Skip to content

Latest commit

 

History

History
77 lines (52 loc) · 3.04 KB

README.md

File metadata and controls

77 lines (52 loc) · 3.04 KB
Raw

Stockbroker Human in the Loop

The code for the Stockbroker Human in the Loop video can be found in this directory. It's setup as a monorepo-style project, with frontend and backend directories. The frontend directory contains a Next.js application which allows you to interact with the Stockbroker agent via a chat interface. The backend contains a LangGraph agent which powers the core functionality of the stockbroker.

Deployment

The stockbroker agent is publicly accessible through two interfaces:

  1. API:

The Cloud API for the stockbroker agent is publicly accessible at the following base URL: https://stockbrokeragent-bracesprouls-projects.vercel.app/api

You can find the REST documentation for the stockbroker agent here.

Note The rest documentation displays a "base URL" which is not exactly correct. To hit the API, you'll need to append /api to the end of the base URL listed.

  1. Web-based Chat Interface:

To go along with the API, we've also deployed this web-based chat interface for the stockbroker agent.

You can access, and interact with it here.

YouTube Video

Setup

To setup the stockbroker, install dependencies from the root of the monorepo:

yarn install

This will install all dependencies required by both the frontend and backend projects. You can also run shared commands from the root of the project:

yarn format

yarn build

Environment variables

Backend

The backend requires Financial Datasets AI, Tavily and OpenAI API keys to run. Sign up here:

Once you have your API keys, create a .env file in the ./backend directory and add the following:

FINANCIAL_DATASETS_API_KEY=YOUR_API_KEY
TAVILY_API_KEY=YOUR_API_KEY
OPENAI_API_KEY=YOUR_API_KEY

Frontend

The frontend requires the production, or local deployment of your agent, along with a LangSmith API key (if calling the production endpoint), and finally the name of the agent to interact with (in this case stockbroker).

For local development, you can find the API endpoint in the bottom left of LangGraph Studio, which defaults to http://localhost:51497. You can find the production URL in the deployment page of your LangGraph cloud deployment.

Then, set the variables in a .env file inside ./frontend:

# Only required for production deployments
# LANGCHAIN_API_KEY=YOUR_API_KEY
LANGGRAPH_API_URL=http://localhost:51497
NEXT_PUBLIC_API_URL=http://localhost:3000/api # Or your production URL + /api
NEXT_PUBLIC_LANGGRAPH_GRAPH_ID=stockbroker

LangGraph Config

The LangGraph configuration file for the stockbroker project is located inside ./backend/langgraph.json. This file defines the stockbroker graph implemented in the project: stockbroker.