Eirene

1. Inspiration

Mental health crises are surging online, but responses are slow and generic. We envisioned an AI system that could detect community stress signals in real-time and generate personalized, clinically-grounded video interventions autonomously—bringing empathy at scale to those who need it most.

2. What It Does

Eirene monitors the web for mental health stress signals, maps them to evidence-based therapeutic frameworks (CBT/DBT), and autonomously generates "Mindful Moment" videos with calming visuals and therapeutic voiceovers. The system validates content for emotional safety, learns from past interventions, and operates end-to-end without human input—from detection to video creation.

3. How It Was Built

We built a multi-agent architecture with five specialized AI agents coordinated by a manager: Monitor (TinyFish MCP + AgentQL for web scraping), Strategy (Yutori knowledge base for therapeutic mapping), Generation (Freepik + ElevenLabs for video/audio), Audit (Modulate API for emotional validation), and Memory (Yutori for self-improvement). The system includes comprehensive error handling, trace logging, and an interactive web dashboard for testing.

4. Challenges We Ran Into

Coordinating five autonomous agents with different API dependencies was complex, requiring robust error handling and retry logic. Ensuring clinical validity while maintaining creative flexibility in generated content proved challenging, as did balancing real-time responsiveness with thorough quality auditing.

5. Accomplishments We're Proud Of

We achieved true zero-touch autonomy—the entire pipeline from stress signal detection to video generation runs without human intervention. Our system grounds every intervention in clinical frameworks (CBT/DBT) and includes self-improvement capabilities that learn from past outcomes to refine future content.

6. What We Learned

Building autonomous multi-agent systems requires meticulous orchestration and observability—our comprehensive trace logging system was essential for debugging. We learned that combining multiple specialized APIs (web scraping, knowledge bases, generative AI, emotional validation) can create emergent capabilities far beyond what each tool offers individually.

7. What's Next

We plan to expand monitoring to more platforms beyond Reddit, integrate real-time sentiment analysis for faster crisis detection, and add A/B testing for intervention effectiveness. Long-term, we envision partnerships with mental health organizations to deploy Eirene as a first-response system that triages users to appropriate resources while providing immediate support.