PeacePulse

Inspiration

Having worked with Mercy Corps and the Interfaith Mediation Center, I’ve been deeply involved in peacebuilding at the grassroots level. I later served at the Emergency Communication Center in Katsina, Nigeria, where I responded to thousands of urgent reports from citizens.

That experience showed me the critical gap between community voices and quick response — and the need for a smart, automated, and real-time early warning system.

What It Does

PeacePulse is an AI-powered conflict reporting and alerting agent. It:

• Accepts incident reports via voice or text (from local communities)
• Transcribes voice to text using Whisper AI
• Stores and indexes all reports in TiDB Cloud with vector and full-text search
• Uses OpenAI to summarize incidents and detect risk signals
• Sends alerts via Twilio or Mailgun to peace actors and mediators

How I Built It

• I created a Flask-based backend to handle incoming reports
• Integrated Whisper API for speech-to-text
• Stored data in TiDB Serverless Cloud and used its vector search features
• Used OpenAI’s GPT API for summarization and risk tagging
• Set up Twilio for SMS alerts and Mailgun for emails
• Connected everything into a simple front-end with optional WhatsApp access

Challenges I Faced

• Getting accurate transcription of low-quality or local-accented voice recordings
• Setting up vector search for context-aware retrieval in TiDB
• Designing a useful yet lightweight UI for users with low digital literacy
• Keeping the flow responsive despite API dependencies

What I Learned

• How to use TiDB Cloud’s vector and full-text indexing for real-time search
• Chaining AI models with external triggers like SMS
• Building AI agents with practical, real-world humanitarian impact
• That even small tools can save lives when designed for the right context

Built With

• css
• flask
• github
• html
• mailgun
• openai-gpt
• python
• render
• rest
• tidb-cloud
• twilio
• whisper-ai