Skip to content

LLM Search Embeddings

The LLM Search Embeddings task accepts an embedding vector directly via the embeddings parameter. If embeddings are not provided, the task instead embeds the query parameter using the specified embedding model provider and model, then searches with the resulting vector. This makes it useful both for reusing a vector you've already generated (for example, via LLM Generate Embeddings) and for searching directly from natural language.

Prerequisites

Task parameters

Configure these parameters for the LLM Search Embeddings task.

Parameter Description Required/ Optional
inputParameters.vectorDB The vector database to search.

Note: If you haven’t configured the vector database on your Orkes Conductor cluster, navigate to the Integrations tab and configure your required provider.		 Required.
inputParameters.index The index in your vector database to search for relevant embeddings.

The terminology of the index field varies depending on the integration:
  • For Weaviate, the index field indicates the collection name.
  • For other integrations, it denotes the index name.
 Required.
inputParameters.namespace Namespaces are separate isolated environments within the database to manage and organize vector data effectively. Enter the namespace the task will utilize.

The usage and terminology of the namespace field vary depending on the integration:
  • For Pinecone, the namespace field is applicable.
  • For Weaviate, the namespace field is not applicable.
  • For MongoDB, the namespace field is referred to as “Collection” in MongoDB.
  • For Postgres, the namespace field is referred to as “Table” in Postgres.
 Required.
inputParameters.embeddingModelProvider The LLM provider used to embed query. Required only when searching by query instead of embeddings. Optional.
inputParameters.embeddingModel The embedding model used to embed query. Required only when searching by query instead of embeddings. Optional.
inputParameters.embeddings The embedding vector to search with. This should be from the same embedding model used to create the embeddings stored in the specified index. Required.
inputParameters.query A natural-language query to embed and search with. Used only when embeddings is not set. Optional.
inputParameters.maxResults The maximum number of results to return. A non-zero integer between 1 and 10000. Required.
inputParameters.dimensions The size of the embedding vector. Optional.
inputParameters.metadata A map of key-value pairs to filter results by document metadata. Optional.

The following are generic configuration parameters that can be applied to the task and are not specific to the LLM Search Embeddings task.

Caching parameters

You can cache the task outputs using the following parameters. Refer to Caching Task Outputs for a full guide.

Parameter Description Required/ Optional
cacheConfig.ttlInSecond The time to live in seconds, which is the duration for the output to be cached. Required if using cacheConfig.
cacheConfig.key The cache key is a unique identifier for the cached output and must be constructed exclusively from the task’s input parameters.
It can be a string concatenation that contains the task’s input keys, such as ${uri}-${method} or re_${uri}_${method}.		 Required if using cacheConfig.
Other generic parameters

Here are other parameters for configuring the task behavior.

Parameter Description Required/ Optional
optional Whether the task is optional.

If set to true, any task failure is ignored, and the workflow continues with the task status updated to COMPLETED_WITH_ERRORS. However, the task must reach a terminal state. If the task remains incomplete, the workflow waits until it reaches a terminal state before proceeding.		 Optional.

Task configuration

This is the task configuration for an LLM Search Embeddings task.

{
  "name": "llm_search_embeddings",
  "taskReferenceName": "llm_search_embeddings_ref",
  "inputParameters": {
    "vectorDB": "pinecone",
    "index": "doc-demo",
    "namespace": "kb",
    "embeddings": "${generate_embedding_ref.output.result}",
    "maxResults": 10
  },
  "type": "LLM_SEARCH_EMBEDDINGS"
}

Task output

The LLM Search Embeddings task will return the following parameters.

Parameter Description
result A JSON array containing the results of the query.
score Represents a value quantifying the degree of likeness between a specific item and a query vector, facilitating ranking and ordering of results. Higher scores denote stronger relevance to the query vector.
metadata An object containing additional metadata related to the retrieved document.
docId The unique identifier of the queried document.
parentDocId An identifier that denotes a parent document in hierarchical or relational data structures.
text The actual content retrieved.

Examples

Here are some examples for using the LLM Search Embeddings task.

Using an LLM Search Embeddings task in a workflow 
{
  "name": "llm_search_embeddings",
  "taskReferenceName": "llm_search_embeddings_ref",
  "inputParameters": {
    "vectorDB": "pinecone",
    "index": "doc-demo",
    "namespace": "kb",
    "embeddings": "${generate_embedding_ref.output.result}",
    "maxResults": 10
  },
  "type": "LLM_SEARCH_EMBEDDINGS"
}