Hey Product Hunt community! 👋 I’m Dmitry, the solo engineer behind Kivara AI. I love interactive fiction and roleplay platforms, but like many of you, I got frustrated with the notorious "goldfish memory" problem in modern LLM chats: after 10–15 messages, characters lose track of the plot, forget earlier decisions, and break the entire immersion. I built Kivara AI to solve this from the ground up: 🧠 Persistent RAG Memory: Using PostgreSQL with pgvector and semantic retrieval, characters recall past details without bloating token limits. 👥 Multi-Character Groups & Branches: Orchestrate dialogues between multiple distinct personas or fork conversations to explore "what if" scenarios. 🎨 Visual Generation: Scene artwork generated directly from the current dialogue context. 🌐 Accessible Everywhere: Responsive PWA, built-in Telegram Mini App, and support for 5 languages. The project is free to start. I’d love for you to try it out, create a character, and test how well they retain your plot twists! I’ll be here all day to answer questions about our memory pipeline, prompt engineering, or anything else. Looking forward to your honest feedback! 🚀

Kivara AI

Interactive AI roleplay & stories with persistent memory

Interactive AI roleplay & stories with persistent memory