🪐 LifeStack Engine

Meta × HuggingFace OpenEnv Hackathon Finale

PRODUCTION HARDENED v3.0

Simulation Control

Subject Persona

Life Conflict

Context Augmentation (RAG)

AI Risk Prediction

Analyzing...

Vital MetricsSTABLE

Domain Risk Heatmap

WORK

MONEY

SOCIAL

BODY

MIND

TIME

--

ID: --

⏱️ --

💵 --

⚖️ What-If Analysis

Retrieved Precedents (RAG) +116% EFFICIENCY

7-DAY LIFE TRAJECTORY

What happens to your six domains over the next week if nothing extraordinary occurs — computed from the real LifeStack env rollout, not estimated.

Discounted Reward (γ=0.9)

Describe Your Situation

💼 Work Stress 💰 Money Stress

❤️ Relationships ⚡ Energy

Digital Life Signals

Click "Sync Digital Life" to pull signals from your connected services.

📧 Gmail

📅 Google Calendar

💪 Fitness

Demo

💪 Health Data Upload

Sleep (h)

HR (bpm)

Steps/day

📅 Calendar Data Upload

Occupancy %

Back-to-back

Critical Deadlines (count)

Personalised Resolution

ID: --

Conflict Scenario

Subject Persona

Select a conflict and run to see how GRPO training changes the LLM's decisions.

No RL Training

Untrained LLM

reward

Strategy

GRPO-Trained LifeStack

LifeStack Agent

reward

Strategy

Memory Effect Demo

Same conflict, same agent. Episode 1 runs cold (no prior context). Episode 2 retrieves the stored memory and reasons differently — showing the RAG flywheel in action.

+116% Efficiency

Conflict

Persona

Run both episodes to see how memory transforms the agent's reasoning.

Episode 1 — Cold Start

No Memory

Episode 2 — With Memory

RAG-Augmented

Constant Conflict

Persona A

Persona B

OCEAN Personality Radar

Openness · Conscientiousness · Extraversion · Agreeableness · Neuroticism

--

Score: --

--

Score: --

🗓️ Longitudinal Context Persistence

By activating Arjun's history, the agent retrieves memories of past decisions from **ChromaDB**. This transforms a zero-shot LLM into a context-aware coach that understands your specific relationship dynamics and work history.

Experience the v3 Flywheel:

Click Activate History to seed ChromaDB with Arjun's Week 1 & 2 data.
Go back to Situational Portal.
Select Arjun (Startup Lead) and run the simulation.
Observe Reasoning: The agent will explicitly mention precedents from his past (e.g., flight cancellations or funding stress).

COUNTERFACTUAL EXPLORER

Causal Reasoning Engine

Run a simulation in the Situational Portal to unlock deep counterfactual analysis.

ACTUAL CHOICE

--

ALTERNATIVE A

--

ALTERNATIVE B

--

EPISODE REPLAY

Audit log of agent reasoning & causal chains

No episodes logged yet. Start a simulation in the Portal!

🎯 Objective: Select a task

🛣️ Viable Routes

⚡ Progressive Milestones

☄️ World Event Log

Learning efficiency (GRPO)

Memory Persistence

Model Strategy

"Prioritizing long-term relationship preservation over short-term career gain in high-neuroticism subjects."

📬 Outcome Signal

Record the actual result of the agent's plan to close the reinforcement loop. High-fidelity feedback stored in ChromaDB improves future planning.

Episode Trace ID

Actual Hours

Success Score (0-10)

Unexpected Deviations