Restez organisé à l'aide des collections
Enregistrez et classez les contenus selon vos préférences.
Présentation de l'IA générative
Google Cloud propose une gamme de produits et d'outils pour l'ensemble du cycle de vie de la création d'applications d'IA générative.
Exploration et hébergement de modèles
Google Cloud fournit un ensemble de modèles de fondation de pointe via Vertex AI, y compris Gemini. Vous pouvez également déployer un modèle tiers vers Vertex AI Model Garden ou l'auto-hébergement sur GKE ou Compute Engine.
Conception et ingénierie des requêtes
La conception de requêtes est le processus de création de paires de requêtes et de réponses pour fournir aux modèles de langage un contexte et des instructions supplémentaires. Une fois que vous avez créé des requêtes, vous les transmettez au modèle en tant qu'ensemble de données de requêtes pour le pré-entraînement. Lorsqu'un modèle génère des prédictions, il répond avec vos instructions intégrées.
Ancrage et RAG
L'ancrage connecte les modèles d'IA à des sources de données pour améliorer la précision des réponses et réduire les hallucinations. Le RAG, une technique d'ancrage courante, recherche des informations pertinentes et les ajoute à la requête du modèle, ce qui garantit que la sortie est basée sur des faits et des informations à jour.
Appels d'agents et de fonctions
Les agents permettent de concevoir et d'intégrer facilement une interface utilisateur de conversation dans votre application mobile, tandis que l'appel de fonction étend les capacités d'un modèle.
Personnalisation et entraînement du modèle
Les tâches spécialisées, telles que l'entraînement d'un modèle de langage sur une terminologie spécifique, peuvent nécessiter plus d'entraînement qu'une conception de requête ou un ancrage seul. Dans ce cas, vous pouvez utiliser le réglage du modèle pour améliorer les performances ou entraîner votre propre modèle.
Configurer votre environnement de développement pour Google Cloud
Configurer LangChain
LangChain est un framework Open Source pour les applications d'IA générative qui vous permet d'intégrer du contexte dans vos requêtes et d'agir en fonction de la réponse du modèle.
Afficher des exemples de code et déployer des applications exemples
Sauf indication contraire, le contenu de cette page est régi par une licence Creative Commons Attribution 4.0, et les échantillons de code sont régis par une licence Apache 2.0. Pour en savoir plus, consultez les Règles du site Google Developers. Java est une marque déposée d'Oracle et/ou de ses sociétés affiliées.
Dernière mise à jour le 2024/12/22 (UTC).
[[["Facile à comprendre","easyToUnderstand","thumb-up"],["J'ai pu résoudre mon problème","solvedMyProblem","thumb-up"],["Autre","otherUp","thumb-up"]],[["Difficile à comprendre","hardToUnderstand","thumb-down"],["Informations ou exemple de code incorrects","incorrectInformationOrSampleCode","thumb-down"],["Il n'y a pas l'information/les exemples dont j'ai besoin","missingTheInformationSamplesINeed","thumb-down"],["Problème de traduction","translationIssue","thumb-down"],["Autre","otherDown","thumb-down"]],["Dernière mise à jour le 2024/12/22 (UTC)."],[[["\u003cp\u003eGoogle Cloud provides comprehensive tools and products for every stage of building generative AI applications, from model exploration to deployment.\u003c/p\u003e\n"],["\u003cp\u003eVertex AI allows users to access, test, tune, and deploy Google's large generative AI models, including Gemini, for use in AI-powered applications.\u003c/p\u003e\n"],["\u003cp\u003ePrompt design and engineering, including using Vertex AI Studio, are crucial for shaping model responses and optimizing their effectiveness.\u003c/p\u003e\n"],["\u003cp\u003eGrounding techniques, like RAG, connect AI models to data sources to improve accuracy and reduce hallucinations, using tools like Google Search, AlloyDB, Cloud SQL, and more.\u003c/p\u003e\n"],["\u003cp\u003eDevelopers can customize and train models, using tools like Cloud TPU, and evaluate performance with Vertex AI to enhance model effectiveness on specialized tasks.\u003c/p\u003e\n"]]],[],null,["# Generative AI\n=============\n\nDocumentation and resources for building and implementing generative AI\napplications with Google Cloud tools and products.\n[Get started for free](https://console.cloud.google.com/freetrial) \n\n#### Start your proof of concept with $300 in free credit\n\n- Get access to Gemini 2.0 Flash Thinking\n- Free monthly usage of popular products, including AI APIs and BigQuery\n- No automatic charges, no commitment \n[View free product offers](/free/docs/free-cloud-features#free-tier) \n\n#### Keep exploring with 20+ always-free products\n\n\nAccess 20+ free products for common use cases, including AI APIs, VMs, data warehouses,\nand more.\n\nLearn about building generative AI applications\n-----------------------------------------------\n\n### [Generative AI on Vertex AI](/vertex-ai/generative-ai/docs/overview)\n\nAccess Google's large generative AI models so you can test, tune, and deploy them for use in your AI-powered applications. \n\n### [Gemini Quickstart](/vertex-ai/generative-ai/docs/start/quickstarts/quickstart-multimodal)\n\nSee what it's like to send requests to the Gemini API through Google Cloud's AI-ML platform, Vertex AI. \n\n### [AI/ML orchestration on GKE](/kubernetes-engine/docs/integrations/ai-infra)\n\nLeverage the power of GKE as a customizable AI/ML platform featuring high performance, cost effective serving and training with industry-leading scale and flexible infrastructure options. \n\n### [When to use generative AI](/docs/ai-ml/generative-ai/generative-ai-or-traditional-ai)\n\nIdentify whether generative AI, traditional AI, or a combination of both might suit your business use case. \n\n### [Develop a generative AI application](/docs/ai-ml/generative-ai/develop-generative-ai-application)\n\nLearn how to address the challenges in each stage of developing a generative AI application. \n\n### [Code samples and sample applications](/docs/generative-ai/code-samples)\n\nView code samples for popular use cases and deploy examples of generative AI applications that are secure, efficient, resilient, high-performing, and cost-effective. \n\n### [Generative AI glossary](/docs/generative-ai/glossary)\n\nLearn about specific terms that are associated with generative AI.\n\nGen AI tools\n------------\n\nGen AI development flow\n-----------------------\n\nModel exploration and hosting\n-----------------------------\n\nGoogle Cloud provides a set of state-of-the-art foundation models through Vertex AI, including Gemini. You can also deploy a third-party model to either Vertex AI Model Garden or self-host on GKE or Compute Engine. \n\n### [Google Models on Vertex AI (Gemini, Imagen)](/vertex-ai/generative-ai/docs/learn/models)\n\nDiscover test, customize, and deploy Google models and assets from an ML model library. \n\n### [Other models in the Vertex AI Model Garden](/vertex-ai/generative-ai/docs/model-garden/explore-models)\n\nDiscover, test, customize, and deploy select OSS models and assets from an ML model library. \n\n### [Text generation models via HuggingFace](/vertex-ai/generative-ai/docs/open-models/use-hugging-face-models)\n\nLearn how to deploy HuggingFace text generation models to Vertex AI or Google Kubernetes Engine (GKE). \n\n### [GPUs on Compute Engine](/compute/docs/gpus/about-gpus)\n\nAttach GPUs to VM instances to accelerate generative AI workloads on Compute Engine.\n\nPrompt design and engineering\n-----------------------------\n\nPrompt design is the process of authoring prompt and response pairs to give language models additional context and instructions. After you author prompts, you feed them to the model as a prompt dataset for pretraining. When a model serves predictions, it responds with your instructions built in. \n\n### [Vertex AI Studio](/vertex-ai/generative-ai/docs/start/quickstarts/quickstart)\n\nDesign, test, and customize your prompts sent to Google's Gemini and PaLM 2 large language models (LLM). \n\n### [Overview of Prompting Strategies](/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies)\n\nLearn the prompt-engineering workflow and common strategies that you can use to affect model responses. \n\n### [Prompt Gallery](/vertex-ai/generative-ai/docs/prompt-gallery)\n\nView example prompts and responses for specific use cases.\n\nGrounding and RAG\n-----------------\n\n*Grounding* connects AI models to data sources to improve the accuracy of responses and reduce hallucinations. *RAG*, a common grounding technique, searches for relevant information and adds it to the model's prompt, ensuring output is based on facts and up-to-date information. \n\n### [Vertex AI grounding](/vertex-ai/generative-ai/docs/grounding/overview)\n\nYou can ground Vertex AI models with Google Search or with your own data stored in Vertex AI Search. \n\n### [Ground with Google Search](/vertex-ai/generative-ai/docs/multimodal/ground-gemini#web-ground-gemini)\n\nUse Grounding with Google Search to connect the model to the up-to-date knowledge available on the internet. \n\n### [Vector embeddings in AlloyDB](/alloydb/docs/ai/work-with-embeddings)\n\nUse AlloyDB to generate and store vector embeddings, then index and query the embeddings using the pgvector extension. \n\n### [Cloud SQL and pgvector](https://github.com/pgvector/pgvector?tab=readme-ov-file#pgvector)\n\nStore vector embeddings in Postgres SQL, then index and query the embeddings using the pgvector extension. \n\n### [Integrating BigQuery data into your LangChain application](https://cloud.google.com/blog/products/ai-machine-learning/open-source-framework-for-connecting-llms-to-your-data)\n\nUse LangChain to extract data from BigQuery and enrich and ground your model's responses. \n[description](/firestore/docs/vector-search) \n\n### [Vector embeddings in Firestore](/firestore/docs/vector-search)\n\nCreate vector embeddings from your Firestore data, then index and query the embeddings. \n\n### [Vector embeddings in Memorystore (Redis)](/memorystore/docs/redis/about-vector-search)\n\nUse LangChain to extract data from Memorystore and enrich and ground your model's responses.\n\nAgents and function calling\n---------------------------\n\nAgents make it easy to design and integrate a conversational user interface into your mobile app, while function calling extends the capabilities of a model. \n\n### [AI Applications](/generative-ai-app-builder/docs/introduction)\n\nLeverage Google's foundation models, search expertise, and conversational AI technologies for enterprise-grade generative AI applications. \n\n### [Vertex AI Function calling](/vertex-ai/generative-ai/docs/multimodal/function-calling)\n\nAdd function calling to your model to enable actions like booking a reservation based on extracted calendar information.\n\nModel customization and training\n--------------------------------\n\nSpecialized tasks, such as training a language model on specific terminology, might require more training than you can do with prompt design or grounding alone. In that scenario, you can use model tuning to improve performance, or train your own model. \n\n### [Evaluate models in Vertex AI](/vertex-ai/generative-ai/docs/models/evaluation-overview)\n\nEvaluate the performance of foundation models and your tuned generative AI models on Vertex AI. \n\n### [Tune Vertex AI models](/vertex-ai/generative-ai/docs/models/tune-models)\n\nGeneral purpose foundation models can benefit from tuning to improve their performance on specific tasks. \n\n### [Cloud TPU](/tpu/docs)\n\nTPUs are Google's custom-developed ASICs used to accelerate machine learning workloads, such as training an LLM.\n\nRelated guides and sites\n------------------------\n\n[description](/architecture/gen-ai-rag-vertex-ai-vector-search) \nIntermediate\n\n### [Infrastructure for a RAG-capable generative AI application using Vertex AI and Vector Search](/architecture/gen-ai-rag-vertex-ai-vector-search)\n\nReference architecture for a RAG-capable generative AI application using Vertex AI and Vector Search. \n[description](/architecture/rag-capable-gen-ai-app-using-vertex-ai) \nIntermediate\n\n### [Infrastructure for a RAG-capable generative AI application using Vertex AI and AlloyDB for PostgreSQL](/architecture/rag-capable-gen-ai-app-using-vertex-ai)\n\nReference architecture for a RAG-capable generative AI application using Vertex AI and AlloyDB for PostgreSQL. \n[description](/architecture/rag-capable-gen-ai-app-using-gke) \nIntermediate\n\n### [Infrastructure for a RAG-capable generative AI application using GKE and Cloud SQL](/architecture/rag-capable-gen-ai-app-using-gke)\n\nReference architecture for a RAG-capable generative AI application using GKE, Cloud SQL, and open source tools like Ray, Hugging Face, and LangChain.\n\nStart building\n--------------\n\n### Set up your development environment for Google Cloud\n\n- [C# and .NET](/dotnet/docs/setup)\n- [C++](/cpp/docs/setup)\n- [Go](/go/docs/setup)\n- [Java](/java/docs/setup)\n- [JavaScript and Node.js](/nodejs/docs/setup)\n- [Python](/python/docs/setup)\n- [Ruby](/ruby/docs/setup)\n\n### Set up LangChain\n\nLangChain is an open source framework for generative AI apps that allows you to build context into your prompts, and take action based on the model's response.\n\n- [Python (LangChain)](https://python.langchain.com/docs/integrations/llms/google_vertex_ai_palm)\n- [JavaScript (LangChain.js)](https://js.langchain.com/docs/integrations/platforms/google)\n- [Java (LangChain4j)](https://docs.langchain4j.dev/integrations/language-models/google-palm/)\n- [Go (LangChainGo)](https://tmc.github.io/langchaingo/docs/)\n\n### View code samples and deploy sample applications\n\nView [code samples for popular use cases and deploy examples of generative AI applications](/docs/generative-ai/code-samples) that are secure, efficient, resilient, high-performing, and cost-effective."]]