Aurora

Marketing home.

Sign in: two-step email/password or SSO.

GTM workspace: product portfolio with onboarding, search, and ready/setup filters.

The 9-stage GTM pipeline (Sources → Ingest → Strategy → … → Publish) with the run log dock.

Executive Brief: accounts, qualified leads, conversations, and GTM readiness, computed without an LLM call.

Sales Action: Discover / Qualify / Outreach / Pipeline tabs for account discovery and CSV import.

Admin: Workflow Stages, an Enterprise-plan feature for tenant-defined pipeline stages.

Customer Sources → Product Discovery → Knowledge Extraction → AI Knowledge Graph
                                                                    ↓
                    Marketing AI ← Supervisor → Sales AI → Solution AI
                                    ↓
                    Omnichannel Publishing → Analytics → Continuous Learning

The public docs site at zyvor.dev/aurora is synced from docs/customer/ via scripts/customer-docs/sync-to-website.mjs — edit the markdown here, not the live site.

Requires Docker Desktop running.

make start    # infra + DB + API + web (background)
# → http://localhost:3000  (web)
# → http://localhost:8000  (api)
# Sign in (2-step): [email protected] → Continue → Admin@321
make stop     # when done

Full install paths (containerized/production, remote K3s, sales CRM), demo logins, and a page-by-page nav table: QUICKSTART.md. Full local setup and troubleshooting: docs/dev-guide.md.

This repository is licensed under the Zyvor Production License v1.0. Non-production use is free. Production use needs a commercial license.

• Onboard a product from a website, docs, CSV, YouTube, GitHub, or an OpenAPI spec — see docs/source-management.md.
• Automatic knowledge graph (Neo4j) + RAG store (Qdrant) built from ingested sources, with grounded Q&A (POST /products/{id}/query).

• Marketing strategy and content generation, sales chat, personalized outreach (suppression-aware publish), solution-architect Q&A, and proposal generation — one LLM factory, swap Ollama ↔ OpenAI-compatible per agent. See docs/ollama-llm-integration.md.
• 11 specialized agents behind a supervisor, lean-hardware-first design. See docs/multi-agent-composition-plan.md.
• Campaign management and channel publishing across email, LinkedIn, X, Medium, dev.to, and Reddit adapters.

• Persona-based default landing after login (exec / sales / marketing). See docs/role-based-landing.md.
• Executive Brief with a 9-signal gtm_readiness status — computed without an LLM call on load.
• Admin: plan/usage, suppression list, tenant data export, tenant purge, and Enterprise-plan custom workflow stages.
• Customer / reseller / salesperson external portals.
• Optional standalone Sales CRM microservice (apps/sales-crm) — Kanban pipeline, SLA timers, round-robin owners; independent of the built-in opportunities pipeline.
• SSO/OIDC — bundled Keycloak demo IdP or bring your own IdP. See docs/sso-oidc.md.

The platform supports both Ollama and OpenAI via a single factory. Switch with env vars only:

If OPENAI_API_KEY is set and LLM_PROVIDER is unset, OpenAI is auto-selected for backward compatibility.

Per-agent model routing:

Check provider status: GET /health

Full plan, architecture, unit tests, and integration test guide: docs/ollama-llm-integration.md

Test case document (223 tests): docs/test-cases.md

Multi-agent composition plan (11 specialized agents, lean hardware / persona-first): docs/multi-agent-composition-plan.md

• POST /api/v1/auth/register — Register tenant
• POST /api/v1/products — Onboard product
• POST /api/v1/products/{id}/sources — Add source
• POST /api/v1/products/{id}/ingest — Trigger ingestion
• GET /api/v1/products/{id}/profile — Product profile
• POST /api/v1/products/{id}/query — Grounded Q&A
• POST /api/v1/products/{id}/strategy — Generate GTM strategy
• POST /api/v1/products/{id}/content — Generate content
• POST /api/v1/products/{id}/chat — Sales agent chat
• POST /api/v1/products/{id}/outreach — Personalized outreach (optional recipient_email for suppression-aware publish)
• POST /api/v1/products/{id}/architect — Solution architect Q&A
• POST /api/v1/products/{id}/proposals — Generate proposal
• POST /api/v1/artifacts/{id}/approve — Approve/reject generated content
• POST /api/v1/artifacts/{id}/publish — Publish to a channel (blocked if recipient is suppressed)
• POST /api/v1/products/{id}/campaigns, GET .../campaigns, GET .../campaigns/{id}/status — Campaign management
• GET /api/v1/products/{id}/opportunities/{opp_id} — Opportunity detail
• GET /api/v1/products/{id}/account-health — Customer success account health
• GET /api/v1/products/{id}/brief — Executive brief incl. gtm_readiness (9 derived, no-LLM status booleans — sources/ingest/profile/strategy/discover/qualify/outreach/proposal/publish)
• GET /api/v1/products/{id}/workflow-runs — Recent/active WorkflowRuns for a product (powers the Workspace run-log dock)
• GET /api/v1/agents/registry — Agent registry catalog
• GET /api/v1/admin/plan — Plan + usage
• GET/POST /api/v1/admin/suppression — Suppression list
• GET /api/v1/admin/export — Tenant data export (admin only)
• POST /api/v1/admin/purge — Tenant knowledge purge, typed-slug confirmation (admin only)
• GET/POST/PUT/DELETE /api/v1/admin/workflow-stages — Tenant-defined custom Workspace stages (Enterprise plan only)

Light-first Apple.com-style system: Apple-blue primary (--primary #0071e3), Mist Blue / Sage / Lavender / Deep Blue accents, SF/system typography, and pill CTAs in apps/web/src/app/globals.css. Dark mode is a real, explicit opt-in (toggle lives only in the app-shell nav; marketing and portal auth stay light regardless of the visitor's OS preference or the toggle's stored state) via html.dark-theme (ThemeContext). One shared Card/Badge/StatGrid/Container primitive set backs cards, status pills, metric tiles, and page-width containers app-wide — legacy .glass* class names remain as thin aliases for the same flat Apple panels (hairline border, no decorative shadow by default); the earlier .tahoe-* glass-morphism direction has been fully removed.

Chrome: GlobalNav (mega-menu flyouts) sits on marketing (MarketingLayout), app (AppShell), product console (ProductConsoleShell), and portal auth pages. Marketing home is / (HomeSections); sign-up/sign-in live at /login with the same Apple-blue auth language as PortalAuthShell. New tenants get an OnboardingChecklist on /dashboard and Workspace until sources are ingested and an agent has run.

The product workspace (/products/[id]/*) keeps a left rail + top tab bar under that same GlobalNav, built around a derived 9-stage pipeline chain. See Frontend workspace.

What's free under the Zyvor Production License vs. what needs a commercial license (full guide):

apps/sales-crm and the platform's built-in opportunities pipeline are deliberately independent services — don't assume one implies the other.

From the 12-phase implementation guide (status, acceptance criteria and test matrix):

→ zyvor.dev

Aurora is source-available under the Zyvor Production License v1.0 (SPDX LicenseRef-Zyvor-Production-1.0).

• Free for development, testing, evaluation, research, education, and non-production labs.
• Production use (production, customer workloads, SaaS, managed services, OEM, and redistribution) requires an annual enterprise subscription. Plans, support levels and terms: docs/SUBSCRIPTION-MODEL.md · licensing guide · Pricing · [email protected].

Commercial terms: https://zyvor.dev.

Contributions: CONTRIBUTING.md. Report vulnerabilities privately per SECURITY.md.