Zy / AskLibra Assistant
A modular personal AI assistant architecture that connects conversational interaction with persistent context and selected AskLibra capabilities while keeping business logic and platform services behind server-side boundaries.
Role
Product Strategist, AI Systems Architect, AI-Assisted Full-Stack Developer
Platform
Web + Desktop Ecosystem
Status
In Development
Year
2026
Private personal assistant environment
The case study describes the architecture and product model without exposing personal memories, credentials, private conversations, or internal system secrets.
Overview
Zy is designed as the conversational and orchestration layer of the AskLibra ecosystem rather than another isolated chatbot. The system is intended to understand the user's context, retain structured information, and connect that context to projects, decisions, artifacts, tools, and selected platform capabilities.
The assistant begins as a modular monolith with a clear path toward a shared context layer that can eventually serve multiple AskLibra interfaces.
Problem
A useful personal AI needs more than chat history. It needs durable context, identity, project state, decisions, and controlled access to tools.
- Context becomes fragmented across conversations
- Important decisions are difficult to retrieve later
- Projects and tasks need persistent relationships
- AI tools often lack a durable user model
- Actions require permission and approval boundaries
- Web and desktop interfaces should share the same intelligence layer
- Private context must remain separate from public product data
Solution
Zy uses a structured context model around identity, profiles, projects, memories, decisions, artifacts, events, and permissions. Conversations become one interface into that context rather than the context itself.
The architecture keeps the assistant runtime behind server-side boundaries and exposes capabilities explicitly, allowing future tools and AskLibra applications to be connected without giving the agent unrestricted access to the platform.
Core assistant workflow
The assistant assembles context before reasoning and keeps tool execution behind explicit capabilities.
- 1
Authenticate
Verify the AskLibra identity and establish the user's assistant profile.
- 2
Assemble context
Load relevant profile, project, memory, decision, artifact, and workspace context.
- 3
Understand
Interpret the request against the available context and current conversation.
- 4
Reason
Plan an answer or action while respecting capability and permission boundaries.
- 5
Ask or act
Use an available capability when appropriate or request user approval when required.
- 6
Persist
Store durable conversations, decisions, artifacts, or events when they are meant to remain available.
System areas
Zy is organized around context, agent execution, identity, and controlled capabilities.
Identity and profile
Connects the assistant to the authenticated AskLibra user without duplicating platform identity.
- AskLibra session verification
- Personal profile
- Access policy
- SSO callback
- Session boundary
Shared context
Persistent information that can support reasoning across conversations and interfaces.
- Projects
- Memories
- Decisions
- Artifacts
- Events
- Permissions
Agent runtime
The server-side boundary where conversational reasoning and capability selection occur.
- Conversation runtime
- Context assembly
- Capability registry
- Tool boundaries
- Model integration
Ecosystem integration
Connects Zy to AskLibra capabilities without recreating the full web platform inside the assistant.
- Workspace context
- Knowledge / RAG
- Selected apps
- Automation
- Desktop extension
Gallery
Product media can be added here without changing the page renderer.
Zy Desktop Application
The desktop assistant interface designed to bring conversations, persistent context, projects, and AI capabilities into one workspace.
Zy Assistant
The web-based assistant login interface designed to connect the user to their personal AI assistant and shared context layer.
Zy Assistant Chat
The web-based chat workspace for interacting with Zy, managing conversations, and working with persistent assistant context.
My role
I defined the assistant's role, architecture, context model, identity integration, capability boundary, and product direction, with AI-assisted implementation and iterative system design.
AI-assisted, human-directed development
AI is used as an implementation and reasoning-development aid, while the assistant's behavior, context model, capability boundaries, and product direction remain human-defined.
AI accelerated
- Implementation scaffolding
- Refactoring
- Debugging
- API development
- Documentation
- Test support
- Architecture exploration
Human direction and ownership
- Assistant persona
- Context architecture
- Permission model
- Capability boundaries
- Product strategy
- Workflow decisions
- Final validation
Architecture
Technology stack
Technical challenges
Persistent context without uncontrolled memory
Structured the context layer around explicit domain objects rather than treating the entire conversation history as memory.
Platform identity integration
Designed a server-side session verification boundary so the assistant can use AskLibra identity without duplicating authentication ownership.
Capability boundaries
Introduced a capability registry so the assistant can access selected functions explicitly instead of receiving unrestricted platform access.
Web and desktop separation
Kept the desktop experience focused while treating AskLibra Web as the complete platform and source of truth.
Future shared intelligence layer
Designed the context layer so multiple applications can eventually consume shared intelligence without forcing an early distributed architecture.
Quality assurance
Build integrity
- TypeScript checking
Functional validation
UI and live validation
Results
Qualitative workflow outcomes
Persistent
Conversations and structured assistant context can survive individual chat sessions.
Explicit
Tool access is modeled as a registry rather than unrestricted agent access.
Shared
The context layer is designed to support multiple AskLibra interfaces.
Web + Desktop
Zy can extend from a focused web assistant into a native desktop workflow.
What this project demonstrates
Zy establishes the foundation for a reusable personal AI system that can understand persistent context, reason across projects and decisions, and orchestrate selected AskLibra capabilities through controlled interfaces.
- Build context as structured data, not just chat history
- Keep agent capabilities explicit and permissioned
- Separate intelligence from interface
- Start as a modular monolith before extracting shared services
More work is being documented.