pi-config

Install pi-config via Docker container or the native installer script, configure prerequisites (GitHub CLI, API keys), and run your first pi session with the orchestrator

Walk through a complete feature implementation: ask pi to make a change, watch the orchestrator delegate to specialist agents, go through the code review loop, and create a PR

Learn how the orchestrator automatically routes tasks to the right specialist agent (python-expert, git-expert, github-expert, etc.) and when to override routing

Use the 3-reviewer parallel code review system (quality, guidelines, security), handle review findings, and iterate until all reviewers approve

Master the built-in slash commands (/implement, /pr-review, /review-local, /release, /external-ai, /scout-and-plan, /btw, /status) and understand how prompt templates work

Use the scored memory system to teach pi your preferences, lessons learned, and project conventions — including memory tools, dreaming consolidation, auto-extraction, and capacity management

Spawn async agents for long-running tasks, monitor their status with /async-status, set up recurring tasks with /cron, and understand fire-and-forget vs tracked agents

Set up P2P (/coms) and networked (/coms-net) communication between multiple pi sessions for collaborative multi-agent workflows

Route prompts to external AI CLIs via /external-ai for peer reviews, code generation, and multi-agent debates using ai-cli-runner

Deploy pi-config in a sandboxed Docker container with proper volume mounts, environment variables, host user mapping, and Docker socket access

Launch the pidash web dashboard to monitor sessions from a browser and use pidiff to view diffs with review comments in a rich UI

Copy-paste patterns for PR reviews, GitHub releases, review handling (autorabbit/autoqodo), implementing with review, scouting and planning, querying the review database, and refining pending reviews

Add a new specialist agent, create custom prompt templates, write reusable skills with /create-skill, configure project settings (.pi/pi-config-settings.json), and set up MCP servers via mcpl

Complete list of all 24 specialist agents with their names, descriptions, tools, model overrides, and when each is invoked

Full reference for all slash commands (/implement, /pr-review, /release, /review-handler, /external-ai, /btw, /status, /cron, /dream, /coms, /pidash, /pidiff, /repair) with arguments and options

Complete reference for the myk-pi-tools Python CLI: memory, pr, release, reviews, db, coderabbit, and ai-cli subcommands with all flags and options

All configuration options: pi-config-settings.json fields, environment variables (PI_CO_AUTHOR, PI_USE_WORKTREES, PI_DREAM_INTERVAL_HOURS, ACPX_AGENTS, GEMINI_API_KEY), rules loading order, and .coderabbit.yaml setup

Reference for all 13 orchestrator rules (00-60): core behavior, issue-first workflow, agent routing, MCP launchpad, code review loop, documentation updates, prompt templates, memory, critical rules, file preview, bug reporting, coms protocol, and task tracking

How the orchestrator extension registers tools, commands, and event hooks (session_start, before_agent_start, turn_end, input) — essential knowledge for writing new extensions or modifying existing ones

Deep dive into the stability-based scoring formula, vector embeddings (bge-small-en-v1.5), topic tree organization, situation report budget allocation, and the auto-injection pipeline

How to configure and use the ACPX provider extension to route pi's LLM requests through external agents (Cursor, Claude, Gemini, Copilot). Covers ACPX_AGENTS configuration, model discovery, session management, and how ACPX provider differs from /external-ai. Includes setup prerequisites (acpx runtime), available models per agent, and troubleshooting.

Set up and use the Discord bot built into the pidash server to remotely control pi sessions. Covers DISCORD_BOT_TOKEN and DISCORD_ALLOWED_USERS configuration, slash commands (/sessions, /status, /prompt, /stop), watching sessions, sending prompts and images via DM, receiving real-time output and ask_user prompts, and security considerations for allowed users.

Using pi inside Neovim terminal: automatic detection via NVIM environment variable, /nvim-changed-files command to send git changed files to the quickfix list, how pi communicates with the parent Neovim instance via RPC, and quickfix workflow for navigating review findings or changed files.

How to use pi's generate_image tool powered by the Gemini API. Covers setup (PI_IMAGE_MODEL and GEMINI_API_KEY environment variables), available parameters (subject, action, scene, composition, lighting, style, text, aspect_ratio), supported aspect ratios, automatic HTTP preview server in containers, and example prompts for different image styles.

How pi's enforcement system protects your environment: blocked commands (direct python/pip, git add ., remote script piping), git branch protection (protected branches, merged branch detection, detached HEAD), dangerous command confirmation prompts, repeat command detection, worktree enforcement, docker-safe wrapper in containers, and memory write restrictions for subagents. What to do when a command is blocked and how to work with the guardrails.