Configuration

All configuration is done via .env.ragpack, created by ragpack init. The file is read on startup — restart the stack after making changes.

ragpack stop && ragpack start

Embedding providers

RagPack needs an embedding model to index and query documents. Set DEFAULT_EMBED_PROVIDER to one of ollama, openai, or tei.

Ollama (default, fully local)

DEFAULT_EMBED_PROVIDER=ollama
OLLAMA_BASE_URL=http://ollama:11434
OLLAMA_EMBED_MODEL=nomic-embed-text

Start the stack with the Ollama profile to include the Ollama service:

ragpack start --profile ollama

OpenAI

DEFAULT_EMBED_PROVIDER=openai
OPENAI_API_KEY=sk-...
OPENAI_EMBED_MODEL=text-embedding-3-small

HuggingFace TEI (self-hosted)

DEFAULT_EMBED_PROVIDER=tei
TEI_BASE_URL=http://localhost:8080
TEI_EMBED_MODEL=BAAI/bge-small-en-v1.5

LLM providers

An LLM is required to use the RAG endpoint (collection.rag()). Set DEFAULT_LLM_PROVIDER to one of openai, ollama, or anthropic.

If no LLM provider is configured, semantic search (collection.findSimilar()) still works — only RAG is unavailable.

OpenAI

DEFAULT_LLM_PROVIDER=openai
OPENAI_API_KEY=sk-...
OPENAI_LLM_MODEL=gpt-4o-mini

Ollama

DEFAULT_LLM_PROVIDER=ollama
OLLAMA_BASE_URL=http://ollama:11434
OLLAMA_LLM_MODEL=llama3.2

Anthropic

DEFAULT_LLM_PROVIDER=anthropic
ANTHROPIC_API_KEY=sk-ant-...
ANTHROPIC_MODEL=claude-haiku-4-5-20251001

Ingestion

WORKER_COUNT=5        # concurrent ingestion workers
EMBED_RATE_LIMIT=10   # max embedding API calls per second
MAX_UPLOAD_SIZE_MB=20 # max file upload size

EMBED_RATE_LIMIT is especially important when using an external provider like OpenAI — set it to stay within your API tier's rate limit.

Chunking defaults

These apply to all collections unless overridden at collection creation time.

CHUNK_STRATEGY=auto   # auto | paragraph | sliding_window | section | unit | row_group
CHUNK_SIZE=2000       # max characters per chunk
CHUNK_OVERLAP=200     # characters carried into the next chunk

RAG defaults

DEFAULT_PROMPT_SLUG=basic_rag  # prompt template used when none is specified

Full reference

Variable	Default	Description
DEFAULT_EMBED_PROVIDER	ollama	Embedding provider: ollama | openai | tei
OLLAMA_BASE_URL	http://ollama:11434	Ollama server URL
OLLAMA_EMBED_MODEL	nomic-embed-text	Ollama embedding model
OPENAI_API_KEY	—	OpenAI API key
OPENAI_EMBED_MODEL	text-embedding-3-small	OpenAI embedding model
TEI_BASE_URL	http://localhost:8080	TEI server URL
TEI_EMBED_MODEL	BAAI/bge-small-en-v1.5	TEI embedding model
DEFAULT_LLM_PROVIDER	—	LLM provider for RAG: openai | ollama | anthropic
OPENAI_LLM_MODEL	gpt-4o-mini	OpenAI LLM model
OLLAMA_LLM_MODEL	llama3.2	Ollama LLM model
ANTHROPIC_API_KEY	—	Anthropic API key
ANTHROPIC_MODEL	claude-haiku-4-5-20251001	Anthropic model
WORKER_COUNT	5	Concurrent ingestion workers
EMBED_RATE_LIMIT	10	Max embedding API calls per second
MAX_UPLOAD_SIZE_MB	20	Max file upload size in MB
CHUNK_STRATEGY	auto	Default chunking strategy
CHUNK_SIZE	2000	Max characters per chunk
CHUNK_OVERLAP	200	Overlap characters between chunks
DEFAULT_PROMPT_SLUG	basic_rag	Default RAG prompt template