?”.

User prompt:
Context: Climate change threatens people with increased flooding, 
extreme heat, increased food and water scarcity, more disease, and economic loss. 
Human migration and conflict can also be a result. The World Health Organization (WHO) 
calls climate change the greatest threat to global health in the 21st century. 
Adapting to climate change through efforts like flood control measures 
or drought-resistant crops partially reduces climate change risks, 
although some limits to adaptation have already been reached. 
Poorer communities are responsible for a small share of global emissions, 
yet have the least ability to adapt and are most vulnerable to climate change. 
The expense, time required, and limits of adaptation mean its success hinge 
on limiting global warming.

Question: What organization calls climate change the greatest threat 
to global health in the 21st century?

(Source of prompt: https://en.wikipedia.org/wiki/Climate_change)

AI21 Labs Jurassic responses with the correct name of the organization according to the context provided in the prompt.

Output:
The World Health Organization (WHO) calls climate change 
the greatest threat to global health in the 21st century.

(Model used: AI21 Labs Jurassic-2 Ultra v1)

Few-shot prompting vs. zero-shot prompting

It is sometimes useful to provide a few examples to help LLMs better calibrate their output to meet your expectations, also known as few-shot prompting or in-context learning, where a shot corresponds to a paired example input and the desired output. To illustrate, first here is an example of a zero-shot sentiment classification prompt where no example input-output pair is provided in the prompt text:

User prompt:
Tell me the sentiment of the following headline and categorize it 
as either positive, negative or neutral: 
New airline between Seattle and San Francisco offers a great opportunity 
for both passengers and investors.

(Source of prompt: AWS)

Output:
Positive

(Model used: Amazon Titan Text)

Here is the few-shot version of a sentiment classification prompt:

User prompt: 
Tell me the sentiment of the following headline and categorize it 
as either positive, negative or neutral. Here are some examples:

Research firm fends off allegations of impropriety over new technology.
Answer: Negative

Offshore windfarms continue to thrive as vocal minority in opposition dwindles.
Answer: Positive

Manufacturing plant is the latest target in investigation by state officials.
Answer:

(Source of prompt: AWS)

Output: 
Negative

(Model used: Amazon Titan Text)

The following example uses Anthropic Claude models. When using Anthropic Claude models, it’s a good practice to use <example></example> tags to include demonstration examples. We also recommend using different delimiters such as H: and A: in the examples to avoid confusion with the delimiters Human: and Assistant: for the whole prompt. Notice that for the last few-shot example, the final A: is left off in favor of Assistant:, prompting Anthropic Claude to generate the answer instead.

User prompt: 


Human: Please classify the given email as "Personal" or "Commercial" related emails.
Here are some examples.

<example>
H: Hi Tom, it's been long time since we met last time. We plan to have a party at my house this weekend. Will you be able to come over?
A: Personal
</example>

<example>
H: Hi Tom, we have a special offer for you. For a limited time, our customers can save up to 35% of their total expense when you make reservations within two days. Book now and save money!
A: Commercial
</example>

H: Hi Tom, Have you heard that we have launched all-new set of products. Order now, you will save $100 for the new products. Please check our website.

Assistant:

Output: 
Commercial

(Source of prompt: AWS, model used: Anthropic Claude)

Prompt template

A prompt template specifies the formatting of the prompt with exchangeable content in it. Prompt templates are “recipes” for using LLMs for different use cases such as classification, summarization, question answering, and more. A prompt template may include instructions, few-shot examples, and specific context and questions appropriate for a given use case. The following example is a template that you can use to perform few-shot sentiment classification using Amazon Bedrock text models:

Prompt template: 
"""Tell me the sentiment of the following 
{{Text Type, e.g., “restaurant review”}} and categorize it 
as either {{Sentiment A}} or {{Sentiment B}}. 
Here are some examples:

Text: {{Example Input 1}} 
Answer: {{Sentiment A}}

Text: {{Example Input 2}}
Answer: {{Sentiment B}}

Text: {{Input}}
Answer:"""

Users can copy and paste this template, filling in their own text and few-shot examples to complete prompts when using LLMs on Amazon Bedrock. More templates are provided in this section.

Note

The double curly braces {{ and }} mark the places to put in data-specific information in the template, and should not be included in the prompt text.

Maintain recall over Amazon Bedrock inference requests

If you access Amazon Bedrock models by using API calls, be aware that models don't recall prior prompts and previous requests, unless the previous interaction is included within the current prompt. Including previous prompts in a request lets you have conversational-style interactions or make follow up requests.

If you are accessing Anthropic Claude models using API calls instead of the Amazon Bedrock text playground, you wrap your prompts in a conversational style to get desired response (for the Titan Text and AI21 Labs Jurassic model family, it's not necessary). The following example demonstrates a conversational-style prompt:

User prompt: 
	
	
Human: Describe where sea otters live in two sentences.

Assistant:

Output: 
Here is a two sentence description of where sea otters live:

Sea otters live in the northern Pacific Ocean along the coasts of 
Russia, Alaska, Canada, and the continental United States. 
They prefer coastal habitats like rocky shores, kelp forests, 
and estuaries where they have access to shallow, nearshore waters.

(Source of prompt: AWS, model used: Anthropic Claude)

Notice how the main content of the prompt is wrapped like this: \n\nHuman: {{Main Content}}\n\nAssistant:. For Anthropic Claude models, prompts sent via the API must contain \n\nHuman: and \n\nAssistant:.

To use conversational mode on Titan, you can use the format of User: {{}} \n Bot: when prompting the model.