Configuration Reference

HiveMind OS is configured through a YAML file that controls daemon settings, AI providers, security, personas, and more.

Config File Location

The default config file is located at:

~/.hivemind/config.yaml

Override with the HIVEMIND_CONFIG_PATH environment variable:

HIVEMIND_CONFIG_PATH=/path/to/custom-config.yaml hive daemon start

Config Precedence

Settings are resolved in the following order (highest priority first):

1. Environment variables — HIVEMIND_CONFIG_PATH, HIVEMIND_HOME, and other HIVEMIND_* prefixed vars
2. Config file — ~/.hivemind/config.yaml
3. Defaults — built-in sensible defaults

Environment Variable Substitution

API keys and secrets can be injected from environment variables using the env: prefix in auth fields:

models:
  providers:
    - id: openai
      auth: "env:OPENAI_API_KEY"

Schema Overview

The config file has the following top-level sections:

Section	Description
daemon	Logging and event bus settings
api	HTTP API bind address and toggle
security	Override policies, prompt injection scanning, command policies, sandboxing
models	AI provider connections (models.providers[]) and timeout settings
local_models	On-device model inference configuration
embedding	Embedding model definitions and file-pattern rules
compaction	Context compaction settings
personas	Agent persona configurations (usually managed via persona YAML files)
skills	Skill configurations
afk	AFK (idle) mode settings
python	Managed Python runtime configuration
node	Managed Node.js runtime configuration
tool_limits	Per-tool rate and size limits
web_search	Web search configuration

Complete Example

# ~/.hivemind/config.yaml

# --- Daemon ---
daemon:
  log_level: info
  event_bus_capacity: 512

# --- API ---
api:
  bind: "127.0.0.1:9180"
  http_enabled: true

# --- AI Providers ---
models:
  request_timeout_secs: 60
  stream_timeout_secs: 120
  providers:
    - id: openai
      kind: open-ai-compatible
      name: OpenAI
      base_url: "https://api.openai.com/v1"
      auth: "env:OPENAI_API_KEY"
      models:
        - gpt-4o
        - gpt-4o-mini
      channel_class: public
      priority: 100
      enabled: true

    - id: anthropic
      kind: anthropic
      name: Anthropic
      base_url: "https://api.anthropic.com"
      auth: "env:ANTHROPIC_API_KEY"
      models:
        - claude-sonnet-4-20250514
      channel_class: public
      priority: 90

    - id: local-ollama
      kind: ollama-local
      name: Local Ollama
      models:
        - llama3.1:70b
      channel_class: local-only
      priority: 50

# --- Local Models ---
local_models:
  enabled: false
  storage_path: ~/.hivemind/models
  max_loaded_models: 2
  max_download_concurrent: 1
  auto_evict: true
  isolate_runtimes: false

# --- Security ---
security:
  override_policy:
    internal: prompt
    confidential: prompt
    restricted: block
  prompt_injection:
    enabled: true
    action_on_detection: prompt
    confidence_threshold: 0.7
  command_policy:
    enabled: true
  sandbox:
    enabled: true
    allow_network: true

# --- Embedding ---
embedding:
  default_model: bge-small-en-v1.5

TIP

Run hive config show to view the current configuration, or hive config validate to check your config for errors. Changes to the config file are picked up automatically — no restart needed.

WARNING

Never commit API keys directly in the config file. Always use environment variable substitution (env:VAR_NAME) or a secrets manager.

See Also

• CLI Overview — global flags that override config