Google Maps Recommendations Agent + Automation

Overview

A small and scrappy app to explore a few technologies I've been interested in. These are Langgraph, OpenAI Agents SDK, and Playwright. This app uses Playwright to login to a user's Google account and add's AI generated recommendations to places in a Google Maps saved list (interestingly, Google doesn't expose an API for Saved Places). The recommendations are specific to the user based on information saved in their traveller profile.

A fun next step would be to make the browser automation add recommended locations to the user's map as new saved locations.

App Setup

This app uses uv as the package manager. The app requires the following environment variable to be set:

OPENAI_API_KEY: str = <your openai api key>
LANGSMITH_TRACING: bool = <flag to enable/disable Langsmith tracing>
LANGSMITH_ENDPOINT: str = <Langsmith endpoint>
LANGSMITH_API_KEY: str = <your Langsmith api key>
LANGSMITH_PROJECT: str = <your Langsmith project name>
SERPER_API_KEY: str = <your Serper API key>
GOOGLE_EMAIL: str = <your google email address>
GOOGLE_PASSWORD: str = <your google password>
GOOGLE_2FA_SECRET: str = <secret key for your google two factor auth code>
GOOGLE_MAPS_LIST_NAME = <the name of your Google maps saved place list>

To install a venv with dependencies

uv venv
uv sync
source .venv/bin/activate

To add a traveller profile

Save a JSON object as app/llm_agents/traveller_profile_agent/input/traveller_profile.json that adheres to the TravellerProfile schema.

To run the app

cd <path_to_project>
source .venv/bin/activate
uv run -m app.main

Note: The app code isn't optimised and it doesn't include automated tests.