Calendar Environment#

This environment exposes a Calendar Gym tools through the OpenEnv reset/step/state interface. The server runs a FastAPI app that serves the OpenEnv endpoints.

Server Setup#

Docker (Recommended)#

cd envs/calendar_env
docker build -t calendar-env:latest .
docker run --rm -p 8004:8004 calendar-env:latest
curl http://localhost:8004/health

On Server health success response will be: {"status":"healthy","service":"calendar-env"}

Without Docker#

cd envs/calendar_env
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
uvicorn server.app:app --host 0.0.0.0 --port 8004

Client Setup#

Quick Start (Demo)#

For a quick demo, simply update llm_api_key in scenario_config.json and run:

python client.py --scenario scenario_config.json

The existing config includes a sample scenario for testing.

Configure Scenario#

To customize for your use case, edit scenario_config.json and update these fields:

llm variables:

llm_api_key - Your OpenAI/Anthropic/Google API key (or set via env var)
llm_model - Model name (e.g., gpt-4o-mini, claude-3-5-sonnet-20241022)
llm_provider - Provider: openai, anthropic, or google

Scenario Variables

user_prompt - Task for the agent to complete
system_prompt - Instructions for agent behavior
context - The auth headers for gym like (x-access-token)
seed_database_file - Path to SQL file for custom data
verifiers - SQL queries to validate task completion
expected_tools - Tools agent should use (for tracking)

Run Client#

Run scenario-based benchmark:

python client.py --scenario scenario_config.json 

Output will be saved to response_output/ folder with execution details, tool calls, and verification results.

Notebook Evaluation: For interactive evaluation and testing, use the: Jupyter notebook