Knowledge, RAG, and Workflows
Knowledge & RAG
Upload documents for agents to search using vector embeddings.
- Supported formats: PDF, text files, markdown, web URLs.
- Embedding providers: OpenAI, Cohere, Jina, Mistral, VoyageAI.
- Vector stores: Pinecone, Qdrant, pgvector.
- How to set up: Go to Knowledge in the sidebar → create a knowledge source → upload documents → attach the source to an agent.
- Agents with knowledge sources attached can search relevant documents when answering questions via the built-in
search_knowledge_basetool.
Workflows
Visual drag-and-drop builder for multi-step automations.
- Node types: Trigger (start), Action (do something), Condition (if/else), Delay (wait), Loop (repeat).
- Triggers: Webhook, schedule, manual, or event-based.
- Building: Go to Workflows → Create Workflow → use the visual builder to add and connect nodes.
Was this article helpful?
Your feedback helps us improve our docs.