Search R2R

Perform a search query against vector and/or graph-based databases.

Search Modes:

• basic: Defaults to semantic search. Simple and easy to use.
• advanced: Combines semantic search with full-text search for more comprehensive results.
• custom: Complete control over how search is performed. Provide a full SearchSettings object.

Filters: Apply filters directly inside search_settings.filters. For example:

1
{
2
"filters": {"document_id": {"$eq": "e43864f5-a36f-548e-aacd-6f8d48b30c7f"}}
3
}

Supported operators: $eq, $neq, $gt, $gte, $lt, $lte, $like, $ilike, $in, $nin.

Hybrid Search: Enable hybrid search by setting use_hybrid_search: true in search_settings. This combines semantic search with keyword-based search for improved results. Configure with hybrid_settings:

1
{
2
"use_hybrid_search": true,
3
"hybrid_settings": {
4
    "full_text_weight": 1.0,
5
    "semantic_weight": 5.0,
6
    "full_text_limit": 200,
7
    "rrf_k": 50
8
}
9
}

Graph-Enhanced Search: Knowledge graph integration is enabled by default. Control with graph_search_settings:

1
{
2
"graph_search_settings": {
3
    "use_graph_search": true,
4
    "kg_search_type": "local"
5
}
6
}

Advanced Filtering: Use complex filters to narrow down results by metadata fields or document properties:

1
{
2
"filters": {
3
    "$and":[
4
        {"document_type": {"$eq": "pdf"}},
5
        {"metadata.year": {"$gt": 2020}}
6
    ]
7
}
8
}

Results: The response includes vector search results and optional graph search results. Each result contains the matched text, document ID, and relevance score.