Skip to main content

Configuration

OWL's configuration is stored in ~/.owl/config.yaml.

Configuration File​

Create or edit ~/.owl/config.yaml:

llm:
  provider: ollama
  host: http://localhost:11434
  model: glm-4.6:cloud
  timeout: 60
  temperature: 0.7

daemon:
  socket_path: ~/.owl/owl.sock
  log_level: INFO

LLM Settings​

provider​

The LLM provider to use. Currently only ollama is supported.

llm:
  provider: ollama

host​

Ollama server address:

llm:
  host: http://localhost:11434

model​

The model to use for chat:

llm:
  model: glm-4.6:cloud  # Fast cloud model
  # model: llama2:latest
  # model: mistral:latest
  # model: codellama:latest

timeout​

Request timeout in seconds:

llm:
  timeout: 60  # Increase for slower models

temperature​

Sampling temperature (0.0 - 1.0):

llm:
  temperature: 0.7  # Lower = more focused, Higher = more creative

Daemon Settings​

socket_path​

Unix socket location:

daemon:
  socket_path: ~/.owl/owl.sock

log_level​

Logging verbosity:

daemon:
  log_level: INFO  # DEBUG, INFO, WARNING, ERROR

Environment Variables​

Override config values with environment variables:

# LLM settings
export OWL_LLM_HOST=http://localhost:11434
export OWL_LLM_MODEL=llama2:latest
export OWL_LLM_TIMEOUT=120
export OWL_LLM_PROVIDER=ollama

# Daemon settings
export OWL_LOG_LEVEL=DEBUG
export OWL_SOCKET_PATH=/custom/path/owl.sock

# Base directory (default: ~/.owl)
export OWL_HOME=/custom/owl/home

Data Directories​

OWL uses these directories (relative to OWL_HOME or ~/.owl):

PathPurpose
config.yamlConfiguration file
soul.yamlOWL's character and personality
memory.mdHuman-readable memory
memory/owl.dbSQLite database
knowledge/chroma/Vector database
owl.sockUnix socket (runtime)

Example Configurations​

Development (Fast)​

llm:
  provider: ollama
  host: http://localhost:11434
  model: glm-4.6:cloud
  timeout: 30
  temperature: 0.7

daemon:
  log_level: DEBUG

Production (Reliable)​

llm:
  provider: ollama
  host: http://localhost:11434
  model: llama2:70b
  timeout: 120
  temperature: 0.5

daemon:
  log_level: WARNING

Remote Ollama​

llm:
  provider: ollama
  host: http://<host-ip>:11434
  model: mixtral:latest
  timeout: 90
  temperature: 0.7

Changing Models at Runtime​

You can switch models without restarting:

# In the CLI
/status  # See current model

# Restart daemon with different model
# Terminal 1: Ctrl+C to stop owld
OWL_LLM_MODEL=mistral:latest owld