flow.js

// This file contains a definition of a form flow.
// It contains a list of steps and a list of transitions, with the descprion of the steps.


const elements = {
    "dense-vector-configuration": [
        {
            "type": "number-with-suggestions",
            "name": "dimensions",
            "suggestions": [
                {
                    "label": "CLIP",
                    "value": 512
                },
                {
                    "label": "openai-ai/text-embedding-3-small",
                    "value": 1536
                },
                {
                    "label": "openai-ai/text-embedding-3-large",
                    "value": 3072
                }
            ]
        },
        {
            "type": "dropdown",
            "name": "metric",
            "options": ["Cosine", "Euclid", "Dot", "Manhattan"],
            "default": "Cosine"
        },
    ],
    "sparse-vector-configuration": [
        {
            "type": "checkbox",
            "title": "Use IDF?",
            "name": "use_idf",
            "default": true,
        }
    ]
}


const steps = {
    "use-case-step": {
        // In this step user should select from 2 cards, which type of search they want to perform
        "title": "Create new collection",
        "description": "What's your use case?",
        "cards": [
            {
                "title": "Global search",
                "Description": "Search across whole collection of data with optional filters. For example: e-commerce search, website search, etc.",
                "on-select": {
                    "continue-step": "tenant-field-selection-step"
                }
            },
            {
                "title": "Multitenancy",
                "Description": "Search across multiple isolated tenants. For example: per-user documents, chat history search, organization-based isolation",
            }
        ]
    },
    "tenant-field-selection-step": {
        // In this step user should select a field that contains tenant id
        "title": "Tenant field",
        "description": "Which payload field should be used as a tenant id?",
        "long-description": "This field should be used to filter data based on tenant id. For example: user_id, organization_id, etc. Payload field should be of a `keyword` type.",
        "elements": [
            {
                "type": "string-input",
                "name": "tenant_id",
            },
            {
                "type": "button",
                "title": "Continue",
                "on-click": {
                    "continue-step": "index-field-selection-step"
                }
            }

        ],
    },
    "templates-selection-step": {
        // In this step user should select a template for the index
        "title": "What to use for search?",
        "description": "There are some common configurations used for search, maybe you want to use one of them?",
        "cards": [
            {
                "title": "Simple Single embedding",
                "description": "Simplest configuration, only one vector field per document.",
                "on-select": {
                    "continue-step": "simple-dense-embedding-step"
                }
            },
            {
                "title": "Simple Hybrid Search",
                "description": "Dense + Sparse vectors searched simultaneously. Search covers both semantic and keyword-based search.",
                "on-select": {
                    "continue-step": "simple-hybrid-searc-step"
                }
            },
            {
                "title": "Hybrid Search with Late Interaction re-ranking",
                "description": "Dense + Sparse vectors searched simultaneously. Results are combined and re-ranked using heavy multi-vector model like e.g. ColBERT",
            },
            {
                "title": "Visual Latest Interaction",
                "description": "End-to-end retrieval of PDFs, presentations, images, etc. Collection consists of two multi-vector fields: one compressed for pre-fetch and other for re-ranking.",
            },
            {
                "title": "Custom",
                "description": "You can define your own configuration.",
            }
        ]
    },
    "simple-dense-embedding-step": {
        // In this step user should select a field that contains tenant id
        "title": "Vector configuration",
        "description": "Configuration for dense embedding",
        "elements": [
            {
                "type": "string-input",
                "title": "Dense vector name",
                "name": "dense_vector_name",
                "default": "dense"
            },
            {
                "type": "dense-vector-configuration",
                "name": "dense_vector_config",
            },
            {
                "type": "string-input",
                "title": "Sparse vector name",
                "name": "sparse_vector_name",
                "default": "sparse"
            },
            {

            },
            {
                "type": "button",
                "title": "Continue",
                "on-click": {
                    "continue-step": "index-field-selection-step"
                }
            }
        ],
    },
    "simple-hybrid-searc-step": {
        // In this step user should select a field that contains tenant id
        "title": "Vector configuration",
        "description": "Configuration for dense embedding",
        "elements": [
            {
                "type": "dense-vector-configuration",
                "name": "dense_vector_config",
            },
            {
                "type": "button",
                "title": "Continue",
                "on-click": {
                    "continue-step": "index-field-selection-step"
                }
            }
        ],
    },
    "index-field-selection-step": {
        // In this step we let user specify which payload fields should be indexed.
        // User can specify as many fields as they want.
        // For each field user needs to choose which type in index they want and parameters for this index.
        "title": "Payload indexes",
        "Description": "We need to create indexes, if we want to do filtered search.",
        "elements": [
            {
                "type": "repeatable",
                "elements": [
                    {
                        "type": "string-input",
                        "name": "Field name",
                    },
                    {
                        "type": "button-group-with-inputs",
                        "enums": [
                            {
                                "name": "keyword",
                            },
                            {
                                "name": "integer",
                                "fields": [
                                    {
                                        "name": "lookup",
                                        "type": "checkbox",
                                        "default": true
                                    },
                                    {
                                        "name": "range",
                                        "type": "checkbox",
                                        "default": true
                                    }
                                ]
                            },
                            {
                                "name": "float",
                            },
                            {
                                "name": "uuid",
                            },
                            {
                                "name": "datetime",
                            },
                            {
                                "name": "text",
                                "fields": [
                                    {
                                        "name": "tokenizer",
                                        "type": "dropdown",
                                        "options": ["prefix", "whitespace", "word", "multilingual"],
                                    },
                                    {
                                        "name": "lowercase",
                                        "type": "checkbox",
                                        "default": true
                                    },
                                    {
                                        "name": "min_token_length",
                                        "type": "number",
                                        "default": null
                                    },
                                    {
                                        "name": "max_token_length",
                                        "type": "number",
                                        "default": null
                                    }
                                ]
                            },
                            {
                                "name": "geo",
                            },
                            {
                                "name": "bool"
                            }
                        ]
                    }
                ]
            },
            {
                "type": "button",
                "title": "Finish",
                "on-click": "finish"
            }
        ]
    }
}