Logo
Section 1 / 4

基础 RAG 与代理型 RAG 的 LlamaIndex 实现

使用 LlamaIndex 实现并直接对比标准静态 RAG 流水线与基础单工具代理式 RAG 系统。

导入

基础 RAG 与代理型 RAG 的 LlamaIndex 实现

使用 LlamaIndex 实现并直接对比标准静态 RAG 流水线与基础单工具代理式 RAG 系统。

...
Agent implementation frameworks

project_cygnus_overview.txt

1Project Cygnus is a next-generation, AI-powered data analytics platform developed by the fictional
2company "Innovate Inc." The project lead is Dr. Aris Thorne. The primary goal of Project Cygnus is 
3to achieve a beta release by Q4 2025, targeting enterprise-level financial institutions.

project_cygnus_technical_spec.txt

1The backend of Project Cygnus is built on a microservices architecture using Python and Rust.
2For data processing, it leverages a distributed graph database and a real-time event streaming
3pipeline using Apache Kafka. The frontend is a progressive web app built with React.

No configuration found for ID: ae55e04bd60e4bad5285c81ac1a9b9cf721221a6bdb317b991b15022fee6168a

No configuration found for ID: d8f8ac9c6aa0f1924c6c776e1c7def8a6fe5a5a8b31104dfaf3ce7e0d5a57a90

Code Example

1HTTP Request: POST https://api.groq.com/openai/v1/chat/completions "HTTP/1.1 200 OK"

1HTTP Request: POST http://localhost:11434/api/embed ...
2HTTP Request: POST https://api.groq.com/openai/v1/chat/completions ...

Code Example

1HTTP Request: POST https://api.groq.com/openai/v1/chat/completions "HTTP/1.1 200 OK"

AspectStatic RAG PipelineAgentic RAG System
Core ObjectA monolithic QueryEngine.A ReActAgent that orchestrates a list of Tools.
InteractionWe call a single .query() method.We interact conversationally using methods like .run().
FlexibilityInflexible. It can only perform RAG.Highly flexible. The agent decides if RAG is needed.
ExtensibilityHard to extend. Adding a new capability (e.g., a calculator) would require rewriting the core logic.Easy to extend. We can simply add a new Tool to the agent’s tools list.

Agentic RAG 的基础

设计我们的代理型 RAG 系统：AI 研究助理