WeaviateEx.GRPC.Services.Search (WeaviateEx v0.7.4)
View SourcegRPC Search service for vector queries.
This module provides high-level functions for performing vector searches against Weaviate using gRPC, including generative search (RAG) support.
Usage
{:ok, channel} = WeaviateEx.GRPC.Channel.connect(config)
{:ok, results} = Search.near_vector(channel, "Article", [0.1, 0.2, ...],
limit: 10,
return_properties: ["title", "content"]
)Generative Search
{:ok, results} = Search.near_text(channel, "Article", "machine learning",
generative: %{
single_prompt: "Summarize this article: {content}",
provider: :openai,
model: "gpt-4"
}
)Summary
Functions
Perform a BM25 keyword search.
Execute a raw SearchRequest.
Perform a hybrid search combining vector and keyword search.
Perform a multimodal image search.
Perform a multimodal media search (audio, video, depth, thermal, IMU).
Perform a search for objects similar to a given object.
Perform a text-based vector search using a text-to-vector model.
Perform a vector similarity search.
Perform a generic search with a filter request map.
Types
@type search_opts() :: [ limit: non_neg_integer(), offset: non_neg_integer(), return_properties: [String.t()], return_references: [WeaviateEx.Query.QueryReference.t()], return_metadata: [atom()], filters: map(), group_by: WeaviateEx.Query.GroupBy.t(), tenant: String.t(), certainty: float(), distance: float(), move_to: map(), move_away: map(), target_vectors: WeaviateEx.Query.TargetVectors.t(), bm25_search_operator: map(), max_vector_distance: float(), fusion_type: atom(), after: String.t(), sort: list(), autocut: non_neg_integer(), generative: WeaviateEx.GRPC.Generative.config() ]
Functions
@spec bm25(GRPC.Channel.t(), String.t(), String.t(), search_opts()) :: {:ok, struct()} | {:error, WeaviateEx.Error.t()}
Perform a BM25 keyword search.
Options
:properties- List of properties to search in
Examples
{:ok, results} = Search.bm25(channel, "Article", "machine learning",
properties: ["title", "content"],
limit: 10
)@spec execute(GRPC.Channel.t(), struct(), keyword()) :: {:ok, struct()} | {:error, WeaviateEx.Error.t()}
Execute a raw SearchRequest.
Useful for complex queries that need full control over the request.
Examples
request = %SearchRequest{
collection: "Article",
limit: 10,
near_vector: %NearVector{vector_bytes: <<...>>}
}
{:ok, results} = Search.execute(channel, request)@spec hybrid(GRPC.Channel.t(), String.t(), String.t(), search_opts()) :: {:ok, struct()} | {:error, WeaviateEx.Error.t()}
Perform a hybrid search combining vector and keyword search.
Options
:alpha- Weight between vector (1.0) and keyword (0.0) search:fusion_type- Fusion algorithm (:ranked or :relative_score):properties- Properties to search for BM25
Examples
{:ok, results} = Search.hybrid(channel, "Article", "machine learning",
alpha: 0.5,
limit: 10
)@spec near_image( GRPC.Channel.t(), String.t(), WeaviateEx.Query.NearImage.t(), search_opts() ) :: {:ok, struct()} | {:error, WeaviateEx.Error.t()}
Perform a multimodal image search.
Options
:limit- Maximum number of results (default: 10):offset- Number of results to skip:return_properties- List of property names to return:return_metadata- List of metadata fields (e.g., [:uuid, :distance, :vector]):tenant- Tenant name for multi-tenant collections
@spec near_media( GRPC.Channel.t(), String.t(), WeaviateEx.Query.NearMedia.t(), search_opts() ) :: {:ok, struct()} | {:error, WeaviateEx.Error.t()}
Perform a multimodal media search (audio, video, depth, thermal, IMU).
@spec near_object(GRPC.Channel.t(), String.t(), String.t(), search_opts()) :: {:ok, struct()} | {:error, WeaviateEx.Error.t()}
Perform a search for objects similar to a given object.
Examples
{:ok, results} = Search.near_object(channel, "Article", "uuid-of-object",
limit: 10
)@spec near_text( GRPC.Channel.t(), String.t(), String.t() | [String.t()], search_opts() ) :: {:ok, struct()} | {:error, WeaviateEx.Error.t()}
Perform a text-based vector search using a text-to-vector model.
Options
:move_to- Concepts to move towards:move_away- Concepts to move away from- Plus all options from
near_vector/4
Examples
{:ok, results} = Search.near_text(channel, "Article", "machine learning",
limit: 10
)@spec near_vector(GRPC.Channel.t(), String.t(), [float()], search_opts()) :: {:ok, struct()} | {:error, WeaviateEx.Error.t()}
Perform a vector similarity search.
Options
:limit- Maximum number of results (default: 10):offset- Number of results to skip:return_properties- List of property names to return:return_metadata- List of metadata fields (e.g., [:uuid, :distance, :vector]):tenant- Tenant name for multi-tenant collections:certainty- Minimum certainty threshold (0.0 to 1.0):distance- Maximum distance threshold
Examples
{:ok, results} = Search.near_vector(channel, "Article", vector,
limit: 10,
return_properties: ["title", "content"]
)@spec search(GRPC.Channel.t(), String.t(), map(), keyword()) :: {:ok, struct()} | {:error, WeaviateEx.Error.t()}
Perform a generic search with a filter request map.
This is a lower-level function used by the Debug module for protocol comparison.
Options
:metadata- gRPC metadata headers
Examples
request = %{collection: "Article", filters: %{...}, limit: 1}
{:ok, results} = Search.search(channel, "Article", request, metadata: metadata)