Hi, I'm Psyche 👋

Self-hosted AI builder · Agent workflow operator · Web / CLI / Telegram automation

I build AI systems that can actually live inside real workflows: self-hosted workspaces, Codex bridges, token accounting services, and small infrastructure pieces that make agent execution easier to run and maintain.

Public Projects

• Psyche: a self-hosted AI workspace for multi-provider chat, provider and model management, attachments, and production-ready deployment.
• teledex: a lightweight Telegram bridge that turns persistent Codex sessions into a practical remote workflow with bound directories and live progress previews.
• token-account: a persistent token usage service for Codex sessions with incremental sync, multi-device aggregation, and a live dashboard.
• codex-bark-notify-hook: a small notification adapter that turns Codex notify callbacks into Bark pushes or webhook relays with deduplication and compressed summaries.

What I'm Doing

• Moving AI from "can answer" to "can plug into real workflows and keep operating"
• Building self-hosted setups where models, providers, data, and deployment all stay under control
• Connecting Web, CLI, and Telegram entry points so the same agent capability can be reused across contexts
• Using lightweight systems to validate product direction quickly, then hardening what proves useful

Focus Areas

• Self-hosted AI products and multi-model integration
• Codex and agent-oriented engineering workflows
• Telegram, Web, and CLI automation bridges
• Service delivery with Fastify and FastAPI
• Token accounting, observability, and usage analytics
• Quant research and tooling infrastructure

Tech Stack

• Frontend: React, TypeScript, Vite, Tailwind CSS
• Backend: Fastify, FastAPI, Node.js, Python
• Data and deployment: PostgreSQL, SQLite, Docker, PM2, Nginx
• AI integration: Codex, OpenAI-compatible APIs, Anthropic-compatible workflows

GitHub Activity

Open To Talk About

• How to turn self-hosted AI apps into maintainable products
• How to plug Codex and agents into real engineering workflows
• Remote control patterns, automation bridges, and observability design
• The tradeoffs between fast MVPs and long-term maintainability

---

I would rather ship a working system first, then refine it into something worth maintaining for the long run.