Ecostream

A voice-driven AI navigator using Confluent 'Data in Motion' to guide sustainable local shopping, streaming real-time carbon-savings analytics to solve SDG 12.

Comment

Inspiration

Growing up in Nigeria, I witnessed firsthand how difficult it is for citizens in mega-cities like Lagos to shop sustainably. While everyone wants to reduce their carbon footprint, finding local, eco-friendly markets is a manual, frustrating task. I wanted to build a bridge between global sustainability goals (SDG 12) and local action. EcoStream was inspired by the idea that "Green Living" should be as easy as having a conversation.

How confluence was used

  • Streaming: Confluent Cloud acts as the central nervous system, handling "Data in Motion" via the market-activity Kafka topic.

ENV files

We included a zip file in our submission which includes our env and also data dog hardware requirements.

What it does

EcoStream is a voice-driven AI navigator that helps users find sustainable local markets in real-time.

  1. Natural Interaction: Users speak to the "Eco-Guide" naturally. Their voice is processed by Google Cloud Speech-to-Text, optimized for Nigerian English.
  2. AI Intelligence: Gemini 2.5 Flash (via Vertex AI) analyzes the request to find the best sustainable market and calculates potential CO2 savings.
  3. Carbon Transparency: The app provides immediate visual and audio feedback (via ElevenLabs) on the impact of the user's choice (e.g., "0.5kg CO2 saved").
  4. Data in Motion: Behind the scenes, every eco-action is streamed in real-time via Confluent Cloud to provide a live "Sustainability Pulse" of the city.

How we built it

We architected a multi-partner cloud ecosystem on Google Cloud:

  • Voice Input: Google Cloud Speech-to-Text for high-accuracy transcription.
  • Brain: Vertex AI (Gemini 2.5 Flash) for high-speed reasoning and JSON-structured responses.
  • Voice Output: ElevenLabs (Rachel) providing human-like, encouraging feedback using the latest v2.27.0 Python SDK.
  • Voice Output 2: ** Eleven labs agent widget(>30 lang) specific for only conversing about ecostream matters and reduction of carbon foot print
  • Streaming: Confluent Cloud acts as the central nervous system, handling "Data in Motion" via the market-activity Kafka topic.
  • Warehouse: BigQuery for long-term persistence and analytics of urban shopping patterns.
  • Observability: Datadog (US5) for full-stack monitoring, using custom business events and automated "Actionable Cases" for incident response.
  • Map: Google Maps JavaScript API with modern Vector Maps and Advanced Markers.
  • Identity: Google OAuth 2.0 for secure, single-click entry.
  • Deployment: Containerized Python/Flask backend hosted on Google Cloud Run.

Challenges we ran into

The biggest challenge was navigating regional identity verification hurdles for Google Cloud billing in Nigeria, which we overcame to activate the full Vertex AI suite. Technically, we faced difficulties with ElevenLabs SDK version conflicts and capturing streaming audio bytes in a serverless environment. We also solved the challenge of configuring "Agentless" Datadog tracing by routing telemetry directly to the Datadog US5 cloud intake.

Accomplishments that we're proud of

We are incredibly proud of achieving a "Triple-Sync" architecture. Seeing a user's voice command in Nigeria travel through Google STT and Gemini, trigger a real-time message in the Confluent Cloud in the US, and appear in a Datadog dashboard instantly is a massive technical milestone. We also successfully integrated 10 different Google Cloud products into a single, cohesive application.

What we learned

We learned the true power of Observability—that monitoring isn't just about "is the server up," but about tracking "AI Business Logic." We also mastered the concept of Data in Motion, realizing that sustainability data is most valuable when it is streamed live rather than sitting "at rest" in a database.

What's next for EcoStream

The next phase is to partner with Lagos city planners to use the Confluent stream for urban development insights. We plan to expand the AI's vision capabilities to allow users to scan "Carbon Receipts" and to integrate ElevenLabs Agents for even more complex, multi-turn sustainability coaching sessions.

Built With

Share this project:

Updates