Vector Operations

Store and query vector embeddings for similarity search and RAG applications

Vector operations allow you to store and retrieve vector embeddings from vector databases like Qdrant. These actions are essential for building semantic search, recommendation systems, and Retrieval-Augmented Generation (RAG) applications.

---

storevector

Stores vector embeddings in a vector database along with associated metadata.

Integration ID

The vector database integration to store vectors in.

YAML Key	integrationID
Type	string
Required	Yes

Vectors

The vector data to store, typically as a JSON array of floats.

YAML Key	vectors
Type	string
Required	Yes

Use {{ .generate_embedding }} to reference vectors generated by a previous action, such as an AI embedding model.

Fields

Metadata fields to store alongside the vector. These fields can be used for filtering during retrieval.

YAML Key	fields
Type	map
Required	Yes

Options

Additional storage options such as the collection name.

YAML Key	options
Type	map
Required	No

Common options:

• collection — The collection name to store vectors in

Example

actions:
  store_embedding:
    type: storevector
    config:
      integrationID: my_qdrant
      vectors: "{{ .generate_embedding }}"
      fields:
        content: "{{ param \"text\" }}"
        source: "user_input"
        timestamp: "{{ now }}"
      options:
        collection: documents
    next: response.success
    fail: response.error

---

fetchvectors

Retrieves vectors from a vector database for similarity search. Returns the most similar vectors to the provided query vector.

Integration ID

The vector database integration to query.

YAML Key	integrationID
Type	string
Required	Yes

Vector

The query vector to search for similar vectors. This is typically an embedding of the search query.

YAML Key	vector
Type	string
Required	No

Options

Query options such as collection name and result limit.

YAML Key	options
Type	map
Required	No

Common options:

• collection — The collection name to search in
• limit — Maximum number of results to return

Example

actions:
  search_similar:
    type: fetchvectors
    config:
      integrationID: my_qdrant
      vector: "{{ .query_embedding }}"
      options:
        collection: documents
        limit: 10
    next: response.results
    fail: response.error

---

Common Patterns

RAG Pipeline

A typical RAG (Retrieval-Augmented Generation) workflow combines vector search with an AI agent:

actions:
  embed_query:
    type: agent
    config:
      integrationID: my_openai
      userPrompt: "Generate an embedding for: {{ param \"question\" }}"
    next: action.search_docs

  search_docs:
    type: fetchvectors
    config:
      integrationID: my_qdrant
      vector: "{{ .embed_query }}"
      options:
        collection: knowledge_base
        limit: 5
    next: action.generate_answer

  generate_answer:
    type: agent
    config:
      integrationID: my_openai
      systemPrompt: |
        Answer the user's question using only the provided context.
        Context: {{ .search_docs }}
      userPrompt: "{{ param \"question\" }}"
    next: response.answer

Document Ingestion

Store documents with their embeddings for later retrieval:

actions:
  generate_embedding:
    type: agent
    config:
      integrationID: my_openai
      userPrompt: "Generate an embedding for: {{ param \"content\" }}"
    next: action.store_document

  store_document:
    type: storevector
    config:
      integrationID: my_qdrant
      vectors: "{{ .generate_embedding }}"
      fields:
        content: "{{ param \"content\" }}"
        title: "{{ param \"title\" }}"
        category: "{{ param \"category\" }}"
      options:
        collection: documents
    next: response.stored

---

Next Steps

AI Agents

Use AI agents to generate embeddings and process search results.

Data Operations

Combine vector search with traditional database queries.

Actions Overview

Learn the fundamentals of ServFlow actions.

Configuration Reference

Explore all ServFlow configuration options.