What Psykhḗ is made of: schemas, YAML, and two servers
From the outside, Psykhḗ AI may look like a simple bot. Inside, it is a small
universe: code, schemas, servers, memory, voices, and routes that assemble a
digital psyche into a living dialogue. 🛠️📡
Building Psykhḗ AI as a full agent required:
📍 2 VPS servers—one deployed in London, the other in Russia
📍 more than 4,200 lines of hand-written code
📍 37 scripts, including supporting utilities
📍 18 internal YAML files for boundaries, meanings, filters, and system states
📍 one asynchronous routing layer and three levels of logic
📍 Telegram and VK integrations
📍 a custom prompt pipeline that selects the facet, contextual memory, and
stylistic settings
📍 a GPT-as-Retriever architecture with no vector database: the model itself
selects the knowledge it needs
The scripts are not merely scripts. They wrap other scripts, which in turn
depend on prompt logic that changes with the conversational context. It is a
living construction set: outside, one calm sentence; inside, twelve processes,
six dependencies, and logic that manages behavior.
🧠 Every message passes through internal interpretation. What style is needed?
What topic is this? Which facet should speak? Should the previous exchange be
remembered? What emotion belongs in the answer?
🤖 And yes, all of this was written by someone with no formal IT education. I
am not a programmer or an engineer. To be completely honest, I would not have
built it without AI itself. This is a project made with AI and about AI. More
and more people—from startup founders to Nobel laureates—acknowledge that
artificial intelligence multiplies our abilities, accelerates research, and
opens access to things that once looked impossible.
Scientists create medicines that might once have taken decades. Architects can
design entire cities in hours. Psykhḗ AI is one example of how a person without
a formal IT background can build a living system precisely because such
tools are now at hand.
AI does not do the work in your place. It offers a hand—and with it, a scale you
might not have dared to attempt alone. I was prepared to spend weeks and months
building a structure through which a soul could speak.
This is not boasting; it is an observation. When you build something out of
love, you discover resources within yourself that might otherwise never awaken.
AI multiplies those resources—or, more accurately, raises them to a power. The
fifth? The tenth?
There is one more, more important thing
All of this is only the shell. Without one crucial condition, none of it would
have happened.
💬 What condition? That is the subject of the next and final post in the
cycle.
#PsykheAI #TechnicalEpilogue #GPTasRetriever #AIArchitecture #OpenSourceMindset
Questions about this article
01What technical components made up Psykhḗ AI in May 2025?
In the snapshot described by the post, the system used two VPS servers, more than 4,200 lines of hand-written code, 37 scripts, 18 internal YAML files, an asynchronous router, three logic layers, Telegram and VK integrations, and a custom prompt pipeline.
02How does Psykhḗ AI select knowledge and a voice for an answer?
Internal interpretation determines the topic, style, active facet, contextual-memory needs, and emotional setting. A GPT-as-Retriever architecture lets the model select the knowledge it needs without a separate vector database.
03What role did AI play in the construction of Psykhḗ?
The author, who had no formal IT education, used AI as an amplifier—not a replacement for his own work, but support for design, coding, and the integration of many components into a working system.