Agent Workflow Starter Kit

Broadly speaking, the future of getting things done with our new magical AI APIs are “Agents.” The definition varies, but I think the folks at Anthropic did a great job on the blog post Building effective agents.

I wanted to try out two appraoches.

• Workflow: a flowchart of decisions and tool calls to answer a prompt from the user. It commonly will have steps to determine which tools to use and evaluate the results.
• Agent: an LLM (like OpenAI) is given a prompt from the user and a set of tool definitions and decides how to use them to answer the prompt.

In this app, I made it easy to try out both to talk to Salesforce data (fake data checked into a duckdb database in this repo).

Quickstart

To get started with this example, follow these steps:

1. Ensure you have Python and Poetry installed on your system.
2. Clone the repository and navigate to this directory.
3. poetry env use 3.10.15
4. Install the required dependencies by running poetry install.
5. Set your OpenAI API key as an environment variable: cp .env.example .env and add your key to .env.
6. Download the embedding model: poetry run download_embedding_model
7. Run the application: ./dev.sh.
8. Open http://localhost:8080/chat

Running the test suite

poetry run pytest -s app/tests

More

Read more about this on the blog post here