 Voir sur ai.google.dev
|
 Exécuter dans Google Colab
|
 Afficher la source sur GitHub
|
Présentation
Ce guide de démarrage rapide explique comment utiliser l'API PaLM, qui vous donne accès aux derniers grands modèles de langage de Google, avec le SDK PaLM Node.js, en particulier pour les cas d'utilisation axés sur les dialogues, tels que les chatbots.
Obtenir une clé API
Pour commencer, vous devez obtenir une clé API. Définissez-la en tant que variable d'environnement:
import os
os.environ["API_KEY"] = "<YOUR API KEY>"
Installer le client API
Dans un nouveau répertoire, initialisez un projet Node.js à l'aide de npm et installez la bibliothèque google-auth:
npm init -ynpm install google-auth-library
Wrote to /content/package.json:
{
"name": "content",
"version": "1.0.0",
"main": "index.js",
"scripts": {
"test": "echo \"Error: no test specified\" && exit 1"
},
"keywords": [],
"author": "",
"license": "ISC",
"dependencies": {
"@google-ai/generativelanguage": "^1.0.1",
"google-auth-library": "^9.0.0"
},
"devDependencies": {},
"description": ""
}
+ google-auth-library@9.0.0
updated 1 package and audited 74 packages in 0.905s
3 packages are looking for funding
run `npm fund` for details
found 0 vulnerabilities
npm WARN content@1.0.0 No description
npm WARN content@1.0.0 No repository field.
Installez ensuite la bibliothèque cliente:
npm install @google-ai/generativelanguage
+ @google-ai/generativelanguage@1.0.1 updated 1 package and audited 74 packages in 2.274s 3 packages are looking for funding run `npm fund` for details found 0 vulnerabilities npm WARN content@1.0.0 No description npm WARN content@1.0.0 No repository field.
Obtenir une clé API
Suivez les instructions de la page de configuration afin de créer une clé API pour votre application. Vous en aurez besoin à l'étape suivante.
Générer des messages
Créez un fichier index.js et ajoutez le code suivant, en fournissant votre clé API PaLM via la variable d'environnement API_KEY:
%%writefile index.js
const { DiscussServiceClient } = require("@google-ai/generativelanguage");
const { GoogleAuth } = require("google-auth-library");
const MODEL_NAME = "models/chat-bison-001";
const API_KEY = process.env.API_KEY;
const client = new DiscussServiceClient({
authClient: new GoogleAuth().fromAPIKey(API_KEY),
});
async function main() {
const result = await client.generateMessage({
model: MODEL_NAME, // Required. The model to use to generate the result.
temperature: 0.5, // Optional. Value `0.0` always uses the highest-probability result.
candidateCount: 1, // Optional. The number of candidate results to generate.
prompt: {
// optional, preamble context to prime responses
context: "Respond to all questions with a rhyming poem.",
// Optional. Examples for further fine-tuning of responses.
examples: [
{
input: { content: "What is the capital of California?" },
output: {
content:
`If the capital of California is what you seek,
Sacramento is where you ought to peek.`,
},
},
],
// Required. Alternating prompt/response messages.
messages: [{ content: "How tall is the Eiffel Tower?" }],
},
});
console.log(result[0].candidates[0].content);
}
main();
Overwriting index.js
Exécutez ensuite le script suivant:
node index.js
The Eiffel Tower is 324 meters tall, It's a sight to behold, tall and not small. It's made of iron and weighs 10,100 tons, It's a wonder of the world, it's a must-see for all.
Continuez la conversation
Pour poursuivre une conversation après un premier message d'invite, vous devez ajouter le message de candidat renvoyé ainsi que l'invite suivante:
%%writefile index.js
const { DiscussServiceClient } = require("@google-ai/generativelanguage");
const { GoogleAuth } = require("google-auth-library");
const MODEL_NAME = "models/chat-bison-001";
const API_KEY = process.env.API_KEY;
const client = new DiscussServiceClient({
authClient: new GoogleAuth().fromAPIKey(API_KEY),
});
async function main() {
let first = "Tell me a one short animal fact."
let messages = [{ content: first }];
const result = await client.generateMessage({
model: MODEL_NAME,
prompt: { messages },
});
console.log("User:\n\n", first, "\n\n")
console.log("Palm:\n\n", result[0].candidates[0].content, "\n\n");
let second = "Oh, where do those live?"
messages.push({ content: result[0].candidates[0].content });
messages.push({ content: second });
const secondResult = await client.generateMessage({
model: MODEL_NAME,
prompt: { messages },
});
console.log("User:\n\n", second, "\n\n")
console.log("Palm:\n\n", secondResult[0].candidates[0].content, "\n\n");
}
main();
Overwriting index.js
node index.js
User: Tell me a one short animal fact. Palm: The world's smallest mammal is the bumblebee bat, which weighs less than a penny. User: Oh, where do those live? Palm: The bumblebee bat is found in the rainforests of Thailand, Myanmar, and Laos. It is a small, nocturnal bat that feeds on insects. The bumblebee bat is the smallest mammal in the world, weighing only about 2 grams. It is about the size of a bumblebee, hence its name. The bumblebee bat is a very important part of the rainforest ecosystem. It helps to control insect populations and pollinate plants.