We built three agents; Chat, Search, Job Notification, and Supervisor Agent. The Supervisor Agent determines which agent to use and routes the user prompt accordingly. All agents run on the LangGraph.js server.
When we developed the server agents, we had in mind a web-based application that we are going to build. This application will make requests to the LangGraph Server agents and render the responses. It will connect to any agent, including the supervisor, and display all previous thread similar to ChatGPT where users can revisit and continue conversations. The app will also allow users to edit prompts and resubmit them (fork them), just like in ChatGPT
LangGraph.js is a library for building stateful, multi-actor applications with LLMs, used to create agent and multi-agent workflows. Compared to other LLM frameworks, it offers these core benefits; cycles, controllability, and persistence.
The LangGraph Server is a component of the LangGraph platform, designed to facilitate the development and deployment of applications built with LangChain's LangGraph framework. It serves as an API server that integrates various tools and technologies to support the execution and management of LangGraph applications.
The Chat Agent is used for answering simple questions, such as Generate 5-letter English words. Note: You can edit the question to receive a new answer (fork it).
The Search Agent searches the web in real time using Tavily Search.
I developed a Job Notification agent in Node.js. You can explore the project.
In the Job Notification agent, the AI agent will select the matching job categories based on a prompt from the Categories.json file. Then, from the respective categories, matching candidates in the Candidates.json file will be selected based on their skills. The app follows a human-in-the-loop (HITL) approach, requiring human intervention before proceeding with certain tasks. If the selected best candidate does not satisfy the recruiter, they can skip it and ask for the next best candidate. If the best candidate is selected, an email notification generated by AI will be sent to the candidate. If the generated email does not satisfy the recruiter; they can skip it and ask for another one with different content.
Supervisor Agent determines which agent to use and routes the user prompt accordingly.
To get started.
# Clone the repository
git clone https://github.com/Ashot72/Ashot72-LangGrap.js-Server-Agents
cd Ashot72-LangGrap.js-Server-Agents
# Create the .env file based on the env.example.txt file and include the respective keys.
# installs dependencies
npm install
# to run locally
npm start
Go to LangGraph.js Server Agents Video page
Go to LangGraph.js Server Agents Description page