Knowledge Base

Give your agent domain-specific knowledge

Knowledge Base

The knowledge base lets you upload documents and websites that your agent can reference during conversations. When a user asks a question, the agent searches the knowledge base for relevant context and uses it to generate accurate answers.

How it works

1. You upload content (files or URLs)
2. Content is split into chunks and converted to vector embeddings
3. When the agent receives a message, it searches for the most relevant chunks
4. Matching chunks are injected into the agent’s context before generating a response

Supported formats

Type	Formats
Documents	PDF, DOCX, TXT, MD
Web pages	Any public URL (content is crawled and extracted)
Plain text	Paste text directly

Adding sources

1. Open your agent’s settings
2. Go to the Knowledge Base tab
3. Click Add Source
4. Choose the source type and upload/enter content
5. Wait for processing (usually a few seconds per document)

Best practices

• Be specific — Upload only content relevant to your agent’s domain. More data isn’t always better.
• Keep content fresh — Re-upload documents when they change. Old content can lead to outdated answers.
• Chunk size matters — The system automatically chunks documents. Very long documents work fine.
• Test retrieval — Ask your agent questions that should be answered by the knowledge base and verify accuracy.

Vector store

Aimna uses Pinecone as the vector store with OpenAI embeddings for semantic search. The minimum similarity score for retrieval is 0.3 — this ensures relevant results without being too restrictive.