LLM integration

[zmsdev]

ZMS now supports multiple Large Language Model (LLM) providers through an abstract interface. This allows you to use different AI backends including OpenAI, local Ollama deployments, and RAG (Retrieval-Augmented Generation) with Qdrant vector database.

1. OpenAI - Cloud-based GPT models
2. Ollama - Local LLM deployment
3. RAG - Retrieval-Augmented Generation with Qdrant vector database

The interface is located in the new module: src/Products/zms/llmapi.py and it's zmi/gui in src/Products/zms/zpt/object/manage_llm.zpt. For testing on a local system a simple docker-file is added : src/docker/addons/llm/docker-compose.llm.yml

Architecture

User Interface (manage_llm.zpt)
         ↓
REST API (rest_api.py)
         ↓
LLM API (llmapi.py)
         ↓
Provider Factory (_get_provider)
         ↓
    ┌────┴───┬─────────┐
    ↓        ↓         ↓
OpenAI    Ollama      RAG
Provider  Provider    Provider

Configuration-Guide: https://github.com/zms-publishing/ZMS/blob/fb_chatgpt/docs/llm_configuration.md

[zmsdev]

Implementation ready for test.
Unfortunately any openapi.api.key returns error:

[insufficient_quota] You exceeded your current quota, please check your plan and billing details. For more information on this error, read the docs: https://platform.openai.com/docs/guides/error-codes/api-errors.