Integration with Third-party LLM providers

Overview

The following diagram shows how the Luna WMS integrates with third-party LLM/AI providers.

graph TD
  %% System 1: Source
  subgraph Luna_WMS [Luna WMS Infrastructure]
    A[Luna Core App Engine]
  end

  %% System 2: Universal Abstraction Layer
  subgraph Rhea_Framework [Rhea Generative Framework]
    B[rhea-api-explorer-f3cd1f.gitlab.io]
    C{Provider Router / Orchestrator}
    B <--> C
  end

  %% System 3: The AI Ecosystem
  subgraph AI_Providers [AI / LLM Processing Layer]
    subgraph Latent_Loom [Latent Loom Engine]
      D[rhea-latent-loom-api-explorer-b6d55f.gitlab.io]
      D_Embed[Vector Embedding Engine]
      D_Infer[Text Inference Generator]
      D --- D_Embed
      D --- D_Infer
    end

    E[External LLMs <br/> OpenAI / Anthropic / Gemini / Ollama]
  end

  %% System 4: Vector Storage
  subgraph Vector_Storage [Vector Infrastructure]
    F[(Vector Database)]
  end

  %% Data Flow Connections
  A -->|1. Universal Payload <br/> Prompt, Text, or Context| B

  %% Rhea Routing to Latent Loom (Dual Capabilities)
  C -->|2a. Inference or Vector Task| D
  D_Embed <-->|3a. Index / Query| F

  %% Rhea Routing to External
  C -->|2b. Inference Request| E

  %% Returns to Rhea
  D -->|4a. Vectors OR Generated Text| C
  E -->|4b. Generated Text Response| C

  B -->|5. Unified JSON Response| A

Any API that supports standard Chat Completion and Responses object types would be compatible ¹.

---

1. Minor tweaks maybe required. ↩