Sage AI

Hey Indie Hackers! 👋 I’m Yashwanth, co-founder of Advent AI.

After spending time in the FAANG world, my co-founder and I noticed a glaring problem in how businesses were adopting AI: most of the "assistants" hitting the market were just glorified FAQ search bars.

They were entirely passive. If a customer had a real issue, or an internal team needed to trigger a process, the bot would just drop a link to a knowledge base and say, "Here's an article that might help." We realized the market didn't need another chat interface; it needed an Active AI Agent that actually executes tasks. So, we built Sage.

What is Sage? Sage is a reasoning-native AI platform designed to replace passive support with specialized agents that handle end-to-end workflows.

Initially, we explored strictly autonomous e-commerce, but we quickly realized the need for specialized chatbots spans everywhere. Now, Sage deploys universally via adapters to web storefronts, WhatsApp, Slack, and Microsoft Teams.

How we built it (The Tech Stack): We wanted Sage to feel like a concierge with perfect memory, which meant the architecture had to be incredibly fast and capable of holding cross-context state.

• The Brain: We are leveraging Claude for its deep contextual reasoning and strict brand-safety guardrails, alongside Gemini’s native multimodal capabilities (so users can literally upload a picture of a broken cable and the bot understands it).

• The Memory: We use Zilliz as our vector database to power the "Customer Memory Engine." This allows Sage to instantly recall a user's past purchases, known bugs, or preferences, even if they switch from a web chat to WhatsApp.

• The Skeleton: The backend is powered by FastAPI for high-performance, asynchronous multi-agent communication, with a dynamic React interface for the merchant dashboards.

Building an agent that actually does things—like processing a return or updating a database directly from chat—was a massive technical hurdle, but getting backed by the NVIDIA Inception Program and AWS Activate gave us the infrastructure runway to pull it off.

I’d love to hear from the community: What is the most tedious operational or support workflow you are currently dealing with that you wish an active agent could just execute for you? Happy to answer any questions about our stack, our pivot to omnichannel, or building AI out of Bengaluru!