Agent Chain

Agent Chain is a simple CLI built on top of LangChain for running AI agents. It is meant for testing and experimenting with different AI agents.

Dependencies

• Poetry
• Docker

Getting started

Start a Redis-search instance:

docker run --name redis-baby-agi -p 6379:6379 -d redislabs/redisearch

Enter the poetry shell in order to access the CLI

poetry shell
python -m agent_chain --help

In order to use the BabyAGI or AutoGPT commands, the environment-variable OPENAI_API_KEY must be set. For convenience, you can create a .env file in the root of the project with the following content:

OPENAI_API_KEY=<your-key-here>

The CLI will automatically load this file if it exists.

For the AI to write to file, the bin-directory must exist in the working directory.

Examples

Use AutoGPT to plan a romantic trip to Paris

python -m agent_chain auto-gpt "Plan a romantic weekend in Paris, write the plan to file"

Use AutoGPT to research lobster-fishing using Ocean Info Hub:

python -m agent_chain auto-gpt "Research the best places to fish for lobster around Norway. Prioritize using Ocean Info Hub, but research the Ocean Info Hub graph before attempting any queries. Make a todo-list before anything else. Write results to file"

Use BabyAGI to plan a romantic trip to Paris

python -m baby-agi --max-iterations 5 "Plan a romantic weekend in Paris"