Retrieval-augmented generation (RAG) enhances LLM responses by retrieving relevant information from external knowledge bases before generating answers. This reduces hallucination and enables LLMs to answer questions about private or domain-specific data without fine-tuning.
PAI provides tools to build, deploy, and manage RAG applications. Deploy a RAG service on EAS with a choice of vector databases and LLMs, or use LangStudio to develop and deploy RAG application flows for specialized domains.
Capabilities
| Capability | Description |
|---|---|
| Deploy a RAG chatbot on EAS | Deploy a RAG service that combines vector retrieval with LLM generation. Access the service through WebUI or API. WebUI lets you configure inference parameters and upload knowledge base files. |
| Develop RAG applications in LangStudio | Use LangStudio to develop and deploy RAG application flows in a visual, flow-based environment. Build RAG solutions tailored to specific domains such as finance and healthcare. |
| Integrate with messaging platforms | Use AppFlow to connect a PAI RAG service with third-party messaging platforms to build AI-powered chatbots and intelligent customer service agents. |
Supported components
RAG service on EAS supports flexible configuration of both retrieval and generation components.
| Component type | Supported options |
|---|---|
| Vector databases | Faiss, Elasticsearch, Hologres, OpenSearch, and RDS PostgreSQL |
| LLMs | Deploy models from Model Gallery, or connect to any LLM service that supports the OpenAI API. |
| Access methods | WebUI, API |
Get started
Approaches and documentation for building RAG applications on PAI:
| Approach | Description | Documentation |
|---|---|---|
| LangStudio | Build a domain-specific RAG application in a visual, flow-based interface. | Use LangStudio to create a DeepSeek- and RAG-based Q&A application flow for finance and healthcare |
| EAS scenario-based deployment | Deploy a general-purpose RAG chatbot with full control over vector databases and LLMs. | Deploy and call a RAG chatbot service |