Vector Stores

Vector Stores are knowledge bases that your AI agents can search during conversations and other supported workflows. After you upload documents, agents can use the indexed content to answer questions more accurately.

What are Vector Stores?

A vector store converts your documents into searchable embeddings. When an agent needs information, it performs a semantic search across the indexed content to find the most relevant passages.

Managing Vector Stores

Navigate to Settings > Vector Stores to manage your knowledge bases.

Creating a Vector Store

1. Click + Create Vector Store
2. Enter a name and description
3. Upload documents:
   • Supported formats: PDF, TXT, DOCX, MD
   • Upload one or multiple files
4. Documents are automatically processed, chunked, and indexed

Assigning to Agents

After creating a vector store, assign it to your agents:

1. Open the agent configuration (voice, chat, SMS, or task agent)
2. Enable the Knowledge Base tool
3. Select which vector store(s) the agent can search

How Agents Use Vector Stores

When an agent needs factual information from uploaded documents:

1. The agent determines a search query based on the conversation
2. It searches the vector store using semantic similarity
3. Relevant passages are retrieved and used to inform the response
4. The agent cites the information naturally in the conversation

Best Practices

• Keep documents focused — Upload topic-specific documents rather than massive general files
• Update regularly — Re-upload documents when information changes
• Name descriptively — Use clear names so you know what each store contains
• Multiple stores — Create separate stores for different topics and assign them to relevant agents

Viewing a Vector Store

Open a vector store to review details such as:

• List of uploaded documents
• Document count and total size
• Chunk count (how many searchable segments)
• Last updated date