sean.
Back to projects

The Brain

Project Year: 2026

The Brain is an extensible AI Agent platform. The system supports importing Skill files and also allows custom Provider implementations. The platform provides user registration and authentication, multi-chatroom contextual memory, as well as API key management and Skill permission control.

Tech Stack

FastAPISQLSolidJSOllamaAI AgentSkill System

Architecture

Loading architecture diagram...

  • SolidJS frontend delivers login, chat, and settings workflows with a responsive UI for daily operations.
  • FastAPI backend exposes APIs for authentication, chat orchestration, skill registry, provider execution, and API key governance.
  • SQL database persists users, sessions, API keys, skill metadata, provider configuration, and chat context history.
  • Skill runtime parses markdown skill files into structured capabilities that can be safely invoked by the agent.
  • Provider abstraction layer supports both custom provider code and local Ollama inference endpoints for flexible model operations.

Features

  • Skill MD import pipeline that turns markdown-defined capabilities into reusable agent tools.
  • Custom Provider extension model for hard-coding team workflows and domain-specific execution logic.
  • API key management with per-key skill access control for safer external integration.
  • Authentication system with user registration/login and account-level security controls.
  • Multi-room chat with persisted context to support long-running and parallel tasks.
  • Provider routing layer that can switch between Ollama local models and other model providers.
  • Unified control plane for skill configuration, provider policy, and chat workflow governance.

Outcomes

  • Built a practical AI Agent foundation that can be extended through both declarative skills and programmable providers.
  • Reduced operational risk by introducing key-level skill permissions and centralized provider governance.
  • Improved usability with account-based access plus persistent multi-room chat for real-world task continuity.

Demo Screens

Roadmap

  1. Core architecture setup with FastAPI, SQL, and SolidJS

    01

    Established frontend-backend boundaries, baseline APIs, and persistent data models for users, keys, chats, skills, and provider settings.

  2. User authentication and registration system

    02

    Implemented registration, login, and session/token flows to support secure account-level access control across all features.

  3. Skill markdown import and parsing pipeline

    03

    Built a loader to ingest Skill MD definitions and convert them into executable capability metadata for agent orchestration.

  4. Custom provider framework for workflow hardening

    04

    Designed Provider adapters so domain-specific logic can be fixed into reusable tools and customized operational processes.

  5. Multi-key API management and skill permission policy

    05

    Added API key management with per-key allowlists to control which skills can be invoked by each integration.

Other Projects