Echo: Designing a Context-Aware Local AI Assistant
An inside look at how we built a zero-retention, context-aware AI assistant that lives directly in your workflow.
Elena Rostova
AI Architect
Raw calculations can sometimes feel abstract. A number like $1,240/mo or 24.3 BMI has value, but what does it mean for your life? That's why we built Echo — a local assistant that gives context to your figures.
Context Over Computation
Most AI integrations are glorified chatbots that require you to copy-paste data back and forth. We designed Echo to be context-aware: it automatically reads the outputs of your active calculator widget. If you are calculating compound interest, Echo already knows the principal, rate, and term, letting you ask questions like "Is this a good rate compared to historical inflation?" directly.
Building for Zero Retention
Integrating AI while maintaining our privacy-first guarantee was a massive engineering challenge. Traditional APIs log prompt histories for training. We implemented a multi-layered privacy approach:
- API Proxying: We pass requests through a stateless gateway that strips all identifying headers, cookies, and IP addresses.
- No Databases: The assistant is completely transient. The moment you close the chat tab, the conversation history is wiped from the browser's temporary state. Nothing is stored on any server.
- Explicit Consent: Echo never reads inputs unless you actively click the chat button, maintaining clear boundaries between your workflow and the AI.
Enhancing the Experience
Echo makes numerical sheets approachable. By bridging the gap between raw quantitative math and conversational qualitative reasoning, we're making precision tools useful for everyone, not just domain experts.
Enjoyed this read?
Get monthly updates on privacy engineering and web performance straight to your inbox.