Launched this week
Akasha
AI-native e-commerce with a Gemini assistant that knows your
14 followers
AI-native e-commerce with a Gemini assistant that knows your
14 followers
Akasha is a production-ready e-commerce platform built with Next.js 14, NestJS, and PostgreSQL — powered by a Gemini AI shopping assistant that understands your entire catalog through semantic search and a real-time knowledge graph. Unlike generic templates, every Akasha store is a fully customized, single-tenant deployment on Google Cloud Platform (Cloud Run, CDN, WAF, load balancer). You own the code, you customize freely.
Owning the code is a big deal, so it'd be great to see a built-in visual editor or admin UI for non-developers to make quick content changes without touching the codebase.
@ferhat186901 Thank you very much for the feedback. I will post images of the admin panel; all site information can be modified. Although it is not possible to edit the frontend directly for now, users can switch the color palette between light and dark modes.
You own the code and get single-tenant deployment on Cloud Run, that part alone is a huge win compared to template shops. The Gemini-powered semantic search actually seems to understand my weird product queries surprisingly well.
@seldaetinit9pb Thank you very much for the feedback; that is exactly what motivated me to create this solution. There is a solution where the client owns the code, but it is built on Magento—which, unfortunately, has flaws and is difficult to maintain (I’ve worked with it myself). So, I thought: why not create a solution using NestJS with a cutting-edge architecture? As for AI, I didn't find any existing solutions with native AI capabilities, which led me to develop this one using LightRAG, Neo4j, Graphiti, and Gemini. Thanks for the comment.
Love that it's single-tenant and you own the code, that's a big differentiator. One thing that would make this even more practical for me would be built-in support for headless CMS integrations like Sanity or Contentful out of the box, so non-developer team members could actually manage landing pages and blog content without needing a deploy every time.
@nazlvkk Thank you very much for the tip. Currently, the system has an administrative panel, but you gave me an excellent idea to use already known CMSs. I also liked the single rental model, where the client has full ownership, and it's much better not having to pay subscriptions and platform fees, plus the client gets a much higher quality.
The Gemini assistant actually understood weird long-tail queries from my catalog, which surprised me for an out-of-the-box setup. Liking that each store gets its own deployment instead of sharing some tenant soup.
@hlyasoyserwd24 Sorry, but I just noticed that my Google AI Studio credits (tokens) have run out; I’m going to make a payment so you can test our chat again. The response turned out really well—I created an exceptional RAG architecture. The project I published uses a multi-tenant and multi-frontend model approach within a monorepo. However, I won't be enabling subscriptions in the system just yet, as I still need to finalize details regarding security and scalability. Currently, I am working on a full implementation for a client—who has access to the source code for any desired modifications—at a higher price point, although I do plan to adopt a SaaS model in the future as well. Thank you very much for the feedback.