Refactor Headless Backend

[louis-menlo]

Description

Refactor the Jan business/backend logic to Cortex, a headless backend platform equivalent to OpenAI. The UI components (Desktop/Mobile/Web) should seamlessly integrate with Cortex endpoints, ensuring that server logics such as file system (FS) and process are decoupled from UI components.

Key results

• Cortex serves as a headless backend platform, supporting multiple users and offering full equivalence to OpenAI.
• Cortex can be deployed via npm repository or Kubernetes (k8s) / Docker.
• There should be no strict external service dependencies; every integration should be injectable.
• The Cortex Backend serves high-level business services and should be adaptable to new structure
• Jan UI applications seamlessly interact with Cortex endpoints and can be readily scaled to multiple platforms.

Subtasks

• janhq/cortex#550
• janhq/cortex#579
• Headless Backend Extensions #2782
• Refactor Inference Engine extensions to Backend #2783
• janhq/cortex#548
• NPM Publish
• feat: Get local API server mode to also work for Nitro-TRT-LLM #2373
• feat: Configurable EOS token when using the server #2758
• janhq/cortex#682
• janhq/cortex#629

Out of scope:

• janhq/cortex#551
• feat: deprecate core's Node.js module  #2823
• janhq/cortex#676
• janhq/cortex#580
• ...

[louis-menlo]

Scoped out inactive actions