goal: Jan Server is Functionally Ready for Enterprise Self-Hosting

[LazyYuuki]

🎯 Goal

Deliver a functionally enterprise-ready Jan Server that businesses can self-host on-premise (with focus on SMEs first).

📖 Context

• Provides core enterprise-like features: model routing, usage tracking, user/admin management, serving infrastructure, and model lifecycle management.

• There are two main users of this product:

  1. The LLM Consumer => they don't care how or what service, they just know they login to chat.jan.ai or use api.jan.ai key, they can chat with a model (cheap and fast)
  2. The Server Admin => they set up all the model routing and connection (e.g other Remote Provider, vLLM, llama.cpp, ...), can track cost and usage, and API key provision to LLM consumer

  For a home server user, both are the same person.
  For a business, LLM consumer is the office worker and Server Admin is the IT department

• ⚠️ Not yet production-grade for security audits and compliance certifications — targeted at SMEs for first adoption.

✅ Scope

• Model Routing (text only): OpenRouter + self-hosted (Jan-v1, Qwen3, etc.)
• Usage Tracking: Per-user interaction monitoring and cost attribution
• User Management & Admin: Back Office UI, centralized API key management, conversation history retention, billing & quota enforcement
• Serving Infrastructure: Queuing, autoscaling, batching
• Model Management: CRUD, repository pulls, start/stop/scale (e.g https://web.archive.org/web/20240326183045/https://jan.ai/api-reference/#description/introduction)

❌ Out of Scope

• Image model support
• Payment gateway integration
• SaaS offering

🛠 Deliverables

• Deployable Jan Server with features above
• Basic admin UI and documentation
• Foundation for future security/compliance work

📊 Success Criteria

• SMEs can reliably self-host Jan Server with admin controls, usage tracking, and scalable serving.
• Core enterprise functions work, but compliance/security hardening is deferred.

❓ Open Question