Knowledge Base
Overview
Store and manage documents to provide context to LLMs in your flows.
Knowledge Base
Store documents to provide context to LLMs in your flows. Each knowledge base stores text and image embeddings separately for retrieval and context augmentation.
Use the LLM Context Indexer block to ingest documents and LLM Context Retriever block to retrieve chunks that provide context to your LLM.
Database Support
| Database | Text Embeddings | Image Embeddings |
|---|---|---|
| MongoDB | ✓ | ✗ |
| MongoDB Atlas | ✓ | ✗ |
| QdrantDB | ✓ | ✓ |
Supported Document Types
| Type | Description |
|---|---|
| PDF / WORD | Layout or markdown extraction, optional image RAG pass |
| MD / TXT | Configurable chunk size and overlap |
| Excel | Unstructured or pandas style processing. Ingest to Qdrant/MongoDB Atlas or return extraction directly |
| JSON | Raw JSON ingestion (non-embedding) or embed a chosen content field |
| IMAGE | Image embedding ingestion (Qdrant only) |
Documents
Click View Documents to see ingested documents.
Each document shows:
- Status - Failed documents can be re-ingested
- Chunks - Text and image segments
- Retrieval Count - Usage statistics
- Ingestion Parameters - Settings used during ingestion
- Metadata - Document properties