This is Day#2 of the #*ServerlessChallenge **challenge, where I'm creating/building serverless apps using different AWS services.*
Short description: An API that accepts a list of food/ingredients that I have and returns meal ideas that I can prepare.
The long/tech description
Using Amazon Bedrock (Jurassic-2 Mid model) to generate some meal ideas and the steps for preparing it after getting the input of the available ingredients and the number of needed meals:
The prompt text is so simple but it's enough to make sure that the output will be as I want, also the max number of tokens is limited to 200 which is enough Don't want to be surprised by the invoice 🏃-
const command: InvokeModelCommand = new InvokeModelCommand({
modelId: process.env.BEDROCK_TEXT_MODEL,
contentType: "application/json",
accept: "*/*",
body: JSON.stringify(
{
prompt: `Recommend only ${numberOfMeals} meal(s) Which i can make only by using ${ingredients.join(',')}`,
maxTokens: 200,
temperature: 0.7,
topP: 1
}
)
});
The trigger of the lambda function is the endpoint /meals (API Gateway) with the following event schema which has the list of the ingredients & the needed number of meals:
default {
type: "object",
properties: {
ingredients: {type: 'string'},
count: {type: 'number'}
},
required: ['ingredients', 'count']
} as const;
events: [
{
http: {
method: 'get',
path: 'meals',
request: {
schemas: {
'application/json': schema,
},
},
},
},
]
So, to wrap that up, simple the function is:
Receiving a list of ingredients and the number of meals
Invoking the model to get some meal ideas
Responding with the ai response
GitHub repo: https://github.com/Rochdy/mealsRecommender
That was the app of Day#2! looking forward to posting soon about the next app
