Abstract:In the current technological landscape, there have been the emergence of application programming interfaces (APIs) based on large-scale language models, such as OpenAI’s GPT-4 API and ChatGPT. These APIs allow developers to directly call the interface to retrieve response information from large-scale language models. However, these technologies lack the capability to connect large-scale language models with local data sources and interact with the environment.To address this issue, this paper proposes an improved system and method for developing large-scale model applications based on the LangChain+ChatGLM+Rasa framework. The system utilizes the LangChain framework to establish connections between fine-tuned AI models like ChatGLM-6B and local vector data sources, as well as enable interaction with the local environment. This improvement brings the following advantages: it enables the connection and interaction between fine-tuned AI models like ChatGLM-6B and local vector data sources, thereby building more intelligent applications. It provides an easy-to-use modular abstraction and customization capability for specific chains, allowing for rapid application development and reducing development time and costs. It achieves customization of specific chains by combining components in specific ways to meet specific requirements, and it can be flexibly modified and expanded according to the needs. It integrates the technologies of the Rasa dialogue system and embedded vector retrieval system. By using the Rasa system for intent recognition and determining the query type, it achieves efficient matching and semantic analysis for text summarization applications in work order scenarios.This method explores the practical development of large-scale AI applications and provides comprehensive AI application development services for cloud network monitoring and scheduling systems. It accelerates technological exploration and promotes the practical application of these scenarios.

ChainStream: A Stream-based LLM Agent Framework for Continuous Context Sensing and Sharing

Chain of Agents: Large Language Models Collaborating on Long-Context Tasks

Large-Scale Real-Time Semantic Processing Framework for Internet of Things

ChainBuddy: An AI Agent System for Generating LLM Pipelines

StreamBench: Towards Benchmarking Continuous Improvement of Language Agents

Edge-Cloud Collaborative Streaming Video Analytics with Multi-agent Deep Reinforcement Learning

A Streaming End-to-End Framework for Spoken Language Understanding.

StreamChat: Chatting with Streaming Video

Adaptive Stream Processing on Edge Devices through Active Inference

Prompt Sapper: A LLM-Empowered Production Tool for Building AI Chains

AgentVerse: Facilitating Multi-Agent Collaboration and Exploring Emergent Behaviors

Internet of Agents: Weaving a Web of Heterogeneous Agents for Collaborative Intelligence

AgentLite: A Lightweight Library for Building and Advancing Task-Oriented LLM Agent System

ChatDev: Communicative Agents for Software Development

Chat with the Environment: Interactive Multimodal Perception Using Large Language Models

Multi-Agent Software Development through Cross-Team Collaboration

Experiential Co-Learning of Software-Developing Agents

Application development exploration and practice based on LangChain+ChatGLM+Rasa

Creating Large Language Model Applications Utilizing LangChain: A Primer on Developing LLM Apps Fast

Stream-based perception for cognitive agents in mobile ecosystems