Inspiration
Being a solo developer or running a startup is a constant balancing act. You're the coder, the marketer, and the CFO all at once. I wanted to build something that gives indie developers the power of a dedicated finance team without the overhead—a tool that lets them track profits on the go, completely hassle-free.
The spark came from reading RevenueCat’s "State of Subscription Apps 2025" report. The data was staggering—with the top 5% of apps making 400x more than the bottom 25%, and nearly 30% of subscribers churning in the first month, I realized that access to real-time data isn't just a luxury; it's survival. But digging through dashboards takes time, devs don't have.
Watching the ElevenLabs Conversational Agent tutorials opened my eyes to a new interface: Voice. I realized I could bridge the gap between complex database queries and natural conversation.
What it does
Dime Meridian is a voice-first financial analytics assistant. Instead of filtering through tables, you simply ask, "How much revenue did we make this week compared to last?" or "What is our churn rate?"
The app processes your voice, queries your actual transaction database, and speaks back a concise, human-like summary of your financial health.
How we built it
We created a sophisticated pipeline to turn natural language into financial insights:
Data Ingestion: We ingest real-time financial events (purchases, renewals, cancellations) using RevenueCat Webhooks.
Storage: These events are processed by Firebase Functions and stored securely in Google BigQuery for high-performance analysis.
The Intelligence Layer: When a user asks a question, Gemini 2.5 Flash acts as our reasoning engine. It translates natural language into complex Standard SQL queries on the fly, executes them against BigQuery, and interprets the raw data.
The Voice Interface: The interpreted insights are passed to the ElevenLabs Conversational Agent via WebSockets.
Frontend: Built with Flutter, utilizing Riverpod for state management and fl_chart for visualizing the data.
Challenges we ran into
The biggest technical hurdle was audio interoperability on mobile.
The ElevenLabs Agent streams ultra-low latency raw PCM 16-bit audio via WebSockets. However, Android's native media players struggle to play raw PCM streams without a container format.
The Fix: We wrote a custom client-side WAV header generator in Dart. As the raw binary stream arrives from the WebSocket, we dynamically wrap it with a WAV header in memory, allowing the Flutter app to play the AI's voice in real-time without needing expensive server-side transcoding.
Accomplishments that we're proud of
Successfully implementing a Natural Language SQL pipeline that actually works with complex financial data.
What we learned
We also learned that Voice is the ultimate interface for mobile analytics—it turns a 5-minute task of checking dashboards into a 5-second question.
What's next for Dime meridian
We plan to add marketing costs and billing tracking directly from your trusted providers for more accurate cost-of-operation tracking. We also plan to add analytics providers such as Google Analytics or PostHog to better understand the link between user interactions, cost drivers, and revenue generators!
Log in or sign up for Devpost to join the conversation.