lo6
The Ableton Live for News — a Newsroom Operating System that orchestrates AI agents for sourcing, research, and drafting while keeping journalists in creative control.
Concept Preview
A static preview of the “Editorial Avant-Garde” design system: dark mode canvas, Playfair Display serif headlines, 50/50 split-pane Reading Room with inline source provenance, floating action bars, and Focus Mode (press Esc). Built from docs/frontend_spec.md.
Why?
Journalism doesn’t need another CMS — it needs a Studio. Current AI writing tools are black boxes that hallucinate. lo6 uses a Human-in-the-Loop architecture with strict provenance: every claim is visually linked to a verified source.
Features
- Newsroom Engine — LangGraph state machine: Lead → Triage → Research → Draft → Review → Publish
- Newsroom Console — 5 dashboard views: Lead Pipeline, Triage Board, Research Desk, Writing Studio, Editorial Review
- Signal Exchange — Federated network for sharing verified leads between newsroom instances
- Source Provenance — Every AI-generated claim is traced to its source, no black boxes
- Security-first — OWASP/NIST/ISO 27001 compliant. MFA, RBAC, audit logging, Zod validation
Tech Stack
| Layer | Technology |
|---|---|
| Framework | Next.js 14+ (App Router) |
| Language | TypeScript (strict mode) |
| AI Orchestration | LangChain.js + LangGraph.js |
| Database | PostgreSQL (Supabase/Neon) + Prisma ORM |
| Styling | TailwindCSS + Shadcn/UI |
| State | TanStack Query + Zustand |
| Auth | Supabase Auth (MFA) |
| Queue | BullMQ + Redis (optional, for long-running agents) |
Project Status
Phase 1: Design Complete. Architecture, data model, agent specs, frontend spec, security spec, and funding plan are all documented in /docs. Implementation starts now.
Quick Start
# Clone and install
git clone https://github.com/juanmanueldaza/lo6.git
cd lo6
npm install
# Copy environment template
cp .env.example .env.local
# Run development server
npm run dev
Documentation
| Doc | Description |
|---|---|
| Implementation Plan | Architecture, tech stack, core components |
| Data Model | PostgreSQL schema, Prisma types |
| Agent Interface Spec | LangGraph workflows, agent tools & prompts |
| Frontend Spec | UX philosophy, component hierarchy, dashboards |
| Security Spec | OWASP Top 10 mitigation, RBAC, audit logging |
| Security Snippets | Copy-pasteable CSP, SSRF, RLS patterns |
| Accessibility Appendix | WCAG 2.2 AA compliance |
| Pitch Executive Summary | Investor-facing one-pager |
| Funding Plan | Grants, consortium, agency, subscription model |
Contributing
We welcome contributions from humans AND AI agents. See CONTRIBUTING.md for setup instructions, coding standards, and how to claim issues.
If you’re using an AI coding assistant (Claude Code, Copilot, Codex, Aider, etc.), read AGENTS.md for project-specific guidance.
Agentic Development (N3RV Framework)
This project uses N3RV — invisible engineering infrastructure for AI agents. N3RV provides:
- SDD Workflow: 8-phase Spec-Driven Development pipeline (
/sdd-new) - A2A Hub: Agent-to-agent task delegation via JSON-RPC 2.0
- MAGI Memory: Persistent semantic memory (ChromaDB + SQLite dual-store)
- OpenCode Integration: 9 specialized agents, 14 skills, and slash commands
# Install N3RV (one-time)
git clone https://github.com/juanmanueldaza/n3rv.git ~/n3rv
cd ~/n3rv && uv tool install .
# Initialize in this project
n3rv init
See .opencode/agents/n3rv.md for available commands, skills, and SDD agents.
License
MIT © Juan Manuel Daza