# TrustGraph Configuration Builder - Dialog Flow (State Machine)
# Each step declares its own transitions to the next state(s)

flow:
  id: "trustgraph-config"
  title: "TrustGraph Configuration Builder"
  version: "1.0"
  start: version

steps:

  # ─────────────────────────────────────────────────────────────
  # Version Selection
  # ─────────────────────────────────────────────────────────────
  version:
    title: "Which TrustGraph version?"
    state_key: version
    input:
      type: select
      options:
        - label: "TrustGraph 2.2"
          value: "2.2"
          description: "Additional document format support"
          badge: pre-release
          recommended: false
        - label: "TrustGraph 2.1"
          value: "2.1"
          description: "Explainability and provenance support"
          badge: stable
          recommended: true
        - label: "TrustGraph 1.9"
          value: "1.9"
          description: "Improvements to ontology and knowledge extraction prompts"
          badge: pre-release
          recommended: false
        - label: "TrustGraph 1.8"
          value: "1.8"
          description: "Pluggable messaging fabric, switched to Garage for object store"
          badge: stable
          recommended: false
    transitions:
      - next: platform

  # ─────────────────────────────────────────────────────────────
  # Platform Selection
  # ─────────────────────────────────────────────────────────────
  platform:
    title: "Which platform?"
    description: "Choose how you'll run the TrustGraph containers"
    state_key: platform
    input:
      type: select
      options:
        - label: "Docker Compose"
          value: "docker-compose"
          icon: "docker"
          description: "Easy to install on MacOS and Linux. Great for evaluation and learning. May not be well suited to production deployments."
          recommended: true
        - label: "Podman Compose"
          value: "podman-compose"
          icon: "podman"
          description: "Feature-compatible with Docker Compose, default for most Linux systems. May not be well suited to production deployments."
        - label: "Minikube"
          value: "minikube"
          icon: "minikube"
          description: "Stand-alone Kubernetes for small-to-medium deployments. Good for learning Kubernetes. Single node cluster limits production use."
        - label: "Google Kubernetes Engine (GKE)"
          value: "gke"
          icon: "google-cloud"
          description: "Managed Kubernetes on Google Cloud. Production-ready with full observability and scalability."
        - label: "AWS EKS"
          value: "eks"
          icon: "aws"
          description: "Elastic Kubernetes Service on AWS infrastructure. Suitable for production deployment."
        - label: "Azure AKS"
          value: "aks"
          icon: "azure"
          description: "Azure Kubernetes Service on Microsoft infrastructure. Suitable for production deployment."
        - label: "Scaleway Kubernetes"
          value: "scw"
          icon: "scaleway"
          description: "Managed Kubernetes on Scaleway European cloud infrastructure. Suitable for production deployment."
        - label: "OVHCloud Kubernetes"
          value: "ovh"
          icon: "ovhcloud"
          description: "Managed Kubernetes on OVHCloud European infrastructure. Suitable for production deployment."
    transitions:
      - next: graph-store

  # ─────────────────────────────────────────────────────────────
  # Graph Store
  # ─────────────────────────────────────────────────────────────
  graph-store:
    title: "Which graph store?"
    description: "Where TrustGraph stores the knowledge graph"
    state_key: graph_store
    input:
      type: select
      options:
        - label: "Apache Cassandra"
          value: "cassandra"
          icon: "cassandra"
          description: "NoSQL database with fast read/write at scale. TrustGraph integrates a graph schema."
          recommended: true
        # - label: "Neo4j"
        #   value: "neo4j"
        #   icon: "neo4j"
        #   description: "Native graph database with Cypher query language"
        # - label: "Memgraph"
        #   value: "memgraph"
        #   icon: "memgraph"
        #   description: "In-memory graph database with Cypher query language. High performance for real-time analytics."
        # - label: "FalkorDB"
        #   value: "falkordb"
        #   icon: "falkordb"
        #   description: "High-performance graph database optimized for low-latency queries. Based on GraphBLAS."
    transitions:
      - next: vector-db

  # ─────────────────────────────────────────────────────────────
  # Vector Database
  # ─────────────────────────────────────────────────────────────
  vector-db:
    title: "Which vector database?"
    description: "Stores embeddings for semantic search and RAG"
    state_key: vector_db
    input:
      type: select
      options:
        - label: "Qdrant"
          value: "qdrant"
          icon: "qdrant"
          description: "High RPS, minimal latency, fast indexing with accuracy control"
          recommended: true
        # - label: "Milvus"
        #   value: "milvus"
        #   icon: "milvus"
        #   description: "Open-source vector database built for scale"
        # - label: "Pinecone"
        #   value: "pinecone"
        #   icon: "pinecone"
        #   description: "Managed vector database service (requires API key)"
    transitions:
      - next: row-store

  # ─────────────────────────────────────────────────────────────
  # Row Store
  # ─────────────────────────────────────────────────────────────
  row-store:
    title: "Which row data store?"
    description: "Storage for row/tabular data"
    state_key: row_store
    input:
      type: select
      options:
        - label: "Apache Cassandra"
          value: "cassandra"
          icon: "cassandra"
          description: "NoSQL database with fast read/write performance at scale. Used for row storage in TrustGraph."
    transitions:
      - next: model-deployment

  # ─────────────────────────────────────────────────────────────
  # Model Deployment
  # ─────────────────────────────────────────────────────────────
  model-deployment:
    title: "How will you run the LLM?"
    description: "Choose where the language model runs"
    state_key: model_deployment
    input:
      type: select
      options:
        # Local / Self-hosted
        - label: "Ollama"
          value: "ollama"
          icon: "ollama"
          description: "Run LLMs locally. Bundles model weights, configs, and data into a single package. No external API dependencies."
        - label: "Llamafile"
          value: "llamafile"
          icon: "llamafile"
          description: "Single-file LLM distribution with embedded inference server. Runs on most computers with no dependencies."
        - label: "LM Studio"
          value: "lm-studio"
          icon: "lm-studio"
          description: "Easy local LLM running. Very usable, great support. Commercial, free for personal use."
        # Cloud API Providers
        - label: "OpenAI"
          value: "openai"
          icon: "openai"
          description: "Industry-leading models (GPT-4, etc). Requires API key."
        - label: "Claude (Anthropic)"
          value: "claude"
          icon: "anthropic"
          description: "Helpful, honest, and harmless AI models. Designed to be trustworthy and reliable. Requires API key."
        - label: "Mistral"
          value: "mistral"
          icon: "mistral"
          description: "Efficient models excelling in multilingual tasks, code generation, and reasoning. Strong performance relative to size. Requires API key."
        - label: "Cohere"
          value: "cohere"
          icon: "cohere"
          description: "Cloud-agnostic AI with high security, privacy, and customization. Supports on-premises and private cloud. Requires API key."
        # Cloud Platform Services
        - label: "Azure AI"
          value: "azure"
          icon: "azure"
          description: "Azure Endpoint Services for building, deploying, and managing AI applications. Requires Azure subscription."
        - label: "Azure OpenAI"
          value: "azure-openai"
          icon: "azure"
          description: "OpenAI models hosted on Azure infrastructure. Requires Azure subscription."
        - label: "Amazon Bedrock"
          value: "bedrock"
          icon: "aws"
          description: "Fully managed service for generative AI on AWS. Multiple model providers available. Requires AWS account."
        - label: "Google AI Studio"
          value: "google-ai-studio"
          icon: "google"
          description: "Integrated environment to prototype and experiment with Google's generative AI models. Requires Google account."
        - label: "Vertex AI"
          value: "vertex-ai"
          icon: "google-cloud"
          description: "Google Cloud platform for training and deploying ML models. Requires GCP account."
        # Self-hosted inference servers
        - label: "vLLM"
          value: "vllm"
          icon: "vllm"
          description: "High-throughput LLM serving engine. Self-hosted with OpenAI-compatible API."
        - label: "Text Generation Inference (TGI)"
          value: "tgi"
          icon: "huggingface"
          description: "Hugging Face's production-ready inference server. Self-hosted with high performance."
    transitions:
      - next: max-output-tokens

  # ─────────────────────────────────────────────────────────────
  # Max Output Tokens
  # ─────────────────────────────────────────────────────────────
  max-output-tokens:
    title: "Maximum output tokens?"
    description: "Limits the length of LLM responses"
    state_key: max_output_tokens
    input:
      type: number
      default: 4096
      min: 256
      max: 65536
      step: 256
    transitions:
      - next: ocr-enabled

  # ─────────────────────────────────────────────────────────────
  # OCR Pipelines Toggle
  # ─────────────────────────────────────────────────────────────
  ocr-enabled:
    title: "Enable OCR processing?"
    description: "Replace standard PDF decoding with Optical Character Recognition for scanned documents and images"
    state_key: ocr.enabled
    input:
      type: toggle
      default: false
    transitions:
      - when: "ocr.enabled = true"
        next: ocr-engine
      - next: embeddings-enabled

  # ─────────────────────────────────────────────────────────────
  # OCR Engine (conditional)
  # ─────────────────────────────────────────────────────────────
  ocr-engine:
    title: "Which OCR engine?"
    state_key: ocr.engine
    input:
      type: select
      options:
        - label: "PDF Decode"
          value: "pdf-decode"
          description: "Default configuration. Extracts text from PDFs with structured text, but does not perform OCR on images or scanned documents."
          recommended: true
        - label: "Tesseract"
          value: "tesseract"
          description: "Free, open-source embedded OCR engine. Best-in-class for free/open-source. Use with PDF documents containing image scans."
        - label: "Mistral"
          value: "mistral"
          description: "Best-in-class commercial OCR service. Requires a Mistral subscription. Use with PDF documents containing image scans."
    transitions:
      - next: embeddings-enabled

  # ─────────────────────────────────────────────────────────────
  # Embeddings Configuration Toggle
  # ─────────────────────────────────────────────────────────────
  embeddings-enabled:
    title: "Configure embeddings engine?"
    description: "Customize the embeddings model used for semantic search"
    state_key: embeddings.enabled
    input:
      type: toggle
      default: false
    transitions:
      # Version gate: embeddings config not available before 1.6.0
      - when: "version < '1.6.0'"
        next: review
      - when: "embeddings.enabled = true"
        next: embeddings-engine
      - next: review

  # ─────────────────────────────────────────────────────────────
  # Embeddings Engine (conditional)
  # ─────────────────────────────────────────────────────────────
  embeddings-engine:
    title: "Which embeddings engine?"
    state_key: embeddings.engine
    input:
      type: select
      options:
        - label: "FastEmbed"
          value: "fastembed"
          icon: "fastembed"
          description: "Lightweight, fast Python library for embedding generation. Small container image and quick start time."
          recommended: true
        - label: "HuggingFace sentence-transformers"
          value: "huggingface-sentence-transformers"
          icon: "huggingface"
          description: "Support for a large number of open/community models. Use for non-standard models. Larger container image and longer start time due to PyTorch dependencies."
        - label: "Ollama"
          value: "ollama"
          icon: "ollama"
          description: "Use Ollama for embeddings. Requires Ollama to be running with an embedding model loaded."
    transitions:
      - next: review

  # ─────────────────────────────────────────────────────────────
  # Review & Generate (terminal)
  # ─────────────────────────────────────────────────────────────
  review:
    title: "Review your configuration"
    type: review
    actions:
      - id: generate
        label: "Generate"
        icon: "rocket"
        description: "Generate the configuration package as a ZIP file"
    transitions: []  # terminal state