langchain-starter

OpenGradient + LangChain Agent

Combines OpenGradient's verifiable inference and advanced ML tools with LangChain.

Structure

• The agent's tools are defined in tools.py
• The agent's prompts and instructions are in prompts.py
• The agent itself is defined (using the tools and prompts) in agent.py
• The main function for executing the agent is in main.py

Agent

The agent is equipped by the following tools:

• Wikipedia search tool for looking up articles
• OpenGradient AlphaSense tool for Sui price prediction.

Setup

1. Dependencies are defined in requirements.txt, run pip install -r requirements.txt to install all dependencies. We recommend using virtualenv.
2. You'll need an OpenGradient token in order to run inferences. You can use the opengradient CLI to create it, read more on our official docs.
3. Make a copy of the .env.sh.template file in the root folder and rename to .env.sh. Fill in with your OpenGradient token as well as Twitter API keys.

Note: If you run into git permission denied error messages, please generate a Github SSH key.

ssh-keygen -t ed25519 -C "your_email@example.com"
eval "$(ssh-agent -s)"
ssh-add ~/.ssh/id_ed25519

Then proceed to add the output of the below command to your Github SSH keys in the settings page.

cat ~/.ssh/id_ed25519.pub

Execute

To execute the agent, simply run python main.py <your-instruction-here>. For example, python main.py "what is the size of NYC"

Next, you can add new tools, prompts, user input etc to customize the agent's behaviour.