Databases (RAG Configuration)

The Databases section is where you configure Retrieval-Augmented Generation — vector databases, embedding models, and retrieval strategies.

Vector Databases

Create and manage vector databases:

• ChromaDB (default) — local vector storage, zero configuration
• Pinecone, Weaviate, etc. — cloud-hosted options

Each database needs a unique name and storage configuration (connection details for hosted options).

Embedding Strategies

Configure how text is converted to vectors:

Setting	Description
Embedding model	Which model creates embeddings (e.g., nomic-embed-text)
Chunk size	How large each embedded piece should be
Overlap	How much chunks overlap to preserve context

Changing Embedding Models

The Change Embedding Model page lets you switch models for a database. This will re-embed all documents, which can take time for large datasets.

Adding Embedding Strategies

Create new embedding strategies with custom chunk sizes, overlap, and model selection. Local models are shown in a table with download status.

Retrieval Strategies

Define how relevant documents are found when querying:

Method	Description
Similarity Search	Pure vector similarity (default)
Hybrid Search	Combines vector + keyword matching (BM25)
MMR	Maximal Marginal Relevance — balances relevance with diversity
Metadata Filtering	Filter by document properties before similarity search

Managing Strategies

• Add — create a new retrieval strategy
• Edit — modify parameters like top-k, score threshold, or filters
• View — see the full strategy configuration

Testing Queries

Use the built-in query tester to:

1. Enter a test question
2. See which documents are retrieved
3. Check relevance scores and ranking
4. Adjust parameters if needed

API Routes

Action	Method	Route
List databases	GET	/v1/projects/{ns}/{project}/rag/databases
Create database	POST	/v1/projects/{ns}/{project}/rag/databases
RAG query	POST	/v1/projects/{ns}/{project}/rag/query
RAG health	GET	/v1/projects/{ns}/{project}/rag/health

Route

/chat/databases
/chat/databases/add-embedding
/chat/databases/add-retrieval
/chat/rag/:strategyId/change-embedding
/chat/rag/:strategyId/retrieval