lifeops / lifeops-context.md
hilarycc's picture
initial environment setup
d40203a

LifeOps Project Context

You are helping build an MVP for an OpenEnv hackathon project called LifeOps.

LifeOps is a reinforcement learning environment where an AI agent learns to manage a user's life schedule.

The environment simulates calendar management, personal goals, travel constraints, and incoming requests.

The purpose is to train agents to make realistic planning decisions using reinforcement learning.


Core Project Goal

Build a personalized planning environment where an AI agent manages:

  • calendar events
  • personal goals/tasks
  • travel constraints
  • incoming messages or scheduling requests

The environment should expose:

  • a structured state
  • a constrained action space
  • a reward function

This allows reinforcement learning algorithms to train agents to improve planning decisions.


MVP Scope

Keep the system simple and hackathon-friendly.

Constraints:

  • Python only
  • No unnecessary frameworks
  • Clean modular structure
  • Readable code

Design decisions:

  • Use structured JSON-like state instead of raw text
  • Use a one-day planning horizon
  • Include user personas with preferences and behavioral tendencies

Environment should include:

  • calendar events
  • tasks/goals
  • travel times
  • incoming requests/messages

Action Space

The agent should choose structured actions such as:

  • accept_event
  • reject_event
  • reschedule_event
  • propose_new_time
  • block_focus_time

Avoid free-form natural language actions.


Reward Design

Penalize:

  • double booking
  • impossible travel
  • violating strong user preferences
  • missing important obligations

Reward:

  • resolving scheduling conflicts
  • respecting user habits
  • allocating time to goals
  • correctly handling important requests

Coding Rules

Follow these rules when generating code:

  • Keep files small and modular
  • Add docstrings and comments
  • Prefer dataclasses or typed dictionaries for structured data
  • Avoid unnecessary abstractions
  • Avoid complex frameworks
  • Prefer clarity over cleverness

Add basic input validation and avoid unsafe patterns.

Follow general OWASP best practices where applicable.


Initial Repository Targets

env/personas.py env/actions.py env/scenario_generator.py env/reward.py env/lifeops_env.py

tests/test_env.py tests/test_reward.py


Validation Requirements

The code must:

  • run locally
  • include a simple manual episode runner
  • include at least 3 personas
  • include at least 5 sample scenarios
  • include at least 5 tests

Tests should verify:

  • reward calculation
  • environment step logic
  • conflict detection
  • travel feasibility
  • persona preference handling

Implementation Philosophy

This project is a simulation environment, not a full product.

Focus on:

  • clarity
  • correctness
  • simple reinforcement learning compatibility

Do not build UI unless explicitly requested.

If something is ambiguous, make a reasonable assumption and document it.