agentic-ecommerce

An ecommerce Agent demonstrating different concepts and features of LangGraph

Run the agent

Deploy it to cloud run

• Run the deploy.sh script
• Do a post request to /invoke

curl -X POST "[CLOUD_RUN_URL]/invoke" -H "Content-Type: application/json" -d '{
  "messages": ["Recommend me some blue clothing from your catalog for men"],
  "config": {"configurable": {"thread_id": "1"}}
}'

TODO

• Add telemetry for Agent calls
• Create modules to simply imports (nvillaluenga)
• Scripts to deploy web server (cloud run)
• Create basic chat interface (decide framework)
• Define set of tools
• Enable access to Mercado Pago API

Dependency Management

This project recommends using uv for managing dependencies.

What is uv?

• uv is a simple and efficient dependency manager for Python projects.
• It focuses on a minimal, declarative approach to defining and installing dependencies.
• uv uses a uv.toml file to specify project dependencies, similar to requirements.txt but with enhanced features.

How to use uv:

1. Installation:

   • macOS and Linux:

     $ curl -LsSf https://astral.sh/uv/install.sh | sh

   • Windows:

     $ powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

2. Sync the project

   uv sync

3. Activate the virtual environment

   . .venv/bin/activate

Now you are ready to go!

Telemetry with Langfuse

You can track LLM calls using LangFuse web service or you can deploy it on your own GCP environment.

LangFuse Platform

The easiest option is to create a LangFuse account and use their freemium plan. It's very easy to set up

After setting up your account you should complete the following environmental variables:

LANGFUSE_SECRET_KEY="<add your project's secret key>"
LANGFUSE_PUBLIC_KEY="<add your project's public key>"
LANGFUSE_HOST=https://us.cloud.langfuse.com

Setup on Google Cloud

Link to original repository

Prerequisites

• Google Cloud account
• gcloud CLI installed
• Terraform installed
• Enable the following APIs:
  • Serverless VPC Access API
  • Service Networking API
  • Cloud Run Admin API
  • Cloud Deployment Manager V2 API
  • Artifact Registry API
  • Cloud Identity-Aware Proxy API
  • Cloud Build API
  • Cloud Logging API

Getting Started

1. Set your own variables on terraform.tfvars

   • Replace the value surrounded by < >

2. Terraform init and apply

   terraform init
   terraform apply
   

3. Migrate Terraform state to GCS

   terraform init -migrate-state
   

   • Then generate backend.tf and have Terraform state managed by a GCS bucket

4. Access to Langfuse

   • Check your Cloud Run URL on Google Cloud Console (default: https://langfuse-<your-project-number>.<your-region>.run.app)
   • Create a project and generate secret & public keys

LANGFUSE_SECRET_KEY="<add your project's secret key>"
LANGFUSE_PUBLIC_KEY="<add your project's public key>"
LANGFUSE_HOST="https://langfuse-<your-project-number>.<your-region>.run.app"