Agent Workshop

This application is a Symfony-based playground for AI-powered features. It demonstrates how to integrate LLMs (Claude Haiku via Anthropic), vector stores (PostgreSQL with pgvector), and similarity search into a modern web application.

The project features:

• Conversational AI Chat: A streaming chat interface powered by Claude Haiku with tool-calling capabilities.
• Similarity Search: Posts are retrieved via semantic similarity using Voyage AI embeddings stored in pgvector.
• Automatic Tagging: Posts are automatically tagged based on content similarity to existing tags using vector embeddings.
• AI Tools: Custom tools (SimilarityPostTagger, TagCreator, StatusSearch) leverage LLMs to enrich application data.
• Admin CRUD: A full dynamic administration interface built with Symfony UX Live Components.

⚙️ Configuration

Before installing, configure your local environment variables:

1. Create a .env.local file:

   cp .env .env.local

2. Set the required variables in .env.local:

   • DATABASE_URL: PostgreSQL connection string (e.g. postgresql://app:!ChangeMe!@127.0.0.1:5432/app?serverVersion=16&charset=utf8)
   • ADMIN_EMAIL: Login email for the admin account
   • ADMIN_PWD: Password for the admin account
   • ANTHROPIC_API_KEY: Your Anthropic API key (required for AI chat and tool features)
   • VOYAGE_API_KEY: Your Voyage AI API key (required for vector embeddings)

🚀 Installation

The project uses a Makefile to simplify common tasks. Run make help at any time to list all available targets.

1. Full installation (builds containers, installs dependencies, creates DB and loads fixtures):

   make install

2. Set up AI stores (vector tables + indexing — requires API keys):

   make db-full

3. Start / stop the app:

   make start
   make stop

🗄️ Database Commands

Command	Description
make db	Create the database and run migrations
make fixtures	Load fixtures
make db-full	db + fixtures + AI message store + vector stores + indexing
make reset	Drop the database, run db-full, then clear cache

🤖 AI Store Commands

Command	Description
make store <name>	Set up a specific AI store
make store-drop <name>	Drop a specific AI store
make index <name>	Index documents into a specific AI store

🛠️ Development Commands

Command	Description
make composer	Install Composer dependencies
make test	Run PHPUnit tests
make cache-clear	Clear Symfony cache and shared cache pools
make symfony cmd="<cmd>"	Run any Symfony console command

Alternative: Without Make

If you don't have make installed, use Docker Compose commands directly:

docker compose build
docker compose up --wait
docker compose exec php composer install
docker compose exec php bin/console doctrine:database:create --if-not-exists
docker compose exec php bin/console doctrine:migrations:migrate --no-interaction
docker compose exec php bin/console foundry:load --no-interaction
docker compose exec php bin/console ai:message-store:setup ai.message_store.cache.messages
docker compose exec php bin/console ai:store:setup ai.store.postgres.posts
docker compose exec php bin/console ai:store:setup ai.store.postgres.tags
docker compose exec php bin/console ai:store:index posts
docker compose exec php bin/console ai:store:index tags

🏥 Health Check

1. Functional Tests: make test
2. Web Interface: https://localhost — homepage with latest posts
3. Admin Panel: https://localhost/admin — manage Posts, Tags, and Users
4. Docker Logs:

   docker compose logs -f php

5. Trainer Contact: If you can't get the application running, contact your trainer via the email provided in the SymfonyLive workshop instructions.