Configuration¶

MFS stores user configuration at:

~/.mfs/config.toml

Generate or inspect it:

mfs config path
mfs config init
mfs config show

Minimal config¶

[embedding]
provider = "openai"
model = "text-embedding-3-small"

[milvus]
uri = "~/.mfs/milvus.db"
collection_name = "mfs_chunks"
account_id = "default"

All keys have defaults. The generated config comments document the available values.

Embeddings¶

Supported providers:

Provider	Typical model
`openai`	`text-embedding-3-small`
`onnx`	local BGE-M3 ONNX model
`google`	Gemini embeddings
`voyage`	Voyage embeddings
`jina`	Jina embeddings
`mistral`	Mistral embeddings
`ollama`	local Ollama embedding model
`local`	sentence-transformers

Examples:

mfs config set embedding.provider onnx
mfs config set embedding.provider openai
mfs config set embedding.model text-embedding-3-large

Provider API keys can be stored in config, but environment variables are usually better:

export OPENAI_API_KEY="sk-..."
export GOOGLE_API_KEY="..."
export VOYAGE_API_KEY="..."
export JINA_API_KEY="..."
export MISTRAL_API_KEY="..."

Optional LLM and VLM enrichment¶

LLM config is only used when you request summaries or image descriptions.

[llm]
provider = "openai"
model = "gpt-4o-mini"

Text summaries are available through OpenAI, Anthropic, Google, Ollama, and Mistral providers. Image descriptions require a vision-capable provider.

Indexing filters¶

[indexing]
include_extensions = []
exclude_extensions = []

If include_extensions is non-empty, only those extensions are indexed. exclude_extensions always wins.

MFS also reads .gitignore and .mfsignore patterns from the project root.

Cache¶

[cache]
max_size_mb = 500

This controls the converted Markdown cache for PDF and DOCX files.

Milvus¶

[milvus]
uri = "~/.mfs/milvus.db"
collection_name = "mfs_chunks"
account_id = "default"
token = ""

See Milvus Backends for backend-specific examples.