@artale/pi-pai 4.5.0 → 4.6.0

package/AGENTS.md

@@ -0,0 +1,32 @@
+# Context: You
+
+## Who You Are
+
+Customize this file with your own context — role, expertise, preferences.
+
+## Key Preferences
+
+- Concise responses for simple tasks
+- Verification over claims of completion
+- Direct language, no hedging
+- CLI-first approach
+- Simple solutions over clever ones
+- Evidence over claims
+
+## Voice Server (Optional)
+
+If you set up a TTS voice server, configure it here:
+
+```bash
+# Example: ElevenLabs or local TTS at a given endpoint
+curl -s -X POST http://localhost:8888/notify \
+  -H "Content-Type: application/json" \
+  -d '{"message": "MESSAGE_HERE", "voice_id": "YOUR_VOICE_ID", "voice_enabled": true}'
+```
+
+## Memory Location
+
+Persistent memory files at the configured memory directory:
+- `learning/` — Signals and patterns from interactions
+- `state/` — Current work tracking
+- `work/` — PRD files for Algorithm sessions

package/INSTALL.md

@@ -0,0 +1,224 @@
+# Installing PAI on Pi — Step by Step
+
+> **Note:** This is v1.0.0 (alpha) — an early release. Expect rough edges as this evolves.
+
+This guide takes you from zero to a working PAI system on Pi. Total time: ~15 minutes.
+
+## Prerequisites
+
+- **Node.js 18+** — [nodejs.org](https://nodejs.org/) or via your package manager
+- **A model provider** — at least one of:
+  - [Ollama](https://ollama.ai/) for local models (free, private)
+  - An API key for Anthropic, OpenAI, or OpenRouter
+
+## Step 1: Install Pi
+
+```bash
+npm install -g @mariozechner/pi-coding-agent
+```
+
+Verify it's installed:
+
+```bash
+pi --version
+```
+
+## Step 2: Set Up a Model Provider
+
+### Option A: Ollama (Local Models — Free)
+
+```bash
+# Install Ollama
+curl -fsSL https://ollama.ai/install.sh | sh
+
+# Pull a model (pick one)
+ollama pull llama3.1:8b      # Fast, good for simple tasks (~4.7GB)
+ollama pull llama3.1:70b     # Powerful, needs 48GB+ RAM (~40GB)
+ollama pull qwen2.5:32b      # Good balance of speed and quality (~19GB)
+ollama pull deepseek-r1:32b   # Strong reasoning (~19GB)
+
+# Ollama runs automatically at http://localhost:11434
+```
+
+### Option B: Cloud API (Anthropic, OpenAI, OpenRouter)
+
+Get an API key from your provider and have it ready for Step 4.
+
+## Step 3: Copy the PAI Scaffold
+
+```bash
+# Create the config directory
+mkdir -p ~/.config/PAI-pi
+
+# From this directory (Releases/Pi/), copy everything
+cp -r config/* ~/.config/PAI-pi/
+mkdir -p ~/.config/PAI-pi/extensions
+cp -r extensions/* ~/.config/PAI-pi/extensions/
+mkdir -p ~/.config/PAI-pi/skills
+cp -r skills/* ~/.config/PAI-pi/skills/
+mkdir -p ~/.config/PAI-pi/memory
+cp -r memory/* ~/.config/PAI-pi/memory/
+```
+
+## Step 4: Configure Your Model Provider
+
+Edit `~/.config/PAI-pi/models.json`. Here are working examples:
+
+### Ollama (Local)
+
+```json
+{
+  "providers": {
+    "ollama": {
+      "type": "ollama",
+      "baseUrl": "http://localhost:11434",
+      "models": {
+        "default": "llama3.1:8b"
+      }
+    }
+  }
+}
+```
+
+### Anthropic
+
+```json
+{
+  "providers": {
+    "anthropic": {
+      "type": "anthropic",
+      "apiKey": "sk-ant-...",
+      "models": {
+        "default": "claude-sonnet-4-6"
+      }
+    }
+  }
+}
+```
+
+### OpenRouter (Access Many Models)
+
+```json
+{
+  "providers": {
+    "openrouter": {
+      "type": "openai",
+      "baseUrl": "https://openrouter.ai/api/v1",
+      "apiKey": "sk-or-...",
+      "models": {
+        "default": "anthropic/claude-sonnet-4-6"
+      }
+    }
+  }
+}
+```
+
+### Multiple Providers
+
+```json
+{
+  "providers": {
+    "local": {
+      "type": "ollama",
+      "baseUrl": "http://localhost:11434",
+      "models": {
+        "default": "llama3.1:8b",
+        "thinking": "qwen2.5:32b"
+      }
+    },
+    "cloud": {
+      "type": "anthropic",
+      "apiKey": "sk-ant-...",
+      "models": {
+        "default": "claude-sonnet-4-6"
+      }
+    }
+  }
+}
+```
+
+## Step 5: Customize Your Identity
+
+Edit `~/.config/PAI-pi/SYSTEM.md` and replace the placeholders:
+
+- `{{YOUR_NAME}}` → Your name (e.g., "Alex")
+- `{{YOUR_AI_NAME}}` → Your AI's name (e.g., "Nova", "Atlas", "Friday")
+
+Also update:
+- The **Projects** table with your actual projects
+- The **Environment** section with your setup
+- Any **Behavioral Rules** you want to add or modify
+
+Edit `~/.config/PAI-pi/AGENTS.md` with your context and preferences.
+
+## Step 6: Launch
+
+```bash
+pi
+```
+
+You should see the PAI status line and your AI is ready.
+
+### Test the Algorithm
+
+Try a complex request to see the 7-phase Algorithm in action:
+
+```
+Create a TypeScript function that parses CSV files with proper error handling
+```
+
+Your AI should enter Algorithm mode, generate ISC criteria, and work through all 7 phases.
+
+### Test Skills
+
+```
+/status          # Check PAI system status
+/algorithm       # Force Algorithm mode for next task
+research quick: what is Pi coding agent
+```
+
+## Step 7: Optional — Voice Notifications
+
+If you want your AI to speak:
+
+1. Set up a TTS server (ElevenLabs, Coqui, or any HTTP-based TTS)
+2. Set environment variables:
+
+```bash
+export PAI_VOICE_ENABLED=true
+export PAI_VOICE_ENDPOINT=http://localhost:8888/notify
+export PAI_VOICE_ID=your-voice-id
+```
+
+3. Restart Pi. Voice will announce Algorithm phase transitions.
+
+## Troubleshooting
+
+### Pi can't find the extension
+
+Make sure the extension is at `~/.config/PAI-pi/extensions/pai-core/index.ts`. Pi looks for extensions in the `extensions/` subdirectory of its config path.
+
+### Ollama connection refused
+
+```bash
+# Check if Ollama is running
+curl http://localhost:11434/api/tags
+
+# Start it if not
+ollama serve
+```
+
+### Model too slow
+
+Use a smaller model for everyday tasks and a larger one for complex work. Configure multiple models in `models.json` and switch between them.
+
+### Skills not loading
+
+Skills are loaded from `~/.config/PAI-pi/skills/`. Each skill needs a `SKILL.md` file with proper frontmatter (the `---` delimited header).
+
+## Next Steps
+
+1. **Add your own skills** — Create new directories in `skills/` for your domains
+2. **Build your memory** — The `memory/` directory grows with usage
+3. **Customize the Algorithm** — Adjust effort levels and ISC rules in SYSTEM.md
+4. **Explore full PAI** — For the complete experience with 63+ skills, try [PAI v4.0.3 on Claude Code](../v4.0.3/)

package/README.md

@@ -1,139 +1,193 @@
-# π-PAI v4.2.0 — Personal AI Infrastructure for Pi
-
-A Pi Coding Agent extension implementing [Daniel Miessler's PAI framework](https://github.com/danielmiessler/Personal_AI_Infrastructure) (9.6K ⭐), synced with PAI v4.0.3 + Daniels official Pi release (v1.0.0). Includes the Ralph Wiggum iteration technique, damage control, and 5 features ported from Miessler's full system.
-
-## Install
-
-```bash
-pi install npm:@artale/pi-pai
-```
-
-## What's New in v4.0
-
-### Synced with Miessler's PAI v4.0.3
-
-| Feature | Source | Implementation |
-|---------|--------|---------------|
-| **v4 Algorithm** | PAI v4.0.3 | OBSERVE → PLAN → DECIDE → EXECUTE → VERIFY (was 7-phase, now 5) |
-| **Sentiment tracking** | PAI Observability | Every rating gets sentiment (positive/neutral/negative) + trend analysis |
-| **Agent personas** | PAI 14 agents | 7 personas: architect, engineer, pentester, designer, reviewer, researcher, qa |
-| **Self-evolution** | PAI self-upgrade | Detects repeating learning patterns (3+ occurrences), triggers `/pai evolve` |
-| **Plans convention** | PAI Plans dir | Auto-creates `.pi/plans/`, lists plans via `/pai plans` |
-
-## Commands
-
-### `/pai` — Goal-Driven Algorithm
-
-```bash
-# Setup
-/pai mission Build a profitable trading system
-/pai goal Deploy live stat-arb strategy
-/pai challenge Overfitting risk on historical data
-
-# Run the v4 algorithm loop
-/pai loop Deploy live strategy    # Start: OBSERVE
-/pai isc Strategy achieves Sharpe >1.5 on 5-year backtest
-/pai next Observed: spread mean-reverts at 3.2 std
-/pai next Plan: backtest 2020-2025, then paper trade 2 weeks
-/pai next Decision: go with mean-reversion, tight stops
-/pai next Executed: deployed to paper trading
-/pai next Verified: 58% win rate, Sharpe 1.8
-
-# Templates
-/pai template trading    # Pre-built mission + goals + challenges
-/pai template saas|devops|research|agent
-
-# Agent personas (NEW in v4)
-/pai agent architect Design the auth system for a multi-tenant SaaS
-/pai agent pentester Review this API for security vulnerabilities
-/pai agent designer Create the onboarding flow for mobile
-/pai agent reviewer Review the feature branch against main
-
-# Sentiment & trends (NEW in v4)
-/pai trend              # Rating trend: avg, recent, sentiment distribution
-/pai evolve             # Self-evolution: repeating pattern report
-
-# Other
-/pai plans              # List .pi/plans/ directory
-/pai status             # Full status with all v4 features
-/pai learn <insight>    # Record a learning
-/pai done g0            # Complete a goal
-/pai block g1           # Block a goal
-/pai reset              # Clear everything
-```
-
-### `/rate` — Sentiment-Aware Ratings
-
-```bash
-/rate 9 Clean architecture, fast execution  # → ⭐9 😊 positive
-/rate 3 Missed edge cases, slow             # → ⭐3 😞 negative → auto-captures learning
-/rate 6                                     # → ⭐6 😐 neutral
-```
-
-Ratings track: score (1-10), context, timestamp, **sentiment** (inferred from score + keywords).
-
-Widget shows trend: `⭐7.2 📈8.0 (15 ratings)` — improving/declining/stable.
-
-### `/ralph` — Deterministic Iteration
-
-```bash
-/ralph Build a REST API with auth, tests, and docs
-# Agent iterates up to 50 times until RALPH_DONE
-/ralph stop
-```
-
-### 🛡️ Damage Control
-
-Guards via YAML rules (`damage-control-rules.yaml`):
-- **Blocked patterns:** `rm -rf`, `git reset --hard`, `git push --force`, `DROP TABLE`
-- **Confirm-first:** `git push --delete`, `git branch -D`
-- **Zero-access:** `.env`, `~/.ssh/`, `~/.aws/`, `*.pem`
-- **Read-only/No-delete:** configurable per project
-
-### 🔧 Agent Tools
-
-| Tool | Description |
-|------|-------------|
-| `pai_status` | Full status with v4 algorithm, trends, personas, patterns |
-| `pai_learn` | Record insight with sentiment |
-| `pai_rate` | Rate 1-10 with sentiment + trend tracking |
-
-### 📊 Live Widget
-
-```
-  🎯 Build a profitable trading system
-  Goals: 2⚡ 0🚫 1✓ │ 5 learnings │ ⭐7.2 📈8.0 (15)
-  Loop: ● ● ◉ ○ ○ [DECIDE] standard 42s
-  ⚠️ 2 repeating pattern(s) — consider /pai evolve
-```
-
-## Agent Personas
-
-| Persona | Focus |
-|---------|-------|
-| `architect` | System design, scalability, API design, failure modes |
-| `engineer` | Production code, error handling, types, tests |
-| `pentester` | Security: injection, auth bypass, SSRF, secrets, deps |
-| `designer` | UX/UI, accessibility, responsive, interaction patterns |
-| `reviewer` | Code review: correctness, edge cases, P1/P2/P3 findings |
-| `researcher` | Deep investigation, evidence-based, source citations |
-| `qa` | Test planning: happy paths, edge cases, boundaries, regression |
-
-## Lineage
-
-| Project | Author | What |
-|---------|--------|------|
-| [Personal AI Infrastructure](https://github.com/danielmiessler/Personal_AI_Infrastructure) | [Daniel Miessler](https://danielmiessler.com) | PAI v4.0.3 framework (9.6K ⭐) |
-| [pi-ralph](https://github.com/Whamp/pi-ralph) | Whamp | Ralph Wiggum iteration technique |
-| [pi-vs-claude-code](https://github.com/disler/pi-vs-claude-code) | disler | Damage control + extension patterns |
-| [Pi Coding Agent](https://github.com/badlogic/pi) | Mario Zechner | The platform |
-
-## v3 → v4 Migration
-
-The algorithm changed from 7 phases to 5:
-- **Old:** OBSERVE → THINK → PLAN → DEFINE → EXECUTE → MEASURE → LEARN
-- **New:** OBSERVE → PLAN → DECIDE → EXECUTE → VERIFY
-
-THINK+PLAN merged into PLAN. DEFINE became DECIDE. MEASURE+LEARN merged into VERIFY.
-
+# π-PAI v4.2.0 — Personal AI Infrastructure for Pi
+
+A Pi Coding Agent extension implementing [Daniel Miessler's PAI framework](https://github.com/danielmiessler/Personal_AI_Infrastructure) (9.6K ⭐), synced with PAI v4.0.3 + Daniels official Pi release (v1.0.0). Includes the Ralph Wiggum iteration technique, damage control, and 5 features ported from Miessler's full system.
+
+## Install
+
+```bash
+pi install npm:@artale/pi-pai
+```
+
+## What's New in v4.0
+
+### Synced with Miessler's PAI v4.0.3
+
+| Feature | Source | Implementation |
+|---------|--------|---------------|
+| **v4 Algorithm** | PAI v4.0.3 | OBSERVE → PLAN → DECIDE → EXECUTE → VERIFY (was 7-phase, now 5) |
+| **Sentiment tracking** | PAI Observability | Every rating gets sentiment (positive/neutral/negative) + trend analysis |
+| **Agent personas** | PAI 14 agents | 7 personas: architect, engineer, pentester, designer, reviewer, researcher, qa |
+| **Self-evolution** | PAI self-upgrade | Detects repeating learning patterns (3+ occurrences), triggers `/pai evolve` |
+| **Plans convention** | PAI Plans dir | Auto-creates `.pi/plans/`, lists plans via `/pai plans` |
+
+## Commands
+
+### `/pai` — Goal-Driven Algorithm
+
+```bash
+# Setup
+/pai mission Build a profitable trading system
+/pai goal Deploy live stat-arb strategy
+/pai challenge Overfitting risk on historical data
+
+# Run the v4 algorithm loop
+/pai loop Deploy live strategy    # Start: OBSERVE
+/pai isc Strategy achieves Sharpe >1.5 on 5-year backtest
+/pai next Observed: spread mean-reverts at 3.2 std
+/pai next Plan: backtest 2020-2025, then paper trade 2 weeks
+/pai next Decision: go with mean-reversion, tight stops
+/pai next Executed: deployed to paper trading
+/pai next Verified: 58% win rate, Sharpe 1.8
+
+# Templates
+/pai template trading    # Pre-built mission + goals + challenges
+/pai template saas|devops|research|agent
+
+# Agent personas (NEW in v4)
+/pai agent architect Design the auth system for a multi-tenant SaaS
+/pai agent pentester Review this API for security vulnerabilities
+/pai agent designer Create the onboarding flow for mobile
+/pai agent reviewer Review the feature branch against main
+
+# Sentiment & trends (NEW in v4)
+/pai trend              # Rating trend: avg, recent, sentiment distribution
+/pai evolve             # Self-evolution: repeating pattern report
+
+# Other
+/pai plans              # List .pi/plans/ directory
+/pai status             # Full status with all v4 features
+/pai learn <insight>    # Record a learning
+/pai done g0            # Complete a goal
+/pai block g1           # Block a goal
+/pai reset              # Clear everything
+```
+
+### `/rate` — Sentiment-Aware Ratings
+
+```bash
+/rate 9 Clean architecture, fast execution  # → ⭐9 😊 positive
+/rate 3 Missed edge cases, slow             # → ⭐3 😞 negative → auto-captures learning
+/rate 6                                     # → ⭐6 😐 neutral
+```
+
+Ratings track: score (1-10), context, timestamp, **sentiment** (inferred from score + keywords).
+
+Widget shows trend: `⭐7.2 📈8.0 (15 ratings)` — improving/declining/stable.
+
+### `/ralph` — Deterministic Iteration
+
+```bash
+/ralph Build a REST API with auth, tests, and docs
+# Agent iterates up to 50 times until RALPH_DONE
+/ralph stop
+```
+
+### 🛡️ Damage Control
+
+Guards via YAML rules (`damage-control-rules.yaml`):
+- **Blocked patterns:** `rm -rf`, `git reset --hard`, `git push --force`, `DROP TABLE`
+- **Confirm-first:** `git push --delete`, `git branch -D`
+- **Zero-access:** `.env`, `~/.ssh/`, `~/.aws/`, `*.pem`
+- **Read-only/No-delete:** configurable per project
+
+### 🔧 Agent Tools
+
+| Tool | Description |
+|------|-------------|
+| `pai_status` | Full status with v4 algorithm, trends, personas, patterns |
+| `pai_learn` | Record insight with sentiment |
+| `pai_rate` | Rate 1-10 with sentiment + trend tracking |
+
+### 📊 Live Widget
+
+```
+  🎯 Build a profitable trading system
+  Goals: 2⚡ 0🚫 1✓ │ 5 learnings │ ⭐7.2 📈8.0 (15)
+  Loop: ● ● ◉ ○ ○ [DECIDE] standard 42s
+  ⚠️ 2 repeating pattern(s) — consider /pai evolve
+```
+
+## Agent Personas
+
+| Persona | Focus |
+|---------|-------|
+| `architect` | System design, scalability, API design, failure modes |
+| `engineer` | Production code, error handling, types, tests |
+| `pentester` | Security: injection, auth bypass, SSRF, secrets, deps |
+| `designer` | UX/UI, accessibility, responsive, interaction patterns |
+| `reviewer` | Code review: correctness, edge cases, P1/P2/P3 findings |
+| `researcher` | Deep investigation, evidence-based, source citations |
+| `qa` | Test planning: happy paths, edge cases, boundaries, regression |
+
+## Lineage
+
+| Project | Author | What |
+|---------|--------|------|
+| [Personal AI Infrastructure](https://github.com/danielmiessler/Personal_AI_Infrastructure) | [Daniel Miessler](https://danielmiessler.com) | PAI v4.0.3 framework (9.6K ⭐) |
+| [pi-ralph](https://github.com/Whamp/pi-ralph) | Whamp | Ralph Wiggum iteration technique |
+| [pi-vs-claude-code](https://github.com/disler/pi-vs-claude-code) | disler | Damage control + extension patterns |
+| [Pi Coding Agent](https://github.com/badlogic/pi) | Mario Zechner | The platform |
+
+## v3 → v4 Migration
+
+The algorithm changed from 7 phases to 5:
+- **Old:** OBSERVE → THINK → PLAN → DEFINE → EXECUTE → MEASURE → LEARN
+- **New:** OBSERVE → PLAN → DECIDE → EXECUTE → VERIFY
+
+THINK+PLAN merged into PLAN. DEFINE became DECIDE. MEASURE+LEARN merged into VERIFY.
+
 Active loops will reset on upgrade. All other state (goals, learnings, ratings) carries forward.
+
+## LifeOS Scaffold Compatibility (from `Releases/Pi`)
+
+This repo also ships the LifeOS Pi scaffold as a compatibility layer and migration path.
+
+### Added in this port
+
+- `SYSTEM.md` and `AGENTS.md` now match the canonical LifeOS scaffold policies.
+- Added root `models.json` + `settings.json` plus `VERSION` (`1.0.0`).
+- Added `INSTALL.md` with the full LifeOS install flow.
+- Added `src/lifeos-pai-core.ts` (a compat extension implementing:
+  - `voice_notify` tool
+  - `prd_create` / `prd_update` tools
+  - `lifeos_capture_learning` tool
+  - optional session voice hooks)
+- Added `memory/learning`, `memory/state`, `memory/work` with `.gitkeep` placeholders.
+- Added canonical 9 skill `SKILL.md` files under `skills/` from LifeOS.
+
+### Merge strategy (recommended)
+
+- **Primary runtime/experience remains this repo's enhanced `v4` PAI loop** (`/pai`, personas, ratings, telemetry, /rate sentiment, damage control).
+- Load `src/lifeos-pai-core.ts` in parallel for LifeOS-specific features when desired.
+- Keep `.pai` variants as rollback references:
+  - `SYSTEM.md.pai`
+  - `INSTALL.md.pai`
+  - `models.json.pai`
+  - `settings.json.pai`
+
+### Scaffold quick tree
+
+```text
+Releases/Pi/
+├── config/
+│   ├── AGENTS.md
+│   ├── SYSTEM.md
+│   ├── settings.json
+│   └── models.json
+├── extensions/
+│   └── pai-core/index.ts   -> src/lifeos-pai-core.ts
+├── skills/
+│   ├── agents/SKILL.md
+│   ├── content-analysis/SKILL.md
+│   ├── investigation/SKILL.md
+│   ├── media/SKILL.md
+│   ├── research/SKILL.md
+│   ├── scraping/SKILL.md
+│   ├── security/SKILL.md
+│   ├── telos/SKILL.md
+│   └── thinking/SKILL.md
+└── memory/
+    ├── learning/
+    ├── state/
+    └── work/
+```