Configuration

Sententia can be configured via CLI arguments, environment variables, and .env files.

Priority of sources (highest to lowest):

1. CLI arguments — values passed via command line override everything
2. Environment variables — SENTENTIA_* variables
3. .env file — values loaded from an env-file
4. Defaults — built-in default values

CLI Arguments

Required

Parameter	Description
data_dir	Positional argument — path to directory with Markdown files
--llm-protocol	LLM provider type: openai, ollama, anthropic
--llm-url	API endpoint URL
--llm-model	Model identifier

Optional

Parameter	Default	Description
--env-file	.env	Path to env-file with configuration
--index-path	—	Path to index file. If not specified, index is built in memory
--llm-token	—	API key (not required for Ollama). Falls back to SENTENTIA_LLM_TOKEN
--mcp	False	MCP Server mode flag. When True, starts MCP Server instead of REST API
--host	0.0.0.0	Server bind address
--port	8000	Server port

Environment Variables

All environment variables use the SENTENTIA_ prefix.

Variable	Field	Default
SENTENTIA_DATA_DIR	data_dir	""
SENTENTIA_INDEX_PATH	index_path	None
SENTENTIA_LLM_PROTOCOL	llm_protocol	""
SENTENTIA_LLM_URL	llm_url	""
SENTENTIA_LLM_MODEL	llm_model	""
SENTENTIA_LLM_TOKEN	llm_token	None
SENTENTIA_MCP	mcp	False
SENTENTIA_HOST	host	"0.0.0.0"
SENTENTIA_PORT	port	8000

Protocol Matching

--llm-protocol	Provider	API
openai	OpenaiProvider	POST {url}/v1/chat/completions
ollama	OpenaiProvider	POST {url}/v1/chat/completions (no auth)
anthropic	AnthropicProvider	POST {url}/v1/messages

Examples

With persistent index

sententia /data/docs \
  --index-path /data/index.faiss \
  --llm-protocol openai \
  --llm-url https://api.openai.com \
  --llm-model gpt-4 \
  --llm-token "$OPENAI_API_KEY"

In-memory index

sententia /data/docs \
  --llm-protocol ollama \
  --llm-url http://localhost:11434 \
  --llm-model llama3

MCP Server

sententia /data/docs \
  --llm-protocol anthropic \
  --llm-url https://api.anthropic.com \
  --llm-model claude-3-sonnet \
  --mcp \
  --port 9000

Using .env file

Create a .env file in the working directory:

SENTENTIA_LLM_PROTOCOL=openai
SENTENTIA_LLM_URL=https://api.openai.com
SENTENTIA_LLM_MODEL=gpt-4
SENTENTIA_LLM_TOKEN=sk-...

Run with minimal arguments:

sententia /data/docs

Custom env-file path

sententia /data/docs --env-file /path/to/config.env