Genesis is a multi-persona conversational AI agent I built in the last six months solo. It supports multiple LLM providers and models, has a vector memory layer (ChromaDB), STT/TTS browser UI, and a domain specific knowledge corpus.
The problem: it would answer confidently even if it was wrong. So I designed and developed Turing a secondary agent built in to flag responses as UNCERTAIN, CLEAN, or CONFIRMED before responding.
I work as an AI Operator: I direct AI to build the agent's code and I design the architecture, decisions and logic. I have no traditional coding background, just a clear vision of what I want.
This video shows Turing in action, including a case where it failed to confirm a Stoic quote which is in its knowledge base. I left it all in rather than cutting, that is the agent's behavior.
Happy to answer questions about the architecture and approach. I am new to HN and looking to connect over an AI build.
Genesis is a multi-persona conversational AI agent I built in the last six months solo. It supports multiple LLM providers and models, has a vector memory layer (ChromaDB), STT/TTS browser UI, and a domain specific knowledge corpus.
The problem: it would answer confidently even if it was wrong. So I designed and developed Turing a secondary agent built in to flag responses as UNCERTAIN, CLEAN, or CONFIRMED before responding.
I work as an AI Operator: I direct AI to build the agent's code and I design the architecture, decisions and logic. I have no traditional coding background, just a clear vision of what I want.
This video shows Turing in action, including a case where it failed to confirm a Stoic quote which is in its knowledge base. I left it all in rather than cutting, that is the agent's behavior.
Happy to answer questions about the architecture and approach. I am new to HN and looking to connect over an AI build.