Knowledge Base

The Knowledge Base lets you upload documents, data, and other content that your agents can reference when answering questions. It powers retrieval-augmented generation (RAG) to ground agent responses in your proprietary data.

Supported Formats

• PDF documents
• Markdown / plain text files
• CSV and structured data
• HTML pages and web content
• JSON and JSONL data files

Uploading Content

Via Dashboard

Navigate to the Knowledge Base section in the dashboard and drag-and-drop your files, or click to browse. Files are automatically processed, chunked, and indexed.

Via API

await lens.knowledgeBase.upload({
  file: './data/product-docs.pdf',
  metadata: {
    category: 'product',
    version: '2.0',
  },
})

How Retrieval Works

When an agent receives a question, LensOS automatically:

• Converts the query into an embedding vector
• Searches the knowledge base for semantically similar content
• Retrieves the most relevant chunks
• Injects the context into the agent's prompt

This process is fully automatic — you don't need to manage the retrieval pipeline yourself.

Configuration

• Chunk Size — Control how documents are split (default: 512 tokens).
• Overlap — Set overlap between chunks for better context (default: 50 tokens).
• Top K — Number of chunks to retrieve per query (default: 5).
• Similarity Threshold — Minimum relevance score for retrieved content.