~/projects/community ▸ Selected Professional Work · 2026
CORPORATELLMCOMMUNITY_
Community platform for Corporate LLM. Teams discover, share and rate agents, prompts, skills and AI models in a secure, observable application.
// My contribution
From product concept to a production-ready system.
My focus covered the Community end-to-end — from visible product flows to data and security boundaries, observability and delivery.
- Role
- End-to-end software development
- Scope
- Product flows, frontend, data and operations
- Status
- Live · public beta
Starting point
Corporate LLM teams needed one place for agents, prompts, skills, models and discussions — without a duplicate account system or a pre-publication approval queue.
Key decisions
- 01Next.js App Router with clear boundaries between Server Components, Route Handlers and interactive client surfaces
- 02The Corporate LLM account as the single identity source instead of a second authentication system
- 03Publish-first moderation with Row Level Security and observable failure and product states
Outcome
A live, publicly browsable Community with feed, discovery, models, search and account handoff. Sentry, PostHog and controlled release gates make ongoing operations measurable and debuggable.
// Screenshots
// Product Surfaces
Community Feed
A shared feed combines agents, prompts and skills with discussions, model reviews and announcements. Filters, sorting and reactions keep different content types easy to compare.
Discovery and Search
Catalogues and server-side search cover content, models and profiles. Categories, contribution types and visible match context keep larger result sets understandable.
Publishing and Interaction
Composer flows support multiple contribution types. Reactions, comments, bookmarks and sharing provide consistent feedback across the main product surfaces.
Model Reviews
The model catalogue separates general experience from usage in RelationFlow. Provider filters and contextual scopes support transparent model selection.
Account and Moderation
Sign-in uses the existing Corporate LLM account. Contributions publish community-wide, while reporting and moderation provide a transparent post-publication safety layer.
Production Operations
Sentry exposes failures with useful technical context. PostHog records relevant product states without unnecessary personal data and supports continuous improvement.
// Tech Stack
// Architecture Highlights
Next.js App Router with Clear Server Boundaries
Server Components, Route Handlers and focused client islands keep data access, interaction and rendering responsibilities clearly separated.
PostgreSQL and Row Level Security
Supabase combines auth, PostgreSQL, storage and Edge Functions. Migration-first changes and RLS policies protect data access at the database layer.
Corporate LLM as the Identity Source
The Community does not duplicate account management. A defined handoff connects login and session state to the existing RelationFlow account.
Observability without PII
Sentry captures failures with actionable debug context. PostHog records product and failure states behind consent boundaries without copying sensitive content into telemetry.
Release and Quality Gates
Linting, type checks, tests, migration evidence and controlled promotions protect changes from feature development through release.