Skip to main content
The init command generates a default configuration file and project structure for a new GraphRAG project.

​
Usage

graphrag init [OPTIONS]

​
Options

​
--root
string
default:"current directory"
The project root directory where the configuration will be created.Aliases: -r
​
--model
string
default:"gpt-4.1"
The default chat/completion model to use for entity extraction, summarization, and other LLM operations.Aliases: -mThis will be written to the settings.yaml configuration file and can be changed later.
​
--embedding
string
default:"text-embedding-3-large"
The default embedding model to use for generating vector embeddings.Aliases: -eThis will be written to the settings.yaml configuration file.
​
--force
boolean
default:"false"
Force initialization even if the project already exists. This will overwrite existing configuration files.Aliases: -f

​
Examples

​
Initialize with default settings

Create a new project in the current directory: 
graphrag init

​
Initialize in a specific directory

graphrag init --root ./my-project

​
Specify custom models

graphrag init --model gpt-4o --embedding text-embedding-3-small

​
Force overwrite existing project

graphrag init --force

​
What gets created

When you run graphrag init, the following files and directories are created:

​
Configuration files

  • settings.yaml - Main configuration file with indexing and search settings
  • .env - Environment variables file for API keys and secrets

​
Directories

  • input/ - Place your source documents here (text files, PDFs, etc.)
  • prompts/ - Contains all prompt templates used during indexing and querying

​
Prompt files

The following prompt templates are created in the prompts/ directory:
  • extract_graph.txt - Entity and relationship extraction
  • summarize_descriptions.txt - Entity description summarization
  • extract_claims.txt - Claim extraction (if enabled)
  • community_report_graph.txt - Community report generation from graph
  • community_report_text.txt - Community report generation from text
  • drift_search_system_prompt.txt - DRIFT search system prompt
  • drift_reduce_prompt.txt - DRIFT search reduce prompt
  • global_search_map_system_prompt.txt - Global search map phase
  • global_search_reduce_system_prompt.txt - Global search reduce phase
  • global_search_knowledge_system_prompt.txt - Global search knowledge instruction
  • local_search_system_prompt.txt - Local search system prompt
  • basic_search_system_prompt.txt - Basic search system prompt
  • question_gen_system_prompt.txt - Question generation system prompt

​
Next steps

After initializing your project:
  1. Add your API key to the .env file: 
    GRAPHRAG_API_KEY=your-api-key-here
  2. Place your documents in the input/ directory
  3. Optionally customize settings.yaml for your use case
  4. Run the indexing pipeline: 
    graphrag index

​
Error handling

If you try to initialize a project that already exists, you’ll receive an error: 
Project already initialized at /path/to/project
Use the --force flag to overwrite the existing configuration.

Build docs developers (and LLMs) love

Get started for freeTalk to us