What is a Knowledge Base?
A knowledge base is a collection of documents that AI agents can search to provide accurate, context-aware responses. Documents are chunked, embedded, and stored for semantic search.
Embedding Providers
- Ollama: Free, local embeddings. Requires Ollama running locally.
- OpenAI: Cloud-based, high quality. Requires API key.
Ollama Models
- nomic-embed-text-v2-moe: High quality MoE model
- qwen3-embedding: Large multilingual model
- embeddinggemma: Lightweight, fast
Chunking Tips
- Chunk Size: Larger chunks = more context, fewer results
- Overlap: Helps maintain context across chunk boundaries
- Recommended: 512 tokens with 50 overlap for most cases