Add RedisVL integration for advanced vector search

.github/wordlist.txt

@@ -111,4 +111,13 @@ ulid
 coroutine
 compat
 programmatically
-uv
+uv
+RedisVL
+embeddings
+VectorQuery
+VectorRangeQuery
+SearchIndex
+ADHOC
+EF
+DSL
+DSL

aredis_om/redisvl.py

@@ -0,0 +1,237 @@
+"""
+RedisVL integration for Redis OM.
+
+This module provides utilities to convert Redis OM models to RedisVL schemas,
+enabling advanced vector search capabilities through RedisVL.
+
+Example:
+    from redis_om import JsonModel, Field, VectorFieldOptions
+    from aredis_om.redisvl import to_redisvl_schema, get_redisvl_index
+
+    class Document(JsonModel, index=True):
+        title: str = Field(index=True)
+        embedding: list[float] = Field(
+            vector_options=VectorFieldOptions.flat(
+                type=VectorFieldOptions.TYPE.FLOAT32,
+                dimension=384,
+                distance_metric=VectorFieldOptions.DISTANCE_METRIC.COSINE,
+            )
+        )
+
+    # Get a RedisVL IndexSchema for advanced operations
+    schema = to_redisvl_schema(Document)
+
+    # Or get a ready-to-use SearchIndex
+    index = get_redisvl_index(Document)
+    results = await index.query(VectorQuery(...))
+"""
+
+from typing import Any, Dict, List, Optional, Type, Union
+
+from redisvl.index import AsyncSearchIndex, SearchIndex
+from redisvl.schema import IndexSchema
+
+from .model.model import (
+    FieldInfo,
+    JsonModel,
+    RedisModel,
+    VectorFieldOptions,
+    get_outer_type,
+    is_numeric_type,
+    is_supported_container_type,
+    should_index_field,
+)
+
+
+def _get_field_type(
+    field_name: str,
+    field_type: Any,
+    field_info: FieldInfo,
+    is_json: bool,
+) -> Optional[Dict[str, Any]]:
+    """Convert an OM field to a RedisVL field definition."""
+    if not should_index_field(field_info):
+        return None
+
+    vector_options: Optional[VectorFieldOptions] = getattr(
+        field_info, "vector_options", None
+    )
+    sortable = getattr(field_info, "sortable", False) is True
+    full_text_search = getattr(field_info, "full_text_search", False) is True
+    case_sensitive = getattr(field_info, "case_sensitive", False) is True
+
+    # Vector field
+    if vector_options:
+        attrs = {
+            "dims": vector_options.dimension,
+            "distance_metric": vector_options.distance_metric.name.lower(),
+            "algorithm": vector_options.algorithm.name.lower(),
+            "datatype": vector_options.type.name.lower(),
+        }
+        if vector_options.initial_cap:
+            attrs["initial_cap"] = vector_options.initial_cap
+        is_flat = vector_options.algorithm.name == "FLAT"
+        if is_flat and vector_options.block_size:
+            attrs["block_size"] = vector_options.block_size
+        if vector_options.algorithm.name == "HNSW":
+            if vector_options.m:
+                attrs["m"] = vector_options.m
+            if vector_options.ef_construction:
+                attrs["ef_construction"] = vector_options.ef_construction
+            if vector_options.ef_runtime:
+                attrs["ef_runtime"] = vector_options.ef_runtime
+            if vector_options.epsilon:
+                attrs["epsilon"] = vector_options.epsilon
+        return {"name": field_name, "type": "vector", "attrs": attrs}
+
+    # Numeric field
+    if is_numeric_type(field_type):
+        attrs = {"sortable": sortable}
+        return {"name": field_name, "type": "numeric", "attrs": attrs}
+
+    # Boolean - stored as TAG
+    if field_type is bool:
+        return {"name": field_name, "type": "tag"}
+
+    # String field
+    if isinstance(field_type, type) and issubclass(field_type, str):
+        if full_text_search:
+            attrs = {"sortable": sortable}
+            return {"name": field_name, "type": "text", "attrs": attrs}
+        else:
+            attrs = {"sortable": sortable, "case_sensitive": case_sensitive}
+            return {"name": field_name, "type": "tag", "attrs": attrs}
+
+    # List of strings -> TAG
+    if is_supported_container_type(field_type):
+        from typing import get_args
+
+        inner_types = get_args(field_type)
+        if inner_types and inner_types[0] is str:
+            attrs = {"sortable": sortable}
+            return {"name": field_name, "type": "tag", "attrs": attrs}
+
+    # Default to tag for unknown types
+    return {"name": field_name, "type": "tag"}
+
+
+def to_redisvl_schema(model_cls: Type[RedisModel]) -> "IndexSchema":
+    """
+    Convert a Redis OM model to a RedisVL IndexSchema.
+
+    This allows you to use RedisVL's advanced query capabilities with your
+    Redis OM models, including:
+    - VectorQuery with hybrid policies (BATCHES, ADHOC_BF)
+    - VectorRangeQuery for epsilon-based searches
+    - Advanced filter expressions
+    - EF_RUNTIME tuning for HNSW indexes
+
+    Args:
+        model_cls: A HashModel or JsonModel class with index=True
+
+    Returns:
+        A RedisVL IndexSchema that can be used with SearchIndex
+
+    Raises:
+        ValueError: If the model is not indexed
+
+    Example:
+        schema = to_redisvl_schema(MyModel)
+        index = SearchIndex(schema=schema, redis_client=redis)
+        results = await index.query(VectorQuery(...))
+    """
+    # Check if model is indexed
+    # model_config is a dict in Pydantic v2
+    model_config = getattr(model_cls, "model_config", {})
+    if isinstance(model_config, dict):
+        is_indexed = model_config.get("index", False)
+    else:
+        is_indexed = False
+    if not is_indexed:
+        raise ValueError(
+            f"Model {model_cls.__name__} is not indexed. "
+            "Use 'class MyModel(JsonModel, index=True):' to enable indexing."
+        )
+
+    # Determine storage type
+    is_json = issubclass(model_cls, JsonModel)
+    storage_type = "json" if is_json else "hash"
+
+    # Get index name and prefix
+    index_name = model_cls.Meta.index_name
+    key_prefix = model_cls.make_key("")
+
+    # Build field definitions
+    fields: List[Dict[str, Any]] = []
+
+    for name, field in model_cls.model_fields.items():
+        field_type = get_outer_type(field)
+        if field_type is None:
+            continue
+
+        # Get FieldInfo (may be wrapped in metadata)
+        if (
+            not isinstance(field, FieldInfo)
+            and hasattr(field, "metadata")
+            and len(field.metadata) > 0
+            and isinstance(field.metadata[0], FieldInfo)
+        ):
+            field_info = field.metadata[0]
+        elif isinstance(field, FieldInfo):
+            field_info = field
+        else:
+            continue
+
+        field_def = _get_field_type(name, field_type, field_info, is_json)
+        if field_def:
+            fields.append(field_def)
+
+    # Build schema dict
+    schema_dict = {
+        "index": {
+            "name": index_name,
+            "prefix": key_prefix,
+            "storage_type": storage_type,
+        },
+        "fields": fields,
+    }
+
+    return IndexSchema.from_dict(schema_dict)
+
+
+def get_redisvl_index(
+    model_cls: Type[RedisModel],
+    async_client: bool = True,
+) -> Union["AsyncSearchIndex", "SearchIndex"]:
+    """
+    Get a RedisVL SearchIndex for a Redis OM model.
+
+    This provides a ready-to-use SearchIndex connected to the model's
+    Redis database, enabling advanced vector search operations.
+
+    Args:
+        model_cls: A HashModel or JsonModel class with index=True
+        async_client: If True (default), return AsyncSearchIndex.
+                     If False, return sync SearchIndex.
+
+    Returns:
+        A RedisVL SearchIndex (async or sync) connected to Redis
+
+    Raises:
+        ValueError: If the model is not indexed
+
+    Example:
+        index = get_redisvl_index(MyModel)
+        results = await index.query(VectorQuery(
+            vector=query_embedding,
+            vector_field_name="embedding",
+            num_results=10,
+        ))
+    """
+    schema = to_redisvl_schema(model_cls)
+    redis_client = model_cls.db()
+
+    if async_client:
+        return AsyncSearchIndex(schema=schema, redis_client=redis_client)
+    else:
+        return SearchIndex(schema=schema, redis_client=redis_client)