agents-from-scratch

What it is

The repository is a guide to building agents from scratch, culminating in an ambient agent that can manage email via Gmail API. It includes four sections with notebooks and accompanying code in src/email_assistant for agent basics, evaluation, human-in-the-loop, and memory.

How it works

The project demonstrates assembling an email assistant by:

• Building an agent that handles email triage and responses
• Integrating evaluation using an email dataset and Pytest with LangSmith evaluate API
• Adding human-in-the-loop via an interface (Agent Inbox) for reviewing tool calls
• Extending with memory using LangGraph Store to persist memories

Code references:

• src/email_assistant/email_assistant.py for the basic agent
• src/email_assistant/email_assistant_hitl.py for HITL
• src/email_assistant/email_assistant_hitl_memory.py for memory-enabled flow
• src/email_assistant/email_assistant_hitl_memory_gmail.py for Gmail integration

Getting started

Environment and installation steps emphasized in the README:

• Python version: 3.11 or later
• Create a .env file and set environment variables such as LANGSMITH_API_KEY, LANGSMITH_TRACING, OPENAI_API_KEY
• Package installation options:
  • Using uv:

    # Install uv if you haven't already
    pip install uv
    
    # Install the package with development dependencies
    uv sync --extra dev
    
    # Activate the virtual environment
    source .venv/bin/activate
    

  • Using pip (editable install):

    $ python3 -m venv .venv
    $ source .venv/bin/activate
    # Ensure you have a recent version of pip (required for editable installs with pyproject.toml)
    $ python3 -m pip install --upgrade pip
    # Install the package in editable mode
    $ pip install -e .
    

Important: The editable install is required for the notebooks to work correctly; the package is installed as interrupt_workshop with import name email_assistant.

Recent releases

• Release 0: none listed.

Traction

• Stars: 2051
• Forks: 402
• Open issues: 14

Behind the repo

• The project center is langchain-ai/agents-from-scratch, focusing on building agents with memory and HITL capabilities for email tasks.

Caveats

• Language: Jupyter Notebook
• License: MIT
• Created: 2025-03-31
• Last push: 2026-06-15
• Gmail integration requires Google API credentials per Gmail Tools README.