Skip to content

Configuration

Jump to bottom
github-actions[bot] edited this page Feb 12, 2026 · 4 revisions

Configuration

Complete guide to configuring Loki Mode.

Configuration Priority

Configuration is loaded in this order (later overrides earlier):

  1. Built-in defaults
  2. User-global config (~/.config/loki-mode/config.yaml)
  3. Project-local config (.loki/config.yaml)
  4. Environment variables

Config File Locations

Location Scope Path
Project Single project .loki/config.yaml
User All projects ~/.config/loki-mode/config.yaml

Initialize Configuration

# Create project config
loki config init

# View current config
loki config show

# Edit config
loki config edit

# Show config path
loki config path

Complete Configuration Schema

#===============================================================================
# Loki Mode Configuration
# Documentation: https://github.com/asklokesh/loki-mode/wiki/Configuration
#===============================================================================

#-------------------------------------------------------------------------------
# Core Settings
#-------------------------------------------------------------------------------
core:
  # Maximum retry attempts before giving up
  max_retries: 50

  # Base wait time between retries (seconds)
  base_wait: 60

  # Maximum wait time for exponential backoff (seconds)
  max_wait: 3600

  # Skip prerequisite checks (not recommended)
  skip_prereqs: false

#-------------------------------------------------------------------------------
# Dashboard Settings
#-------------------------------------------------------------------------------
dashboard:
  # Enable web dashboard
  enabled: true

  # Dashboard server port
  port: 57374

#-------------------------------------------------------------------------------
# API Server Settings (Legacy Node.js API - separate from dashboard)
#-------------------------------------------------------------------------------
api:
  # API server port (legacy Node.js server; dashboard uses LOKI_DASHBOARD_PORT)
  port: 57374

  # API server host (use 0.0.0.0 for external access)
  host: localhost

  # API authentication token (for remote access)
  token: ""

#-------------------------------------------------------------------------------
# Resource Monitoring
#-------------------------------------------------------------------------------
resources:
  # Check resources every N seconds
  check_interval: 300

  # CPU warning threshold (percentage)
  cpu_threshold: 80

  # Memory warning threshold (percentage)
  mem_threshold: 80

#-------------------------------------------------------------------------------
# Security Settings
#-------------------------------------------------------------------------------
security:
  # Require approval before autonomous execution
  staged_autonomy: false

  # Enable audit logging
  audit_log: false

  # Maximum concurrent agents
  max_parallel_agents: 10

  # Run in Docker sandbox
  sandbox_mode: false

  # Comma-separated allowed paths (empty = all)
  allowed_paths: ""

  # Comma-separated blocked commands
  blocked_commands: "rm -rf /,dd if=,mkfs,:(){ :|:& };:"

  # Enable prompt injection via HUMAN_INPUT.md
  prompt_injection: false

#-------------------------------------------------------------------------------
# Enterprise Features
#-------------------------------------------------------------------------------
enterprise:
  # Enable token-based authentication
  auth: false

  # Enable enterprise audit logging
  audit: false

#-------------------------------------------------------------------------------
# SDLC Phases (all enabled by default)
#-------------------------------------------------------------------------------
phases:
  unit_tests: true
  api_tests: true
  e2e_tests: true
  security: true
  integration: true
  code_review: true
  web_research: true
  performance: true
  accessibility: true
  regression: true
  uat: true

#-------------------------------------------------------------------------------
# Completion & Loop Control
#-------------------------------------------------------------------------------
completion:
  # Explicit stop condition text
  promise: ""

  # Maximum loop iterations
  max_iterations: 1000

  # Ignore ALL completion signals
  perpetual_mode: false

  # Completion Council settings (v5.25.0)
  council:
    # Enable the 3-member completion council
    enabled: true

    # Number of council members
    size: 3

    # Votes needed for completion decision
    threshold: 2

    # Check every N iterations
    check_interval: 5

    # Minimum iterations before council activates
    min_iterations: 3

    # Max iterations with no git changes before force-stop
    stagnation_limit: 5

#-------------------------------------------------------------------------------
# Model Selection & Routing
#-------------------------------------------------------------------------------
model:
  # Enable Haiku for fast tier
  allow_haiku: false

  # Prompt repetition for Haiku
  prompt_repetition: true

  # Confidence-based model routing
  confidence_routing: true

  # Autonomy level: perpetual, checkpoint, supervised
  autonomy_mode: perpetual

  # Context compaction interval
  compaction_interval: 25

#-------------------------------------------------------------------------------
# Parallel Workflows
#-------------------------------------------------------------------------------
parallel:
  # Enable git worktree parallelism
  enabled: false

  # Maximum parallel worktrees
  max_worktrees: 5

  # Maximum concurrent AI sessions
  max_sessions: 3

  # Run testing stream in parallel
  testing: true

  # Run documentation stream in parallel
  docs: true

  # Run blog stream in parallel
  blog: false

  # Auto-merge completed features
  auto_merge: true

#-------------------------------------------------------------------------------
# Complexity Tier
#-------------------------------------------------------------------------------
complexity:
  # Complexity tier: auto, simple, standard, complex
  tier: auto

#-------------------------------------------------------------------------------
# GitHub Integration
#-------------------------------------------------------------------------------
github:
  # Import open issues as tasks
  import: false

  # Create PR when feature complete
  pr: false

  # Sync status back to issues
  sync: false

  # Override repo detection (owner/repo)
  repo: ""

  # Filter by labels (comma-separated)
  labels: ""

  # Filter by milestone
  milestone: ""

  # Filter by assignee
  assignee: ""

  # Max issues to import
  limit: 100

  # Label for PRs
  pr_label: ""

#-------------------------------------------------------------------------------
# Notifications
#-------------------------------------------------------------------------------
notifications:
  # Enable desktop notifications
  enabled: true

  # Play notification sounds
  sound: true

  # Slack incoming webhook URL
  slack_webhook: ""

  # Discord webhook URL
  discord_webhook: ""

  # Generic webhook URL
  webhook_url: ""

  # Active channels (all, slack, discord, webhook)
  channels: all

Configuration Examples

Minimal (Individual Developer)

# .loki/config.yaml
dashboard:
  enabled: true

notifications:
  slack_webhook: "https://hooks.slack.com/services/T00/B00/xxx"

Startup Team

# .loki/config.yaml
core:
  max_retries: 100

parallel:
  enabled: true
  max_worktrees: 3

github:
  import: true
  pr: true
  labels: "loki-mode"

notifications:
  enabled: true
  slack_webhook: "https://hooks.slack.com/..."
  discord_webhook: "https://discord.com/api/webhooks/..."

Enterprise

# .loki/config.yaml
security:
  staged_autonomy: true
  audit_log: true
  sandbox_mode: true
  allowed_paths: "/app/src,/app/tests"

enterprise:
  auth: true
  audit: true

model:
  autonomy_mode: checkpoint

notifications:
  enabled: true
  webhook_url: "https://monitoring.company.com/api/events"

CI/CD Pipeline

# .loki/config.yaml
core:
  max_retries: 20

dashboard:
  enabled: false

complexity:
  tier: simple

completion:
  max_iterations: 100

notifications:
  enabled: false

Environment Variable Override

Any config option can be overridden with environment variables:

# Pattern: LOKI_<SECTION>_<OPTION> (uppercase, underscores)

# Examples:
export LOKI_CORE_MAX_RETRIES=100
export LOKI_DASHBOARD_ENABLED=false
export LOKI_PARALLEL_ENABLED=true
export LOKI_GITHUB_IMPORT=true

Security Considerations

File Permissions

Config files with sensitive data should have restricted permissions:

chmod 600 .loki/config.yaml

Secrets Management

Never commit secrets to version control. Use environment variables for:

  • API tokens
  • Webhook URLs
  • Credentials
# .env (gitignored)
LOKI_SLACK_WEBHOOK=https://hooks.slack.com/...
LOKI_API_TOKEN=secret_token
# Load before running
source .env
loki start ./prd.md

CORS Configuration

For production deployments, restrict CORS origins:

export LOKI_DASHBOARD_CORS="https://dashboard.example.com,http://localhost:57374"

When not set, defaults to http://localhost:57374,http://127.0.0.1:57374 (localhost-only). See Security for details.

Symlink Protection

Loki Mode rejects symlinks in project paths to prevent path traversal attacks. Symlinks are allowed in home directory paths.

Navigation

Home

Getting Started

Reference

Features

Enterprise

Architecture

Resources

Version: 6.80.1

Autonomi | GitHub | npm

Clone this wiki locally