Uploading documents

Add documents to a collection to make them searchable. Documents are automatically chunked and embedded using the collection’s configured model.

Document format

Each document contains the following fields:

Field	Required	Description
content	Yes	The text content (10–10,000 characters)
metadata	No	Key-value pairs for filtering and context
searchableFields	No	Additional fields to include in search

Uploading documents

Navigate to a collection from the Vector Database page, then use the upload interface to add documents. You can upload individual documents or batches.

Auto-chunking

Large documents are automatically split into smaller chunks for more precise search results. You can configure chunking behaviour with the following settings:

Setting	Default	Range	Description
autoChunk	true	–	Enable automatic chunking
chunkSize	1000	500–2000	Target characters per chunk
chunkOverlap	200	0–500	Overlap between consecutive chunks

Tip

Smaller chunks (500–800 characters) work better for precise question-and-answer scenarios. Larger chunks (1000–2000 characters) are better for providing broader context.

Batch upload

Upload up to 100 documents in a single request. Each document’s content must be between 10 and 10,000 characters.

Metadata

Attach key-value metadata to documents for additional context. Metadata is returned with search results and can help agents understand the source and context of retrieved documents.

{
  "content": "Your document text here...",
  "metadata": {
    "source": "knowledge-base",
    "category": "security",
    "lastUpdated": "2025-01-15"
  }
}

Next steps

• Semantic search – query your documents using natural language or vectors
• Agent configuration – assign collections to agents for retrieval-augmented generation