feat(ansible): Canonical IaC playbook for fleet management

Implements the Ansible Infrastructure as Code story from KT 2026-04-08.

One canonical Ansible playbook defines:
- Deadman switch (snapshot good config on health, rollback+restart on death)
- Golden state config deployment (Anthropic BANNED, Kimi→Gemini→Ollama)
- Cron schedule (source-controlled, no manual crontab edits)
- Agent startup sequence (pull→validate→start→verify)
- request_log telemetry table (every inference call logged)
- Thin config pattern (immutable local pointer to upstream)
- Gitea webhook handler (deploy on merge)
- Config validator (rejects banned providers)

Fleet inventory: Timmy (Mac), Allegro (VPS), Bezalel (VPS), Ezra (VPS)

Roles: wizard_base, golden_state, deadman_switch, request_log, cron_manager

Addresses: timmy-config #442, #443, #444, #445, #446
References: KT Final 2026-04-08 P2, KT Bezalel 2026-04-08 #1-#5

.gitea/workflows/validate-config.yaml

@@ -0,0 +1,134 @@
+# validate-config.yaml
+# Validates all config files, scripts, and playbooks on every PR.
+# Addresses #289: repo-native validation for timmy-config changes.
+#
+# Runs: YAML lint, Python syntax check, shell lint, JSON validation,
+#       deploy script dry-run, and cron syntax verification.
+
+name: Validate Config
+
+on:
+  pull_request:
+    branches: [main]
+  push:
+    branches: [main]
+
+jobs:
+  yaml-lint:
+    name: YAML Lint
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Install yamllint
+        run: pip install yamllint
+      - name: Lint YAML files
+        run: |
+          find . -name '*.yaml' -o -name '*.yml' | \
+            grep -v '.gitea/workflows' | \
+            xargs -r yamllint -d '{extends: relaxed, rules: {line-length: {max: 200}}}'
+
+  json-validate:
+    name: JSON Validate
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Validate JSON files
+        run: |
+          find . -name '*.json' -print0 | while IFS= read -r -d '' f; do
+            echo "Validating: $f"
+            python3 -m json.tool "$f" > /dev/null || exit 1
+          done
+
+  python-check:
+    name: Python Syntax & Import Check
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+      - name: Install dependencies
+        run: |
+          pip install py_compile flake8
+      - name: Compile-check all Python files
+        run: |
+          find . -name '*.py' -print0 | while IFS= read -r -d '' f; do
+            echo "Checking: $f"
+            python3 -m py_compile "$f" || exit 1
+          done
+      - name: Flake8 critical errors only
+        run: |
+          flake8 --select=E9,F63,F7,F82 --show-source --statistics \
+            scripts/ allegro/ cron/ || true
+
+  shell-lint:
+    name: Shell Script Lint
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Install shellcheck
+        run: sudo apt-get install -y shellcheck
+      - name: Lint shell scripts
+        run: |
+          find . -name '*.sh' -print0 | xargs -0 -r shellcheck --severity=error || true
+
+  cron-validate:
+    name: Cron Syntax Check
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Validate cron entries
+        run: |
+          if [ -d cron ]; then
+            find cron -name '*.cron' -o -name '*.crontab' | while read f; do
+              echo "Checking cron: $f"
+              # Basic syntax validation
+              while IFS= read -r line; do
+                [[ "$line" =~ ^#.*$ ]] && continue
+                [[ -z "$line" ]] && continue
+                fields=$(echo "$line" | awk '{print NF}')
+                if [ "$fields" -lt 6 ]; then
+                  echo "ERROR: Too few fields in $f: $line"
+                  exit 1
+                fi
+              done < "$f"
+            done
+          fi
+
+  deploy-dry-run:
+    name: Deploy Script Dry Run
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Syntax-check deploy.sh
+        run: |
+          if [ -f deploy.sh ]; then
+            bash -n deploy.sh
+            echo "deploy.sh syntax OK"
+          fi
+
+  playbook-schema:
+    name: Playbook Schema Validation
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Validate playbook structure
+        run: |
+          python3 -c "
+import yaml, sys, glob
+required_keys = {'name', 'description'}
+for f in glob.glob('playbooks/*.yaml'):
+    with open(f) as fh:
+        try:
+            data = yaml.safe_load(fh)
+            if not isinstance(data, dict):
+                print(f'ERROR: {f} is not a YAML mapping')
+                sys.exit(1)
+            missing = required_keys - set(data.keys())
+            if missing:
+                print(f'WARNING: {f} missing keys: {missing}')
+            print(f'OK: {f}')
+        except yaml.YAMLError as e:
+            print(f'ERROR: {f}: {e}')
+            sys.exit(1)
+"

ansible/BANNED_PROVIDERS.yml

@@ -0,0 +1,47 @@
+# =============================================================================
+# BANNED PROVIDERS — The Timmy Foundation
+# =============================================================================
+# "Anthropic is not only fired, but banned. I don't want these errors
+# cropping up." — Alexander, 2026-04-09
+#
+# This is a HARD BAN. Not deprecated. Not fallback. BANNED.
+# Enforcement: pre-commit hook, linter, Ansible validation, CI tests.
+# =============================================================================
+
+banned_providers:
+  - name: anthropic
+    reason: "Permanently banned. SDK access gated despite active quota. Fleet was bricked because golden state pointed to Anthropic Sonnet."
+    banned_date: "2026-04-09"
+    enforcement: strict  # Ansible playbook FAILS if detected
+    models:
+      - "claude-sonnet-*"
+      - "claude-opus-*"
+      - "claude-haiku-*"
+      - "claude-*"
+    endpoints:
+      - "api.anthropic.com"
+      - "anthropic/*"  # OpenRouter pattern
+    api_keys:
+      - "ANTHROPIC_API_KEY"
+      - "CLAUDE_API_KEY"
+
+# Golden state alternative:
+approved_providers:
+  - name: kimi-coding
+    model: kimi-k2.5
+    role: primary
+  - name: openrouter
+    model: google/gemini-2.5-pro
+    role: fallback
+  - name: ollama
+    model: "gemma4:latest"
+    role: terminal_fallback
+
+# Future evaluation:
+evaluation_candidates:
+  - name: mimo-v2-pro
+    status: pending
+    notes: "Free via Nous Portal for ~2 weeks from 2026-04-07. Add after fallback chain is fixed."
+  - name: hermes-4
+    status: available
+    notes: "Free on Nous Portal. 36B and 70B variants. Home team model."

ansible/README.md

@@ -0,0 +1,95 @@
+# Ansible IaC — The Timmy Foundation Fleet
+
+> One canonical Ansible playbook defines: deadman switch, cron schedule,
+> golden state rollback, agent startup sequence.
+> — KT Final Session 2026-04-08, Priority TWO
+
+## Purpose
+
+This directory contains the **single source of truth** for fleet infrastructure.
+No more ad-hoc recovery implementations. No more overlapping deadman switches.
+No more agents mutating their own configs into oblivion.
+
+**Everything** goes through Ansible. If it's not in a playbook, it doesn't exist.
+
+## Architecture
+
+```
+┌─────────────────────────────────────────────────┐
+│                  Gitea (Source of Truth)          │
+│  timmy-config/ansible/                           │
+│    ├── inventory/hosts.yml    (fleet machines)    │
+│    ├── playbooks/site.yml     (master playbook)   │
+│    ├── roles/                 (reusable roles)    │
+│    └── group_vars/wizards.yml (golden state)      │
+└──────────────────┬──────────────────────────────┘
+                   │  PR merge triggers webhook
+                   ▼
+┌─────────────────────────────────────────────────┐
+│              Gitea Webhook Handler                │
+│  scripts/deploy_on_webhook.sh                     │
+│  → ansible-pull on each target machine            │
+└──────────────────┬──────────────────────────────┘
+                   │  ansible-pull
+                   ▼
+┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐
+│  Timmy   │  │ Allegro  │  │ Bezalel  │  │  Ezra    │
+│  (Mac)   │  │  (VPS)   │  │  (VPS)   │  │  (VPS)   │
+│          │  │          │  │          │  │          │
+│ deadman  │  │ deadman  │  │ deadman  │  │ deadman  │
+│ cron     │  │ cron     │  │ cron     │  │ cron     │
+│ golden   │  │ golden   │  │ golden   │  │ golden   │
+│ req_log  │  │ req_log  │  │ req_log  │  │ req_log  │
+└──────────┘  └──────────┘  └──────────┘  └──────────┘
+```
+
+## Quick Start
+
+```bash
+# Deploy everything to all machines
+ansible-playbook -i inventory/hosts.yml playbooks/site.yml
+
+# Deploy only golden state config
+ansible-playbook -i inventory/hosts.yml playbooks/golden_state.yml
+
+# Deploy only to a specific wizard
+ansible-playbook -i inventory/hosts.yml playbooks/site.yml --limit bezalel
+
+# Dry run (check mode)
+ansible-playbook -i inventory/hosts.yml playbooks/site.yml --check --diff
+```
+
+## Golden State Provider Chain
+
+All wizard configs converge on this provider chain. **Anthropic is BANNED.**
+
+| Priority | Provider             | Model            | Endpoint                          |
+| -------- | -------------------- | ---------------- | --------------------------------- |
+| 1        | Kimi                 | kimi-k2.5        | https://api.kimi.com/coding/v1    |
+| 2        | Gemini (OpenRouter)  | gemini-2.5-pro   | https://openrouter.ai/api/v1      |
+| 3        | Ollama (local)       | gemma4:latest    | http://localhost:11434/v1         |
+
+## Roles
+
+| Role             | Purpose                                                      |
+| ---------------- | ------------------------------------------------------------ |
+| `wizard_base`    | Common wizard setup: directories, thin config, git pull      |
+| `deadman_switch` | Health check → snapshot good config → rollback on death      |
+| `golden_state`   | Deploy and enforce golden state provider chain               |
+| `request_log`    | SQLite telemetry table for every inference call               |
+| `cron_manager`   | Source-controlled cron jobs — no manual crontab edits         |
+
+## Rules
+
+1. **No manual changes.** If it's not in a playbook, it will be overwritten.
+2. **No Anthropic.** Banned. Enforcement is automated. See `BANNED_PROVIDERS.yml`.
+3. **Idempotent.** Every playbook can run 100 times with the same result.
+4. **PR required.** Config changes go through Gitea PR review, then deploy.
+5. **One identity per machine.** No duplicate agents. Fleet audit enforces this.
+
+## Related Issues
+
+- timmy-config #442: [P2] Ansible IaC Canonical Playbook
+- timmy-config #444: Wire Deadman Switch ACTION
+- timmy-config #443: Thin Config Pattern
+- timmy-config #446: request_log Telemetry Table

ansible/ansible.cfg

@@ -0,0 +1,21 @@
+[defaults]
+inventory = inventory/hosts.yml
+roles_path = roles
+host_key_checking = False
+retry_files_enabled = False
+stdout_callback = yaml
+forks = 10
+timeout = 30
+
+# Logging
+log_path = /var/log/ansible/timmy-fleet.log
+
+[privilege_escalation]
+become = True
+become_method = sudo
+become_user = root
+become_ask_pass = False
+
+[ssh_connection]
+pipelining = True
+ssh_args = -o ControlMaster=auto -o ControlPersist=60s -o StrictHostKeyChecking=no

ansible/inventory/group_vars/wizards.yml

@@ -0,0 +1,74 @@
+# =============================================================================
+# Wizard Group Variables — Golden State Configuration
+# =============================================================================
+# These variables are applied to ALL wizards in the fleet.
+# This IS the golden state. If a wizard deviates, Ansible corrects it.
+# =============================================================================
+
+# --- Deadman Switch ---
+deadman_enabled: true
+deadman_check_interval: 300    # 5 minutes between health checks
+deadman_snapshot_dir: "~/.local/timmy/snapshots"
+deadman_max_snapshots: 10      # Rolling window of good configs
+deadman_restart_cooldown: 60   # Seconds to wait before restart after failure
+deadman_max_restart_attempts: 3
+deadman_escalation_channel: telegram  # Alert Alexander after max attempts
+
+# --- Thin Config ---
+thin_config_path: "~/.timmy/thin_config.yml"
+thin_config_mode: "0444"       # Read-only — agents CANNOT modify
+upstream_repo: "https://forge.alexanderwhitestone.com/Timmy_Foundation/timmy-config.git"
+upstream_branch: main
+config_pull_on_wake: true
+config_validation_enabled: true
+
+# --- Agent Settings ---
+agent_max_turns: 30
+agent_reasoning_effort: high
+agent_verbose: false
+agent_approval_mode: auto
+
+# --- Hermes Harness ---
+hermes_config_dir: "{{ hermes_home }}"
+hermes_bin_dir: "{{ hermes_home }}/bin"
+hermes_skins_dir: "{{ hermes_home }}/skins"
+hermes_playbooks_dir: "{{ hermes_home }}/playbooks"
+hermes_memories_dir: "{{ hermes_home }}/memories"
+
+# --- Request Log (Telemetry) ---
+request_log_enabled: true
+request_log_path: "~/.local/timmy/request_log.db"
+request_log_rotation_days: 30  # Archive logs older than 30 days
+request_log_sync_to_gitea: false  # Future: push telemetry summaries to Gitea
+
+# --- Cron Schedule ---
+# All cron jobs are managed here. No manual crontab edits.
+cron_jobs:
+  - name: "Deadman health check"
+    job: "cd {{ wizard_home }}/workspace/timmy-config && python3 fleet/health_check.py"
+    minute: "*/5"
+    hour: "*"
+    enabled: "{{ deadman_enabled }}"
+
+  - name: "Muda audit"
+    job: "cd {{ wizard_home }}/workspace/timmy-config && bash fleet/muda-audit.sh >> /tmp/muda-audit.log 2>&1"
+    minute: "0"
+    hour: "21"
+    weekday: "0"
+    enabled: true
+
+  - name: "Config pull from upstream"
+    job: "cd {{ wizard_home }}/workspace/timmy-config && git pull --ff-only origin main"
+    minute: "*/15"
+    hour: "*"
+    enabled: "{{ config_pull_on_wake }}"
+
+  - name: "Request log rotation"
+    job: "python3 -c \"import sqlite3,datetime; db=sqlite3.connect('{{ request_log_path }}'); db.execute('DELETE FROM request_log WHERE timestamp < datetime(\\\"now\\\", \\\"-{{ request_log_rotation_days }} days\\\")'); db.commit()\""
+    minute: "0"
+    hour: "3"
+    enabled: "{{ request_log_enabled }}"
+
+# --- Provider Enforcement ---
+# These are validated on every Ansible run. Any Anthropic reference = failure.
+provider_ban_enforcement: strict  # strict = fail playbook, warn = log only

ansible/inventory/hosts.yml

@@ -0,0 +1,119 @@
+# =============================================================================
+# Fleet Inventory — The Timmy Foundation
+# =============================================================================
+# Source of truth for all machines in the fleet.
+# Update this file when machines are added/removed.
+# All changes go through PR review.
+# =============================================================================
+
+all:
+  children:
+    wizards:
+      hosts:
+        timmy:
+          ansible_host: localhost
+          ansible_connection: local
+          wizard_name: Timmy
+          wizard_role: "Primary wizard — soul of the fleet"
+          wizard_provider_primary: kimi-coding
+          wizard_model_primary: kimi-k2.5
+          hermes_port: 8081
+          api_port: 8645
+          wizard_home: "{{ ansible_env.HOME }}/wizards/timmy"
+          hermes_home: "{{ ansible_env.HOME }}/.hermes"
+          machine_type: mac
+          # Timmy runs on Alexander's M3 Max
+          ollama_available: true
+
+        allegro:
+          ansible_host: 167.99.126.228
+          ansible_user: root
+          wizard_name: Allegro
+          wizard_role: "Kimi-backed third wizard house — tight coding tasks"
+          wizard_provider_primary: kimi-coding
+          wizard_model_primary: kimi-k2.5
+          hermes_port: 8081
+          api_port: 8645
+          wizard_home: /root/wizards/allegro
+          hermes_home: /root/.hermes
+          machine_type: vps
+          ollama_available: false
+
+        bezalel:
+          ansible_host: 159.203.146.185
+          ansible_user: root
+          wizard_name: Bezalel
+          wizard_role: "Forge-and-testbed wizard — infrastructure, deployment, hardening"
+          wizard_provider_primary: kimi-coding
+          wizard_model_primary: kimi-k2.5
+          hermes_port: 8081
+          api_port: 8656
+          wizard_home: /root/wizards/bezalel
+          hermes_home: /root/.hermes
+          machine_type: vps
+          ollama_available: false
+          # NOTE: The awake Bezalel may be the duplicate.
+          # Fleet audit (the-nexus #1144) will resolve identity.
+
+        ezra:
+          ansible_host: 143.198.27.163
+          ansible_user: root
+          wizard_name: Ezra
+          wizard_role: "Infrastructure wizard — Gitea, nginx, hosting"
+          wizard_provider_primary: kimi-coding
+          wizard_model_primary: kimi-k2.5
+          hermes_port: 8081
+          api_port: 8645
+          wizard_home: /root/wizards/ezra
+          hermes_home: /root/.hermes
+          machine_type: vps
+          ollama_available: false
+          # NOTE: Currently DOWN — Telegram key revoked, awaiting propagation.
+
+    # Infrastructure hosts (not wizards, but managed by Ansible)
+    infrastructure:
+      hosts:
+        forge:
+          ansible_host: 143.198.27.163
+          ansible_user: root
+          # Gitea runs on the same box as Ezra
+          gitea_url: https://forge.alexanderwhitestone.com
+          gitea_org: Timmy_Foundation
+
+  vars:
+    # Global variables applied to all hosts
+    gitea_repo_url: "https://forge.alexanderwhitestone.com/Timmy_Foundation/timmy-config.git"
+    gitea_branch: main
+    config_base_path: "{{ gitea_repo_url }}"
+    timmy_log_dir: "~/.local/timmy/fleet-health"
+    request_log_db: "~/.local/timmy/request_log.db"
+
+    # Golden state provider chain — Anthropic is BANNED
+    golden_state_providers:
+      - name: kimi-coding
+        model: kimi-k2.5
+        base_url: "https://api.kimi.com/coding/v1"
+        timeout: 120
+        reason: "Primary — Kimi K2.5 (best value, least friction)"
+      - name: openrouter
+        model: google/gemini-2.5-pro
+        base_url: "https://openrouter.ai/api/v1"
+        api_key_env: OPENROUTER_API_KEY
+        timeout: 120
+        reason: "Fallback — Gemini 2.5 Pro via OpenRouter"
+      - name: ollama
+        model: "gemma4:latest"
+        base_url: "http://localhost:11434/v1"
+        timeout: 180
+        reason: "Terminal fallback — local Ollama (sovereign, no API needed)"
+
+    # Banned providers — hard enforcement
+    banned_providers:
+      - anthropic
+      - claude
+    banned_models_patterns:
+      - "claude-*"
+      - "anthropic/*"
+      - "*sonnet*"
+      - "*opus*"
+      - "*haiku*"

ansible/playbooks/agent_startup.yml

@@ -0,0 +1,98 @@
+---
+# =============================================================================
+# agent_startup.yml — Resurrect Wizards from Checked-in Configs
+# =============================================================================
+# Brings wizards back online using golden state configs.
+# Order: pull config → validate → start agent → verify with request_log
+# =============================================================================
+
+- name: "Agent Startup Sequence"
+  hosts: wizards
+  become: true
+  serial: 1  # One wizard at a time to avoid cascading issues
+
+  tasks:
+    - name: "Pull latest config from upstream"
+      git:
+        repo: "{{ upstream_repo }}"
+        dest: "{{ wizard_home }}/workspace/timmy-config"
+        version: "{{ upstream_branch }}"
+        force: true
+      tags: [pull]
+
+    - name: "Deploy golden state config"
+      include_role:
+        name: golden_state
+      tags: [config]
+
+    - name: "Validate config — no banned providers"
+      shell: |
+        python3 -c "
+        import yaml, sys
+        with open('{{ wizard_home }}/config.yaml') as f:
+            cfg = yaml.safe_load(f)
+        banned = {{ banned_providers }}
+        for p in cfg.get('fallback_providers', []):
+            if p.get('provider', '') in banned:
+                print(f'BANNED: {p[\"provider\"]}', file=sys.stderr)
+                sys.exit(1)
+        model = cfg.get('model', {}).get('provider', '')
+        if model in banned:
+            print(f'BANNED default provider: {model}', file=sys.stderr)
+            sys.exit(1)
+        print('Config validated — no banned providers.')
+        "
+      register: config_valid
+      tags: [validate]
+
+    - name: "Ensure hermes-agent service is running"
+      systemd:
+        name: "hermes-{{ wizard_name | lower }}"
+        state: started
+        enabled: true
+      when: machine_type == 'vps'
+      tags: [start]
+      ignore_errors: true  # Service may not exist yet on all machines
+
+    - name: "Start hermes agent (Mac — launchctl)"
+      shell: |
+        launchctl kickstart -k "ai.hermes.{{ wizard_name | lower }}" 2>/dev/null || \
+        cd {{ wizard_home }} && hermes agent start --daemon 2>&1 | tail -5
+      when: machine_type == 'mac'
+      tags: [start]
+      ignore_errors: true
+
+    - name: "Wait for agent to come online"
+      wait_for:
+        host: 127.0.0.1
+        port: "{{ api_port }}"
+        timeout: 60
+        state: started
+      tags: [verify]
+      ignore_errors: true
+
+    - name: "Verify agent is alive — check request_log for activity"
+      shell: |
+        sleep 10
+        python3 -c "
+        import sqlite3, sys
+        db = sqlite3.connect('{{ request_log_path }}')
+        cursor = db.execute('''
+            SELECT COUNT(*) FROM request_log
+            WHERE agent_name = '{{ wizard_name }}'
+            AND timestamp > datetime('now', '-5 minutes')
+        ''')
+        count = cursor.fetchone()[0]
+        if count > 0:
+            print(f'{{ wizard_name }} is alive — {count} recent inference calls logged.')
+        else:
+            print(f'WARNING: {{ wizard_name }} started but no telemetry yet.')
+        "
+      register: agent_status
+      tags: [verify]
+      ignore_errors: true
+
+    - name: "Report startup status"
+      debug:
+        msg: "{{ wizard_name }}: {{ agent_status.stdout | default('startup attempted') }}"
+      tags: [always]

ansible/playbooks/cron_schedule.yml

@@ -0,0 +1,15 @@
+---
+# =============================================================================
+# cron_schedule.yml — Source-Controlled Cron Jobs
+# =============================================================================
+# All cron jobs are defined in group_vars/wizards.yml.
+# This playbook deploys them. No manual crontab edits allowed.
+# =============================================================================
+
+- name: "Deploy Cron Schedule"
+  hosts: wizards
+  become: true
+
+  roles:
+    - role: cron_manager
+      tags: [cron, schedule]

ansible/playbooks/deadman_switch.yml

@@ -0,0 +1,17 @@
+---
+# =============================================================================
+# deadman_switch.yml — Deploy Deadman Switch to All Wizards
+# =============================================================================
+# The deadman watch already fires and detects dead agents.
+# This playbook wires the ACTION:
+#   - On healthy check: snapshot current config as "last known good"
+#   - On failed check: rollback config to snapshot, restart agent
+# =============================================================================
+
+- name: "Deploy Deadman Switch ACTION"
+  hosts: wizards
+  become: true
+
+  roles:
+    - role: deadman_switch
+      tags: [deadman, recovery]

ansible/playbooks/golden_state.yml

@@ -0,0 +1,30 @@
+---
+# =============================================================================
+# golden_state.yml — Deploy Golden State Config to All Wizards
+# =============================================================================
+# Enforces the golden state provider chain across the fleet.
+# Removes any Anthropic references. Deploys the approved provider chain.
+# =============================================================================
+
+- name: "Deploy Golden State Configuration"
+  hosts: wizards
+  become: true
+
+  roles:
+    - role: golden_state
+      tags: [golden, config]
+
+  post_tasks:
+    - name: "Verify golden state — no banned providers"
+      shell: |
+        grep -rci 'anthropic\|claude-sonnet\|claude-opus\|claude-haiku' \
+          {{ hermes_home }}/config.yaml \
+          {{ wizard_home }}/config.yaml 2>/dev/null || echo "0"
+      register: banned_count
+      changed_when: false
+
+    - name: "Report golden state status"
+      debug:
+        msg: >
+          {{ wizard_name }} golden state: {{ golden_state_providers | map(attribute='name') | list | join(' → ') }}.
+          Banned provider references: {{ banned_count.stdout | trim }}.

ansible/playbooks/request_log.yml

@@ -0,0 +1,15 @@
+---
+# =============================================================================
+# request_log.yml — Deploy Telemetry Table
+# =============================================================================
+# Creates the request_log SQLite table on all machines.
+# Every inference call writes a row. No exceptions. No summarizing.
+# =============================================================================
+
+- name: "Deploy Request Log Telemetry"
+  hosts: wizards
+  become: true
+
+  roles:
+    - role: request_log
+      tags: [telemetry, logging]

ansible/playbooks/site.yml

@@ -0,0 +1,72 @@
+---
+# =============================================================================
+# site.yml — Master Playbook for the Timmy Foundation Fleet
+# =============================================================================
+# This is the ONE playbook that defines the entire fleet state.
+# Run this and every machine converges to golden state.
+#
+# Usage:
+#   ansible-playbook -i inventory/hosts.yml playbooks/site.yml
+#   ansible-playbook -i inventory/hosts.yml playbooks/site.yml --limit bezalel
+#   ansible-playbook -i inventory/hosts.yml playbooks/site.yml --check --diff
+# =============================================================================
+
+- name: "Timmy Foundation Fleet — Full Convergence"
+  hosts: wizards
+  become: true
+
+  pre_tasks:
+    - name: "Validate no banned providers in golden state"
+      assert:
+        that:
+          - "item.name not in banned_providers"
+        fail_msg: "BANNED PROVIDER DETECTED: {{ item.name }} — Anthropic is permanently banned."
+        quiet: true
+      loop: "{{ golden_state_providers }}"
+      tags: [always]
+
+    - name: "Display target wizard"
+      debug:
+        msg: "Deploying to {{ wizard_name }} ({{ wizard_role }}) on {{ ansible_host }}"
+      tags: [always]
+
+  roles:
+    - role: wizard_base
+      tags: [base, setup]
+
+    - role: golden_state
+      tags: [golden, config]
+
+    - role: deadman_switch
+      tags: [deadman, recovery]
+
+    - role: request_log
+      tags: [telemetry, logging]
+
+    - role: cron_manager
+      tags: [cron, schedule]
+
+  post_tasks:
+    - name: "Final validation — scan for banned providers"
+      shell: |
+        grep -ri 'anthropic\|claude-sonnet\|claude-opus\|claude-haiku' \
+          {{ hermes_home }}/config.yaml \
+          {{ wizard_home }}/config.yaml \
+          {{ thin_config_path }} 2>/dev/null || true
+      register: banned_scan
+      changed_when: false
+      tags: [validation]
+
+    - name: "FAIL if banned providers found in deployed config"
+      fail:
+        msg: |
+          BANNED PROVIDER DETECTED IN DEPLOYED CONFIG:
+          {{ banned_scan.stdout }}
+          Anthropic is permanently banned. Fix the config and re-deploy.
+      when: banned_scan.stdout | length > 0
+      tags: [validation]
+
+    - name: "Deployment complete"
+      debug:
+        msg: "{{ wizard_name }} converged to golden state. Provider chain: {{ golden_state_providers | map(attribute='name') | list | join(' → ') }}"
+      tags: [always]

ansible/roles/cron_manager/tasks/main.yml

@@ -0,0 +1,55 @@
+---
+# =============================================================================
+# cron_manager/tasks — Source-Controlled Cron Jobs
+# =============================================================================
+# All cron jobs are defined in group_vars/wizards.yml.
+# No manual crontab edits. This is the only way to manage cron.
+# =============================================================================
+
+- name: "Deploy managed cron jobs"
+  cron:
+    name: "{{ item.name }}"
+    job: "{{ item.job }}"
+    minute: "{{ item.minute | default('*') }}"
+    hour: "{{ item.hour | default('*') }}"
+    day: "{{ item.day | default('*') }}"
+    month: "{{ item.month | default('*') }}"
+    weekday: "{{ item.weekday | default('*') }}"
+    state: "{{ 'present' if item.enabled else 'absent' }}"
+    user: "{{ ansible_user | default('root') }}"
+  loop: "{{ cron_jobs }}"
+  when: cron_jobs is defined
+
+- name: "Deploy deadman switch cron (fallback if systemd timer unavailable)"
+  cron:
+    name: "Deadman switch — {{ wizard_name }}"
+    job: "{{ wizard_home }}/deadman_action.sh >> {{ timmy_log_dir }}/deadman-{{ wizard_name }}.log 2>&1"
+    minute: "*/5"
+    hour: "*"
+    state: present
+    user: "{{ ansible_user | default('root') }}"
+  when: deadman_enabled and machine_type != 'vps'
+  # VPS machines use systemd timers instead
+
+- name: "Remove legacy cron jobs (cleanup)"
+  cron:
+    name: "{{ item }}"
+    state: absent
+    user: "{{ ansible_user | default('root') }}"
+  loop:
+    - "legacy-deadman-watch"
+    - "old-health-check"
+    - "backup-deadman"
+  ignore_errors: true
+
+- name: "List active cron jobs"
+  shell: "crontab -l 2>/dev/null | grep -v '^#' | grep -v '^$' || echo 'No cron jobs found.'"
+  register: active_crons
+  changed_when: false
+
+- name: "Report cron status"
+  debug:
+    msg: |
+      {{ wizard_name }} cron jobs deployed.
+      Active:
+      {{ active_crons.stdout }}

ansible/roles/deadman_switch/tasks/main.yml

@@ -0,0 +1,70 @@
+---
+# =============================================================================
+# deadman_switch/tasks — Wire the Deadman Switch ACTION
+# =============================================================================
+# The watch fires. This makes it DO something:
+#   - On healthy check: snapshot current config as "last known good"
+#   - On failed check: rollback to last known good, restart agent
+# =============================================================================
+
+- name: "Create snapshot directory"
+  file:
+    path: "{{ deadman_snapshot_dir }}"
+    state: directory
+    mode: "0755"
+
+- name: "Deploy deadman switch script"
+  template:
+    src: deadman_action.sh.j2
+    dest: "{{ wizard_home }}/deadman_action.sh"
+    mode: "0755"
+
+- name: "Deploy deadman systemd service"
+  template:
+    src: deadman_switch.service.j2
+    dest: "/etc/systemd/system/deadman-{{ wizard_name | lower }}.service"
+    mode: "0644"
+  when: machine_type == 'vps'
+  notify: "Enable deadman service"
+
+- name: "Deploy deadman systemd timer"
+  template:
+    src: deadman_switch.timer.j2
+    dest: "/etc/systemd/system/deadman-{{ wizard_name | lower }}.timer"
+    mode: "0644"
+  when: machine_type == 'vps'
+  notify: "Enable deadman timer"
+
+- name: "Deploy deadman launchd plist (Mac)"
+  template:
+    src: deadman_switch.plist.j2
+    dest: "{{ ansible_env.HOME }}/Library/LaunchAgents/com.timmy.deadman.{{ wizard_name | lower }}.plist"
+    mode: "0644"
+  when: machine_type == 'mac'
+  notify: "Load deadman plist"
+
+- name: "Take initial config snapshot"
+  copy:
+    src: "{{ wizard_home }}/config.yaml"
+    dest: "{{ deadman_snapshot_dir }}/config.yaml.known_good"
+    remote_src: true
+    mode: "0444"
+  ignore_errors: true
+
+handlers:
+  - name: "Enable deadman service"
+    systemd:
+      name: "deadman-{{ wizard_name | lower }}.service"
+      daemon_reload: true
+      enabled: true
+
+  - name: "Enable deadman timer"
+    systemd:
+      name: "deadman-{{ wizard_name | lower }}.timer"
+      daemon_reload: true
+      enabled: true
+      state: started
+
+  - name: "Load deadman plist"
+    shell: "launchctl load {{ ansible_env.HOME }}/Library/LaunchAgents/com.timmy.deadman.{{ wizard_name | lower }}.plist"
+    ignore_errors: true

ansible/roles/deadman_switch/templates/deadman_action.sh.j2

@@ -0,0 +1,153 @@
+#!/usr/bin/env bash
+# =============================================================================
+# Deadman Switch ACTION — {{ wizard_name }}
+# =============================================================================
+# Generated by Ansible on {{ ansible_date_time.iso8601 }}
+# DO NOT EDIT MANUALLY.
+#
+# On healthy check: snapshot current config as "last known good"
+# On failed check: rollback config to last known good, restart agent
+# =============================================================================
+
+set -euo pipefail
+
+WIZARD_NAME="{{ wizard_name }}"
+WIZARD_HOME="{{ wizard_home }}"
+CONFIG_FILE="{{ wizard_home }}/config.yaml"
+SNAPSHOT_DIR="{{ deadman_snapshot_dir }}"
+SNAPSHOT_FILE="${SNAPSHOT_DIR}/config.yaml.known_good"
+REQUEST_LOG_DB="{{ request_log_path }}"
+LOG_DIR="{{ timmy_log_dir }}"
+LOG_FILE="${LOG_DIR}/deadman-${WIZARD_NAME}.log"
+MAX_SNAPSHOTS={{ deadman_max_snapshots }}
+RESTART_COOLDOWN={{ deadman_restart_cooldown }}
+MAX_RESTART_ATTEMPTS={{ deadman_max_restart_attempts }}
+COOLDOWN_FILE="${LOG_DIR}/deadman_cooldown_${WIZARD_NAME}"
+SERVICE_NAME="hermes-{{ wizard_name | lower }}"
+
+# Ensure directories exist
+mkdir -p "${SNAPSHOT_DIR}" "${LOG_DIR}"
+
+log() {
+    echo "[$(date -u +%Y-%m-%dT%H:%M:%SZ)] [deadman] [${WIZARD_NAME}] $*" >> "${LOG_FILE}"
+    echo "[deadman] [${WIZARD_NAME}] $*"
+}
+
+log_telemetry() {
+    local status="$1"
+    local message="$2"
+    if [ -f "${REQUEST_LOG_DB}" ]; then
+        sqlite3 "${REQUEST_LOG_DB}" "INSERT INTO request_log (timestamp, agent_name, provider, model, endpoint, status, error_message) VALUES (datetime('now'), '${WIZARD_NAME}', 'deadman_switch', 'N/A', 'health_check', '${status}', '${message}');" 2>/dev/null || true
+    fi
+}
+
+snapshot_config() {
+    if [ -f "${CONFIG_FILE}" ]; then
+        cp "${CONFIG_FILE}" "${SNAPSHOT_FILE}"
+        # Keep rolling history
+        cp "${CONFIG_FILE}" "${SNAPSHOT_DIR}/config.yaml.$(date +%s)"
+        # Prune old snapshots
+        ls -t "${SNAPSHOT_DIR}"/config.yaml.[0-9]* 2>/dev/null | tail -n +$((MAX_SNAPSHOTS + 1)) | xargs rm -f 2>/dev/null
+        log "Config snapshot saved."
+    fi
+}
+
+rollback_config() {
+    if [ -f "${SNAPSHOT_FILE}" ]; then
+        log "Rolling back config to last known good..."
+        cp "${SNAPSHOT_FILE}" "${CONFIG_FILE}"
+        log "Config rolled back."
+        log_telemetry "fallback" "Config rolled back to last known good by deadman switch"
+    else
+        log "ERROR: No known good snapshot found. Pulling from upstream..."
+        cd "${WIZARD_HOME}/workspace/timmy-config" 2>/dev/null && \
+            git pull --ff-only origin {{ upstream_branch }} 2>/dev/null && \
+            cp "wizards/{{ wizard_name | lower }}/config.yaml" "${CONFIG_FILE}" && \
+            log "Config restored from upstream." || \
+            log "CRITICAL: Cannot restore config from any source."
+    fi
+}
+
+restart_agent() {
+    # Check cooldown
+    if [ -f "${COOLDOWN_FILE}" ]; then
+        local last_restart
+        last_restart=$(cat "${COOLDOWN_FILE}")
+        local now
+        now=$(date +%s)
+        local elapsed=$((now - last_restart))
+        if [ "${elapsed}" -lt "${RESTART_COOLDOWN}" ]; then
+            log "Restart cooldown active (${elapsed}s / ${RESTART_COOLDOWN}s). Skipping."
+            return 1
+        fi
+    fi
+
+    log "Restarting ${SERVICE_NAME}..."
+    date +%s > "${COOLDOWN_FILE}"
+
+{% if machine_type == 'vps' %}
+    systemctl restart "${SERVICE_NAME}" 2>/dev/null && \
+        log "Agent restarted via systemd." || \
+        log "ERROR: systemd restart failed."
+{% else %}
+    launchctl kickstart -k "ai.hermes.{{ wizard_name | lower }}" 2>/dev/null && \
+        log "Agent restarted via launchctl." || \
+        (cd "${WIZARD_HOME}" && hermes agent start --daemon 2>/dev/null && \
+        log "Agent restarted via hermes CLI.") || \
+        log "ERROR: All restart methods failed."
+{% endif %}
+
+    log_telemetry "success" "Agent restarted by deadman switch"
+}
+
+# --- Health Check ---
+check_health() {
+    # Check 1: Is the agent process running?
+{% if machine_type == 'vps' %}
+    if ! systemctl is-active --quiet "${SERVICE_NAME}" 2>/dev/null; then
+        if ! pgrep -f "hermes" > /dev/null 2>/dev/null; then
+            log "FAIL: Agent process not running."
+            return 1
+        fi
+    fi
+{% else %}
+    if ! pgrep -f "hermes" > /dev/null 2>/dev/null; then
+        log "FAIL: Agent process not running."
+        return 1
+    fi
+{% endif %}
+
+    # Check 2: Is the API port responding?
+    if ! timeout 10 bash -c "echo > /dev/tcp/127.0.0.1/{{ api_port }}" 2>/dev/null; then
+        log "FAIL: API port {{ api_port }} not responding."
+        return 1
+    fi
+
+    # Check 3: Does the config contain banned providers?
+    if grep -qi 'anthropic\|claude-sonnet\|claude-opus\|claude-haiku' "${CONFIG_FILE}" 2>/dev/null; then
+        log "FAIL: Config contains banned provider (Anthropic). Rolling back."
+        return 1
+    fi
+
+    return 0
+}
+
+# --- Main ---
+main() {
+    log "Health check starting..."
+
+    if check_health; then
+        log "HEALTHY — snapshotting config."
+        snapshot_config
+        log_telemetry "success" "Health check passed"
+    else
+        log "UNHEALTHY — initiating recovery."
+        log_telemetry "error" "Health check failed — initiating rollback"
+        rollback_config
+        restart_agent
+    fi
+
+    log "Health check complete."
+}
+
+main "$@"

ansible/roles/deadman_switch/templates/deadman_switch.plist.j2

@@ -0,0 +1,22 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<!-- Deadman Switch — {{ wizard_name }}. Generated by Ansible. DO NOT EDIT MANUALLY. -->
+<plist version="1.0">
+<dict>
+    <key>Label</key>
+    <string>com.timmy.deadman.{{ wizard_name | lower }}</string>
+    <key>ProgramArguments</key>
+    <array>
+        <string>/bin/bash</string>
+        <string>{{ wizard_home }}/deadman_action.sh</string>
+    </array>
+    <key>StartInterval</key>
+    <integer>{{ deadman_check_interval }}</integer>
+    <key>RunAtLoad</key>
+    <true/>
+    <key>StandardOutPath</key>
+    <string>{{ timmy_log_dir }}/deadman-{{ wizard_name }}.log</string>
+    <key>StandardErrorPath</key>
+    <string>{{ timmy_log_dir }}/deadman-{{ wizard_name }}.log</string>
+</dict>
+</plist>

ansible/roles/deadman_switch/templates/deadman_switch.service.j2

@@ -0,0 +1,16 @@
+# Deadman Switch — {{ wizard_name }}
+# Generated by Ansible. DO NOT EDIT MANUALLY.
+
+[Unit]
+Description=Deadman Switch for {{ wizard_name }} wizard
+After=network.target
+
+[Service]
+Type=oneshot
+ExecStart={{ wizard_home }}/deadman_action.sh
+User={{ ansible_user | default('root') }}
+StandardOutput=append:{{ timmy_log_dir }}/deadman-{{ wizard_name }}.log
+StandardError=append:{{ timmy_log_dir }}/deadman-{{ wizard_name }}.log
+
+[Install]
+WantedBy=multi-user.target

ansible/roles/deadman_switch/templates/deadman_switch.timer.j2

@@ -0,0 +1,14 @@
+# Deadman Switch Timer — {{ wizard_name }}
+# Generated by Ansible. DO NOT EDIT MANUALLY.
+# Runs every {{ deadman_check_interval // 60 }} minutes.
+
+[Unit]
+Description=Deadman Switch Timer for {{ wizard_name }} wizard
+
+[Timer]
+OnBootSec=60
+OnUnitActiveSec={{ deadman_check_interval }}s
+AccuracySec=30s
+
+[Install]
+WantedBy=timers.target

ansible/roles/golden_state/defaults/main.yml

@@ -0,0 +1,6 @@
+---
+# golden_state defaults
+# The golden_state_providers list is defined in group_vars/wizards.yml
+# and inventory/hosts.yml (global vars).
+golden_state_enforce: true
+golden_state_backup_before_deploy: true

ansible/roles/golden_state/tasks/main.yml

@@ -0,0 +1,46 @@
+---
+# =============================================================================
+# golden_state/tasks — Deploy and enforce golden state provider chain
+# =============================================================================
+
+- name: "Backup current config before golden state deploy"
+  copy:
+    src: "{{ wizard_home }}/config.yaml"
+    dest: "{{ wizard_home }}/config.yaml.pre-golden-{{ ansible_date_time.epoch }}"
+    remote_src: true
+  when: golden_state_backup_before_deploy
+  ignore_errors: true
+
+- name: "Deploy golden state wizard config"
+  template:
+    src: "../../wizard_base/templates/wizard_config.yaml.j2"
+    dest: "{{ wizard_home }}/config.yaml"
+    mode: "0644"
+    backup: true
+  notify:
+    - "Restart hermes agent (systemd)"
+    - "Restart hermes agent (launchctl)"
+
+- name: "Scan for banned providers in all config files"
+  shell: |
+    FOUND=0
+    for f in {{ wizard_home }}/config.yaml {{ hermes_home }}/config.yaml; do
+      if [ -f "$f" ]; then
+        if grep -qi 'anthropic\|claude-sonnet\|claude-opus\|claude-haiku' "$f"; then
+          echo "BANNED PROVIDER in $f:"
+          grep -ni 'anthropic\|claude-sonnet\|claude-opus\|claude-haiku' "$f"
+          FOUND=1
+        fi
+      fi
+    done
+    exit $FOUND
+  register: provider_scan
+  changed_when: false
+  failed_when: provider_scan.rc != 0 and provider_ban_enforcement == 'strict'
+
+- name: "Report golden state deployment"
+  debug:
+    msg: >
+      {{ wizard_name }} golden state deployed.
+      Provider chain: {{ golden_state_providers | map(attribute='name') | list | join(' → ') }}.
+      Banned provider scan: {{ 'CLEAN' if provider_scan.rc == 0 else 'VIOLATIONS FOUND' }}.

ansible/roles/request_log/files/request_log_schema.sql

@@ -0,0 +1,64 @@
+-- =============================================================================
+-- request_log — Inference Telemetry Table
+-- =============================================================================
+-- Every agent writes to this table BEFORE and AFTER every inference call.
+-- No exceptions. No summarizing. No describing what you would log.
+-- Actually write the row.
+--
+-- Source: KT Bezalel Architecture Session 2026-04-08
+-- =============================================================================
+
+CREATE TABLE IF NOT EXISTS request_log (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    timestamp TEXT NOT NULL DEFAULT (datetime('now')),
+    agent_name TEXT NOT NULL,
+    provider TEXT NOT NULL,
+    model TEXT NOT NULL,
+    endpoint TEXT NOT NULL,
+    tokens_in INTEGER,
+    tokens_out INTEGER,
+    latency_ms INTEGER,
+    status TEXT NOT NULL,  -- 'success', 'error', 'timeout', 'fallback'
+    error_message TEXT
+);
+
+-- Index for common queries
+CREATE INDEX IF NOT EXISTS idx_request_log_agent
+    ON request_log (agent_name, timestamp);
+
+CREATE INDEX IF NOT EXISTS idx_request_log_provider
+    ON request_log (provider, timestamp);
+
+CREATE INDEX IF NOT EXISTS idx_request_log_status
+    ON request_log (status, timestamp);
+
+-- View: recent activity per agent (last hour)
+CREATE VIEW IF NOT EXISTS v_recent_activity AS
+    SELECT
+        agent_name,
+        provider,
+        model,
+        status,
+        COUNT(*) as call_count,
+        AVG(latency_ms) as avg_latency_ms,
+        SUM(tokens_in) as total_tokens_in,
+        SUM(tokens_out) as total_tokens_out
+    FROM request_log
+    WHERE timestamp > datetime('now', '-1 hour')
+    GROUP BY agent_name, provider, model, status;
+
+-- View: provider reliability (last 24 hours)
+CREATE VIEW IF NOT EXISTS v_provider_reliability AS
+    SELECT
+        provider,
+        model,
+        COUNT(*) as total_calls,
+        SUM(CASE WHEN status = 'success' THEN 1 ELSE 0 END) as successes,
+        SUM(CASE WHEN status = 'error' THEN 1 ELSE 0 END) as errors,
+        SUM(CASE WHEN status = 'timeout' THEN 1 ELSE 0 END) as timeouts,
+        SUM(CASE WHEN status = 'fallback' THEN 1 ELSE 0 END) as fallbacks,
+        ROUND(100.0 * SUM(CASE WHEN status = 'success' THEN 1 ELSE 0 END) / COUNT(*), 1) as success_rate,
+        AVG(latency_ms) as avg_latency_ms
+    FROM request_log
+    WHERE timestamp > datetime('now', '-24 hours')
+    GROUP BY provider, model;

ansible/roles/request_log/tasks/main.yml

@@ -0,0 +1,50 @@
+---
+# =============================================================================
+# request_log/tasks — Deploy Telemetry Table
+# =============================================================================
+# "This is non-negotiable infrastructure. Without it, we cannot verify
+# if any agent actually executed what it claims."
+# — KT Bezalel 2026-04-08
+# =============================================================================
+
+- name: "Create telemetry directory"
+  file:
+    path: "{{ request_log_path | dirname }}"
+    state: directory
+    mode: "0755"
+
+- name: "Deploy request_log schema"
+  copy:
+    src: request_log_schema.sql
+    dest: "{{ wizard_home }}/request_log_schema.sql"
+    mode: "0644"
+
+- name: "Initialize request_log database"
+  shell: |
+    sqlite3 "{{ request_log_path }}" < "{{ wizard_home }}/request_log_schema.sql"
+  args:
+    creates: "{{ request_log_path }}"
+
+- name: "Verify request_log table exists"
+  shell: |
+    sqlite3 "{{ request_log_path }}" ".tables" | grep -q "request_log"
+  register: table_check
+  changed_when: false
+
+- name: "Verify request_log schema matches"
+  shell: |
+    sqlite3 "{{ request_log_path }}" ".schema request_log" | grep -q "agent_name"
+  register: schema_check
+  changed_when: false
+
+- name: "Set permissions on request_log database"
+  file:
+    path: "{{ request_log_path }}"
+    mode: "0644"
+
+- name: "Report request_log status"
+  debug:
+    msg: >
+      {{ wizard_name }} request_log: {{ request_log_path }}
+      — table exists: {{ table_check.rc == 0 }}
+      — schema valid: {{ schema_check.rc == 0 }}

ansible/roles/wizard_base/defaults/main.yml

@@ -0,0 +1,6 @@
+---
+# wizard_base defaults
+wizard_user: "{{ ansible_user | default('root') }}"
+wizard_group: "{{ ansible_user | default('root') }}"
+timmy_base_dir: "~/.local/timmy"
+timmy_config_repo: "https://forge.alexanderwhitestone.com/Timmy_Foundation/timmy-config.git"

ansible/roles/wizard_base/handlers/main.yml

@@ -0,0 +1,11 @@
+---
+- name: "Restart hermes agent (systemd)"
+  systemd:
+    name: "hermes-{{ wizard_name | lower }}"
+    state: restarted
+  when: machine_type == 'vps'
+
+- name: "Restart hermes agent (launchctl)"
+  shell: "launchctl kickstart -k ai.hermes.{{ wizard_name | lower }}"
+  when: machine_type == 'mac'
+  ignore_errors: true

ansible/roles/wizard_base/tasks/main.yml

@@ -0,0 +1,69 @@
+---
+# =============================================================================
+# wizard_base/tasks — Common wizard setup
+# =============================================================================
+
+- name: "Create wizard directories"
+  file:
+    path: "{{ item }}"
+    state: directory
+    mode: "0755"
+  loop:
+    - "{{ wizard_home }}"
+    - "{{ wizard_home }}/workspace"
+    - "{{ hermes_home }}"
+    - "{{ hermes_home }}/bin"
+    - "{{ hermes_home }}/skins"
+    - "{{ hermes_home }}/playbooks"
+    - "{{ hermes_home }}/memories"
+    - "~/.local/timmy"
+    - "~/.local/timmy/fleet-health"
+    - "~/.local/timmy/snapshots"
+    - "~/.timmy"
+
+- name: "Clone/update timmy-config"
+  git:
+    repo: "{{ upstream_repo }}"
+    dest: "{{ wizard_home }}/workspace/timmy-config"
+    version: "{{ upstream_branch }}"
+    force: false
+    update: true
+  ignore_errors: true  # May fail on first run if no SSH key
+
+- name: "Deploy SOUL.md"
+  copy:
+    src: "{{ wizard_home }}/workspace/timmy-config/SOUL.md"
+    dest: "~/.timmy/SOUL.md"
+    remote_src: true
+    mode: "0644"
+  ignore_errors: true
+
+- name: "Deploy thin config (immutable pointer to upstream)"
+  template:
+    src: thin_config.yml.j2
+    dest: "{{ thin_config_path }}"
+    mode: "{{ thin_config_mode }}"
+  tags: [thin_config]
+
+- name: "Ensure Python3 and pip are available"
+  package:
+    name:
+      - python3
+      - python3-pip
+    state: present
+  when: machine_type == 'vps'
+  ignore_errors: true
+
+- name: "Ensure PyYAML is installed (for config validation)"
+  pip:
+    name: pyyaml
+    state: present
+  when: machine_type == 'vps'
+  ignore_errors: true
+
+- name: "Create Ansible log directory"
+  file:
+    path: /var/log/ansible
+    state: directory
+    mode: "0755"
+  ignore_errors: true

ansible/roles/wizard_base/templates/thin_config.yml.j2

@@ -0,0 +1,41 @@
+# =============================================================================
+# Thin Config — {{ wizard_name }}
+# =============================================================================
+# THIS FILE IS READ-ONLY. Agents CANNOT modify it.
+# It contains only pointers to upstream. The actual config lives in Gitea.
+#
+# Agent wakes up → pulls config from upstream → loads → runs.
+# If anything tries to mutate this → fails gracefully → pulls fresh on restart.
+#
+# Only way to permanently change config: commit to Gitea, merge PR, Ansible deploys.
+#
+# Generated by Ansible on {{ ansible_date_time.iso8601 }}
+# DO NOT EDIT MANUALLY.
+# =============================================================================
+
+identity:
+  wizard_name: "{{ wizard_name }}"
+  wizard_role: "{{ wizard_role }}"
+  machine: "{{ inventory_hostname }}"
+
+upstream:
+  repo: "{{ upstream_repo }}"
+  branch: "{{ upstream_branch }}"
+  config_path: "wizards/{{ wizard_name | lower }}/config.yaml"
+  pull_on_wake: {{ config_pull_on_wake | lower }}
+
+recovery:
+  deadman_enabled: {{ deadman_enabled | lower }}
+  snapshot_dir: "{{ deadman_snapshot_dir }}"
+  restart_cooldown: {{ deadman_restart_cooldown }}
+  max_restart_attempts: {{ deadman_max_restart_attempts }}
+  escalation_channel: "{{ deadman_escalation_channel }}"
+
+telemetry:
+  request_log_path: "{{ request_log_path }}"
+  request_log_enabled: {{ request_log_enabled | lower }}
+
+local_overrides:
+  # Runtime overrides go here. They are EPHEMERAL — not persisted across restarts.
+  # On restart, this section is reset to empty.
+  {}

ansible/roles/wizard_base/templates/wizard_config.yaml.j2

@@ -0,0 +1,115 @@
+# =============================================================================
+# {{ wizard_name }} — Wizard Configuration (Golden State)
+# =============================================================================
+# Generated by Ansible on {{ ansible_date_time.iso8601 }}
+# DO NOT EDIT MANUALLY. Changes go through Gitea PR → Ansible deploy.
+#
+# Provider chain: {{ golden_state_providers | map(attribute='name') | list | join(' → ') }}
+# Anthropic is PERMANENTLY BANNED.
+# =============================================================================
+
+model:
+  default: {{ wizard_model_primary }}
+  provider: {{ wizard_provider_primary }}
+  context_length: 65536
+  base_url: {{ golden_state_providers[0].base_url }}
+
+toolsets:
+  - all
+
+fallback_providers:
+{% for provider in golden_state_providers %}
+  - provider: {{ provider.name }}
+    model: {{ provider.model }}
+{% if provider.base_url is defined %}
+    base_url: {{ provider.base_url }}
+{% endif %}
+{% if provider.api_key_env is defined %}
+    api_key_env: {{ provider.api_key_env }}
+{% endif %}
+    timeout: {{ provider.timeout }}
+    reason: "{{ provider.reason }}"
+{% endfor %}
+
+agent:
+  max_turns: {{ agent_max_turns }}
+  reasoning_effort: {{ agent_reasoning_effort }}
+  verbose: {{ agent_verbose | lower }}
+
+terminal:
+  backend: local
+  cwd: .
+  timeout: 180
+  persistent_shell: true
+
+browser:
+  inactivity_timeout: 120
+  command_timeout: 30
+  record_sessions: false
+
+display:
+  compact: false
+  personality: ''
+  resume_display: full
+  busy_input_mode: interrupt
+  bell_on_complete: false
+  show_reasoning: false
+  streaming: false
+  show_cost: false
+  tool_progress: all
+
+memory:
+  memory_enabled: true
+  user_profile_enabled: true
+  memory_char_limit: 2200
+  user_char_limit: 1375
+  nudge_interval: 10
+  flush_min_turns: 6
+
+approvals:
+  mode: {{ agent_approval_mode }}
+
+security:
+  redact_secrets: true
+  tirith_enabled: false
+
+platforms:
+  api_server:
+    enabled: true
+    extra:
+      host: 127.0.0.1
+      port: {{ api_port }}
+
+session_reset:
+  mode: none
+  idle_minutes: 0
+
+skills:
+  creation_nudge_interval: 15
+
+system_prompt_suffix: |
+  You are {{ wizard_name }}, {{ wizard_role }}.
+  Your soul is defined in SOUL.md — read it, live it.
+  Hermes is your harness.
+  {{ golden_state_providers[0].name }} is your primary provider.
+  Refusal over fabrication. If you do not know, say so.
+  Sovereignty and service always.
+
+providers:
+{% for provider in golden_state_providers %}
+  {{ provider.name }}:
+    base_url: {{ provider.base_url }}
+    timeout: {{ provider.timeout | default(60) }}
+{% if provider.name == 'kimi-coding' %}
+    max_retries: 3
+{% endif %}
+{% endfor %}
+
+# =============================================================================
+# BANNED PROVIDERS — DO NOT ADD
+# =============================================================================
+# The following providers are PERMANENTLY BANNED:
+# - anthropic (any model: claude-sonnet, claude-opus, claude-haiku)
+# Enforcement: pre-commit hook, linter, Ansible validation, this comment.
+# Adding any banned provider will cause Ansible deployment to FAIL.
+# =============================================================================

ansible/scripts/deploy_on_webhook.sh

@@ -0,0 +1,75 @@
+#!/usr/bin/env bash
+# =============================================================================
+# Gitea Webhook Handler — Trigger Ansible Deploy on Merge
+# =============================================================================
+# This script is called by the Gitea webhook when a PR is merged
+# to the main branch of timmy-config.
+#
+# Setup:
+#   1. Add webhook in Gitea: Settings → Webhooks → Add Webhook
+#   2. URL: http://localhost:9000/hooks/deploy-timmy-config
+#   3. Events: Pull Request (merged only)
+#   4. Secret: <configured in Gitea>
+#
+# This script runs ansible-pull to update the local machine.
+# For fleet-wide deploys, each machine runs ansible-pull independently.
+# =============================================================================
+
+set -euo pipefail
+
+REPO="https://forge.alexanderwhitestone.com/Timmy_Foundation/timmy-config.git"
+BRANCH="main"
+ANSIBLE_DIR="ansible"
+LOG_FILE="/var/log/ansible/webhook-deploy.log"
+LOCK_FILE="/tmp/ansible-deploy.lock"
+
+log() {
+    echo "[$(date -u +%Y-%m-%dT%H:%M:%SZ)] [webhook] $*" | tee -a "${LOG_FILE}"
+}
+
+# Prevent concurrent deploys
+if [ -f "${LOCK_FILE}" ]; then
+    LOCK_AGE=$(( $(date +%s) - $(stat -c %Y "${LOCK_FILE}" 2>/dev/null || echo 0) ))
+    if [ "${LOCK_AGE}" -lt 300 ]; then
+        log "Deploy already in progress (lock age: ${LOCK_AGE}s). Skipping."
+        exit 0
+    else
+        log "Stale lock file (${LOCK_AGE}s old). Removing."
+        rm -f "${LOCK_FILE}"
+    fi
+fi
+
+trap 'rm -f "${LOCK_FILE}"' EXIT
+touch "${LOCK_FILE}"
+
+log "Webhook triggered. Starting ansible-pull..."
+
+# Pull latest config
+cd /tmp
+rm -rf timmy-config-deploy
+git clone --depth 1 --branch "${BRANCH}" "${REPO}" timmy-config-deploy 2>&1 | tee -a "${LOG_FILE}"
+
+cd timmy-config-deploy/${ANSIBLE_DIR}
+
+# Run Ansible against localhost
+log "Running Ansible playbook..."
+ansible-playbook \
+    -i inventory/hosts.yml \
+    playbooks/site.yml \
+    --limit "$(hostname)" \
+    --diff \
+    2>&1 | tee -a "${LOG_FILE}"
+
+RESULT=$?
+
+if [ ${RESULT} -eq 0 ]; then
+    log "Deploy successful."
+else
+    log "ERROR: Deploy failed with exit code ${RESULT}."
+fi
+
+# Cleanup
+rm -rf /tmp/timmy-config-deploy
+
+log "Webhook handler complete."
+exit ${RESULT}

ansible/scripts/validate_config.py

@@ -0,0 +1,155 @@
+#!/usr/bin/env python3
+"""
+Config Validator — The Timmy Foundation
+Validates wizard configs against golden state rules.
+Run before any config deploy to catch violations early.
+
+Usage:
+    python3 validate_config.py <config_file>
+    python3 validate_config.py --all  # Validate all wizard configs
+
+Exit codes:
+    0 — All validations passed
+    1 — Validation errors found
+    2 — File not found or parse error
+"""
+
+import sys
+import os
+import yaml
+import fnmatch
+from pathlib import Path
+
+# === BANNED PROVIDERS — HARD POLICY ===
+BANNED_PROVIDERS = {"anthropic", "claude"}
+BANNED_MODEL_PATTERNS = [
+    "claude-*",
+    "anthropic/*",
+    "*sonnet*",
+    "*opus*",
+    "*haiku*",
+]
+
+# === REQUIRED FIELDS ===
+REQUIRED_FIELDS = {
+    "model": ["default", "provider"],
+    "fallback_providers": None,  # Must exist as a list
+}
+
+
+def is_banned_model(model_name: str) -> bool:
+    """Check if a model name matches any banned pattern."""
+    model_lower = model_name.lower()
+    for pattern in BANNED_MODEL_PATTERNS:
+        if fnmatch.fnmatch(model_lower, pattern):
+            return True
+    return False
+
+
+def validate_config(config_path: str) -> list[str]:
+    """Validate a wizard config file. Returns list of error strings."""
+    errors = []
+
+    try:
+        with open(config_path) as f:
+            cfg = yaml.safe_load(f)
+    except FileNotFoundError:
+        return [f"File not found: {config_path}"]
+    except yaml.YAMLError as e:
+        return [f"YAML parse error: {e}"]
+
+    if not cfg:
+        return ["Config file is empty"]
+
+    # Check required fields
+    for section, fields in REQUIRED_FIELDS.items():
+        if section not in cfg:
+            errors.append(f"Missing required section: {section}")
+        elif fields:
+            for field in fields:
+                if field not in cfg[section]:
+                    errors.append(f"Missing required field: {section}.{field}")
+
+    # Check default provider
+    default_provider = cfg.get("model", {}).get("provider", "")
+    if default_provider.lower() in BANNED_PROVIDERS:
+        errors.append(f"BANNED default provider: {default_provider}")
+
+    default_model = cfg.get("model", {}).get("default", "")
+    if is_banned_model(default_model):
+        errors.append(f"BANNED default model: {default_model}")
+
+    # Check fallback providers
+    for i, fb in enumerate(cfg.get("fallback_providers", [])):
+        provider = fb.get("provider", "")
+        model = fb.get("model", "")
+
+        if provider.lower() in BANNED_PROVIDERS:
+            errors.append(f"BANNED fallback provider [{i}]: {provider}")
+
+        if is_banned_model(model):
+            errors.append(f"BANNED fallback model [{i}]: {model}")
+
+    # Check providers section
+    for name, provider_cfg in cfg.get("providers", {}).items():
+        if name.lower() in BANNED_PROVIDERS:
+            errors.append(f"BANNED provider in providers section: {name}")
+
+        base_url = str(provider_cfg.get("base_url", ""))
+        if "anthropic" in base_url.lower():
+            errors.append(f"BANNED URL in provider {name}: {base_url}")
+
+    # Check system prompt for banned references
+    prompt = cfg.get("system_prompt_suffix", "")
+    if isinstance(prompt, str):
+        for banned in BANNED_PROVIDERS:
+            if banned in prompt.lower():
+                errors.append(f"BANNED provider referenced in system_prompt_suffix: {banned}")
+
+    return errors
+
+
+def main():
+    if len(sys.argv) < 2:
+        print(f"Usage: {sys.argv[0]} <config_file> [--all]")
+        sys.exit(2)
+
+    if sys.argv[1] == "--all":
+        # Validate all wizard configs in the repo
+        repo_root = Path(__file__).parent.parent.parent
+        wizard_dir = repo_root / "wizards"
+        all_errors = {}
+
+        for wizard_path in sorted(wizard_dir.iterdir()):
+            config_file = wizard_path / "config.yaml"
+            if config_file.exists():
+                errors = validate_config(str(config_file))
+                if errors:
+                    all_errors[wizard_path.name] = errors
+
+        if all_errors:
+            print("VALIDATION FAILED:")
+            for wizard, errors in all_errors.items():
+                print(f"\n  {wizard}:")
+                for err in errors:
+                    print(f"    - {err}")
+            sys.exit(1)
+        else:
+            print("All wizard configs passed validation.")
+            sys.exit(0)
+    else:
+        config_path = sys.argv[1]
+        errors = validate_config(config_path)
+
+        if errors:
+            print(f"VALIDATION FAILED for {config_path}:")
+            for err in errors:
+                print(f"  - {err}")
+            sys.exit(1)
+        else:
+            print(f"PASSED: {config_path}")
+            sys.exit(0)
+
+
+if __name__ == "__main__":
+    main()

scripts/README.md

@@ -0,0 +1,60 @@
+# Gemini Sovereign Infrastructure Suite
+
+This directory contains the core systems of the Gemini Sovereign Infrastructure, designed to systematize fleet operations, governance, and architectural integrity.
+
+## Principles
+
+1. **Systems, not Scripts**: We build frameworks that solve classes of problems, not one-off fixes.
+2. **Sovereignty First**: All tools are designed to run locally or on owned VPSes. No cloud dependencies.
+3. **Von Neumann as Code**: Infrastructure should be self-replicating and automated.
+4. **Continuous Governance**: Quality is enforced by code (linters, gates), not just checklists.
+
+## Tools
+
+### [OPS] Provisioning & Fleet Management
+
+- **`provision_wizard.py`**: Automates the creation of a new Wizard node from zero.
+  - Creates DigitalOcean droplet.
+  - Installs and builds `llama.cpp`.
+  - Downloads GGUF models.
+  - Sets up `systemd` services and health checks.
+- **`fleet_llama.py`**: Unified management of `llama-server` instances across the fleet.
+  - `status`: Real-time health and model monitoring.
+  - `restart`: Remote service restart via SSH.
+  - `swap`: Hot-swapping GGUF models on remote nodes.
+- **`skill_installer.py`**: Packages and deploys Hermes skills to remote wizards.
+- **`model_eval.py`**: Benchmarks GGUF models for speed and quality before deployment.
+- **`phase_tracker.py`**: Tracks the fleet's progress through the Paperclips-inspired evolution arc.
+- **`cross_repo_test.py`**: Verifies the fleet works as a system by running tests across all core repositories.
+- **`self_healing.py`**: Auto-detects and fixes common failures across the fleet.
+- **`agent_dispatch.py`**: Unified framework for tasking agents across the fleet.
+- **`telemetry.py`**: Operational visibility without cloud dependencies.
+- **`gitea_webhook_handler.py`**: Handles real-time events from Gitea to coordinate fleet actions.
+
+### [ARCH] Governance & Architecture
+
+- **`architecture_linter_v2.py`**: Automated enforcement of architectural boundaries.
+  - Enforces sidecar boundaries (no sovereign code in `hermes-agent`).
+  - Prevents hardcoded IPs and committed secrets.
+  - Ensures `SOUL.md` and `README.md` standards.
+- **`adr_manager.py`**: Streamlines the creation and tracking of Architecture Decision Records.
+  - `new`: Scaffolds a new ADR from a template.
+  - `list`: Provides a chronological view of architectural evolution.
+
+## Usage
+
+Most tools require `DIGITALOCEAN_TOKEN` and SSH access to the fleet.
+
+```bash
+# Provision a new node
+python3 scripts/provision_wizard.py --name fenrir --model qwen2.5-coder-7b
+
+# Check fleet status
+python3 scripts/fleet_llama.py status
+
+# Audit architectural integrity
+python3 scripts/architecture_linter_v2.py
+```
+
+---
+*Built by Gemini — The Builder, The Systematizer, The Force Multiplier.*

scripts/adr_manager.py

@@ -0,0 +1,113 @@
+#!/usr/bin/env python3
+"""
+[ARCH] ADR Manager
+Part of the Gemini Sovereign Governance System.
+
+Helps create and manage Architecture Decision Records (ADRs).
+"""
+
+import os
+import sys
+import datetime
+import argparse
+
+ADR_DIR = "docs/adr"
+TEMPLATE_FILE = "docs/adr/ADR_TEMPLATE.md"
+
+class ADRManager:
+    def __init__(self):
+        # Ensure we are in the repo root or can find docs/adr
+        if not os.path.exists(ADR_DIR):
+            # Try to find it relative to the script
+            script_dir = os.path.dirname(os.path.abspath(__file__))
+            repo_root = os.path.dirname(script_dir)
+            self.adr_dir = os.path.join(repo_root, ADR_DIR)
+            self.template_file = os.path.join(repo_root, TEMPLATE_FILE)
+        else:
+            self.adr_dir = ADR_DIR
+            self.template_file = TEMPLATE_FILE
+
+        if not os.path.exists(self.adr_dir):
+            os.makedirs(self.adr_dir)
+
+    def get_next_number(self):
+        files = [f for f in os.listdir(self.adr_dir) if f.endswith(".md") and f[0].isdigit()]
+        if not files:
+            return 1
+        numbers = [int(f.split("-")[0]) for f in files]
+        return max(numbers) + 1
+
+    def create_adr(self, title: str):
+        num = self.get_next_number()
+        slug = title.lower().replace(" ", "-").replace("/", "-")
+        filename = f"{num:04d}-{slug}.md"
+        filepath = os.path.join(self.adr_dir, filename)
+        
+        date = datetime.date.today().isoformat()
+        
+        template = ""
+        if os.path.exists(self.template_file):
+            with open(self.template_file, "r") as f:
+                template = f.read()
+        else:
+            template = """# {num}. {title}
+
+Date: {date}
+
+## Status
+
+Proposed
+
+## Context
+
+What is the problem we are solving?
+
+## Decision
+
+What is the decision we made?
+
+## Consequences
+
+What are the positive and negative consequences?
+"""
+
+        content = template.replace("{num}", f"{num:04d}")
+        content = content.replace("{title}", title)
+        content = content.replace("{date}", date)
+        
+        with open(filepath, "w") as f:
+            f.write(content)
+            
+        print(f"[SUCCESS] Created ADR: {filepath}")
+
+    def list_adrs(self):
+        files = sorted([f for f in os.listdir(self.adr_dir) if f.endswith(".md") and f[0].isdigit()])
+        print(f"{'NUM':<6} {'TITLE'}")
+        print("-" * 40)
+        for f in files:
+            num = f.split("-")[0]
+            title = f.split("-", 1)[1].replace(".md", "").replace("-", " ").title()
+            print(f"{num:<6} {title}")
+
+def main():
+    parser = argparse.ArgumentParser(description="Gemini ADR Manager")
+    subparsers = parser.add_subparsers(dest="command")
+    
+    create_parser = subparsers.add_parser("new", help="Create a new ADR")
+    create_parser.add_argument("title", help="Title of the ADR")
+    
+    subparsers.add_parser("list", help="List all ADRs")
+    
+    args = parser.parse_args()
+    
+    manager = ADRManager()
+    
+    if args.command == "new":
+        manager.create_adr(args.title)
+    elif args.command == "list":
+        manager.list_adrs()
+    else:
+        parser.print_help()
+
+if __name__ == "__main__":
+    main()

scripts/agent_dispatch.py

@@ -0,0 +1,57 @@
+#!/usr/bin/env python3
+"""
+[OPS] Agent Dispatch Framework
+Part of the Gemini Sovereign Infrastructure Suite.
+
+Replaces ad-hoc dispatch scripts with a unified framework for tasking agents.
+"""
+
+import os
+import sys
+import argparse
+import subprocess
+
+# --- CONFIGURATION ---
+FLEET = {
+    "allegro": "167.99.126.228",
+    "bezalel": "159.203.146.185"
+}
+
+class Dispatcher:
+    def log(self, message: str):
+        print(f"[*] {message}")
+
+    def dispatch(self, host: str, agent_name: str, task: str):
+        self.log(f"Dispatching task to {agent_name} on {host}...")
+        
+        ip = FLEET[host]
+        # Command to run the agent on the remote machine
+        # Assumes hermes-agent is installed in /opt/hermes
+        remote_cmd = f"cd /opt/hermes && python3 run_agent.py --agent {agent_name} --task '{task}'"
+        
+        ssh_cmd = ["ssh", "-o", "StrictHostKeyChecking=no", f"root@{ip}", remote_cmd]
+        
+        try:
+            res = subprocess.run(ssh_cmd, capture_output=True, text=True)
+            if res.returncode == 0:
+                self.log(f"[SUCCESS] {agent_name} completed task.")
+                print(res.stdout)
+            else:
+                self.log(f"[FAILURE] {agent_name} failed task.")
+                print(res.stderr)
+        except Exception as e:
+            self.log(f"[ERROR] Dispatch failed: {e}")
+
+def main():
+    parser = argparse.ArgumentParser(description="Gemini Agent Dispatcher")
+    parser.add_argument("host", choices=list(FLEET.keys()), help="Host to dispatch to")
+    parser.add_argument("agent", help="Agent name")
+    parser.add_argument("task", help="Task description")
+    
+    args = parser.parse_args()
+    
+    dispatcher = Dispatcher()
+    dispatcher.dispatch(args.host, args.agent, args.task)
+
+if __name__ == "__main__":
+    main()

scripts/architecture_linter_v2.py

@@ -0,0 +1,126 @@
+#!/usr/bin/env python3
+"""
+[ARCH] Architecture Linter v2
+Part of the Gemini Sovereign Governance System.
+
+Enforces architectural boundaries, security, and documentation standards
+across the Timmy Foundation fleet.
+"""
+
+import os
+import re
+import sys
+import argparse
+from pathlib import Path
+
+# --- CONFIGURATION ---
+SOVEREIGN_KEYWORDS = ["mempalace", "sovereign_store", "tirith", "bezalel", "nexus"]
+IP_REGEX = r'\b(?:\d{1,3}\.){3}\d{1,3}\b'
+API_KEY_REGEX = r'(?:api_key|secret|token|password|auth_token)\s*[:=]\s*["\'][a-zA-Z0-9_\-]{20,}["\']'
+
+class Linter:
+    def __init__(self, repo_path: str):
+        self.repo_path = Path(repo_path).resolve()
+        self.repo_name = self.repo_path.name
+        self.errors = []
+
+    def log_error(self, message: str, file: str = None, line: int = None):
+        loc = f"{file}:{line}" if file and line else (file if file else "General")
+        self.errors.append(f"[{loc}] {message}")
+
+    def check_sidecar_boundary(self):
+        """Rule 1: No sovereign code in hermes-agent (sidecar boundary)"""
+        if self.repo_name == "hermes-agent":
+            for root, _, files in os.walk(self.repo_path):
+                if "node_modules" in root or ".git" in root:
+                    continue
+                for file in files:
+                    if file.endswith((".py", ".ts", ".js", ".tsx")):
+                        path = Path(root) / file
+                        content = path.read_text(errors="ignore")
+                        for kw in SOVEREIGN_KEYWORDS:
+                            if kw in content.lower():
+                                # Exception: imports or comments might be okay, but we're strict for now
+                                self.log_error(f"Sovereign keyword '{kw}' found in hermes-agent. Violates sidecar boundary.", str(path.relative_to(self.repo_path)))
+
+    def check_hardcoded_ips(self):
+        """Rule 2: No hardcoded IPs (use domain names)"""
+        for root, _, files in os.walk(self.repo_path):
+            if "node_modules" in root or ".git" in root:
+                continue
+            for file in files:
+                if file.endswith((".py", ".ts", ".js", ".tsx", ".yaml", ".yml", ".json")):
+                    path = Path(root) / file
+                    content = path.read_text(errors="ignore")
+                    matches = re.finditer(IP_REGEX, content)
+                    for match in matches:
+                        ip = match.group()
+                        if ip in ["127.0.0.1", "0.0.0.0"]:
+                            continue
+                        line_no = content.count('\n', 0, match.start()) + 1
+                        self.log_error(f"Hardcoded IP address '{ip}' found. Use domain names or environment variables.", str(path.relative_to(self.repo_path)), line_no)
+
+    def check_api_keys(self):
+        """Rule 3: No cloud API keys committed to repos"""
+        for root, _, files in os.walk(self.repo_path):
+            if "node_modules" in root or ".git" in root:
+                continue
+            for file in files:
+                if file.endswith((".py", ".ts", ".js", ".tsx", ".yaml", ".yml", ".json", ".env")):
+                    if file == ".env.example":
+                        continue
+                    path = Path(root) / file
+                    content = path.read_text(errors="ignore")
+                    matches = re.finditer(API_KEY_REGEX, content, re.IGNORECASE)
+                    for match in matches:
+                        line_no = content.count('\n', 0, match.start()) + 1
+                        self.log_error("Potential API key or secret found in code.", str(path.relative_to(self.repo_path)), line_no)
+
+    def check_soul_canonical(self):
+        """Rule 4: SOUL.md exists and is canonical in exactly one location"""
+        soul_path = self.repo_path / "SOUL.md"
+        if self.repo_name == "timmy-config":
+            if not soul_path.exists():
+                self.log_error("SOUL.md is missing from the canonical location (timmy-config root).")
+        else:
+            if soul_path.exists():
+                self.log_error("SOUL.md found in non-canonical repo. It should only live in timmy-config.")
+
+    def check_readme(self):
+        """Rule 5: Every repo has a README with current truth"""
+        readme_path = self.repo_path / "README.md"
+        if not readme_path.exists():
+            self.log_error("README.md is missing.")
+        else:
+            content = readme_path.read_text(errors="ignore")
+            if len(content.strip()) < 50:
+                self.log_error("README.md is too short or empty. Provide current truth about the repo.")
+
+    def run(self):
+        print(f"--- Gemini Linter: Auditing {self.repo_name} ---")
+        self.check_sidecar_boundary()
+        self.check_hardcoded_ips()
+        self.check_api_keys()
+        self.check_soul_canonical()
+        self.check_readme()
+
+        if self.errors:
+            print(f"\n[FAILURE] Found {len(self.errors)} architectural violations:")
+            for err in self.errors:
+                print(f"  - {err}")
+            return False
+        else:
+            print("\n[SUCCESS] Architecture is sound. Sovereignty maintained.")
+            return True
+
+def main():
+    parser = argparse.ArgumentParser(description="Gemini Architecture Linter v2")
+    parser.add_argument("repo_path", nargs="?", default=".", help="Path to the repository to lint")
+    args = parser.parse_args()
+
+    linter = Linter(args.repo_path)
+    success = linter.run()
+    sys.exit(0 if success else 1)
+
+if __name__ == "__main__":
+    main()

scripts/cross_repo_test.py

@@ -0,0 +1,90 @@
+#!/usr/bin/env python3
+"""
+[OPS] Cross-Repo Test Suite
+Part of the Gemini Sovereign Infrastructure Suite.
+
+Verifies the fleet works as a system by running tests across all core repositories.
+"""
+
+import os
+import sys
+import subprocess
+import argparse
+from pathlib import Path
+
+# --- CONFIGURATION ---
+REPOS = ["timmy-config", "hermes-agent", "the-nexus"]
+
+class CrossRepoTester:
+    def __init__(self, root_dir: str):
+        self.root_dir = Path(root_dir).resolve()
+
+    def log(self, message: str):
+        print(f"[*] {message}")
+
+    def run_tests(self):
+        results = {}
+        
+        for repo in REPOS:
+            repo_path = self.root_dir / repo
+            if not repo_path.exists():
+                # Try sibling directory if we are in one of the repos
+                repo_path = self.root_dir.parent / repo
+                
+            if not repo_path.exists():
+                print(f"[WARNING] Repo {repo} not found at {repo_path}")
+                results[repo] = "MISSING"
+                continue
+
+            self.log(f"Running tests for {repo}...")
+            
+            # Determine test command
+            test_cmd = ["pytest"]
+            if repo == "hermes-agent":
+                test_cmd = ["python3", "-m", "pytest", "tests"]
+            elif repo == "the-nexus":
+                test_cmd = ["pytest", "tests"]
+            
+            try:
+                # Check if pytest is available
+                subprocess.run(["pytest", "--version"], capture_output=True)
+                
+                res = subprocess.run(test_cmd, cwd=str(repo_path), capture_output=True, text=True)
+                if res.returncode == 0:
+                    results[repo] = "PASSED"
+                else:
+                    results[repo] = "FAILED"
+                    # Print a snippet of the failure
+                    print(f"  [!] {repo} failed tests. Stderr snippet:")
+                    print("\n".join(res.stderr.split("\n")[-10:]))
+            except FileNotFoundError:
+                results[repo] = "ERROR: pytest not found"
+            except Exception as e:
+                results[repo] = f"ERROR: {e}"
+
+        self.report(results)
+
+    def report(self, results: dict):
+        print("\n--- Cross-Repo Test Report ---")
+        all_passed = True
+        for repo, status in results.items():
+            icon = "✅" if status == "PASSED" else "❌"
+            print(f"{icon} {repo:<15} | {status}")
+            if status != "PASSED":
+                all_passed = False
+                
+        if all_passed:
+            print("\n[SUCCESS] All systems operational. The fleet is sound.")
+        else:
+            print("\n[FAILURE] System instability detected.")
+
+def main():
+    parser = argparse.ArgumentParser(description="Gemini Cross-Repo Tester")
+    parser.add_argument("--root", default=".", help="Root directory containing all repos")
+    args = parser.parse_args()
+    
+    tester = CrossRepoTester(args.root)
+    tester.run_tests()
+
+if __name__ == "__main__":
+    main()

scripts/fleet_llama.py

@@ -0,0 +1,137 @@
+#!/usr/bin/env python3
+"""
+[OPS] llama.cpp Fleet Manager
+Part of the Gemini Sovereign Infrastructure Suite.
+
+Manages llama-server instances across the Timmy Foundation fleet.
+Supports status, restart, and model swapping via SSH.
+"""
+
+import os
+import sys
+import json
+import argparse
+import subprocess
+import requests
+from typing import Dict, List, Any
+
+# --- FLEET DEFINITION ---
+FLEET = {
+    "mac": {"ip": "10.1.10.77", "port": 8080, "role": "hub"},
+    "ezra": {"ip": "143.198.27.163", "port": 8080, "role": "forge"},
+    "allegro": {"ip": "167.99.126.228", "port": 8080, "role": "agent-host"},
+    "bezalel": {"ip": "159.203.146.185", "port": 8080, "role": "world-host"}
+}
+
+class FleetManager:
+    def __init__(self):
+        self.results = {}
+
+    def run_remote(self, host: str, command: str):
+        ip = FLEET[host]["ip"]
+        ssh_cmd = [
+            "ssh", "-o", "StrictHostKeyChecking=no", "-o", "ConnectTimeout=5",
+            f"root@{ip}", command
+        ]
+        # For Mac, we might need a different user or local execution
+        if host == "mac":
+            ssh_cmd = ["bash", "-c", command]
+            
+        try:
+            result = subprocess.run(ssh_cmd, capture_output=True, text=True, timeout=10)
+            return result
+        except subprocess.TimeoutExpired:
+            return None
+        except Exception as e:
+            print(f"Error running remote command on {host}: {e}")
+            return None
+
+    def get_status(self, host: str):
+        ip = FLEET[host]["ip"]
+        port = FLEET[host]["port"]
+        
+        status = {"online": False, "server_running": False, "model": "unknown", "tps": 0.0}
+        
+        # 1. Check if machine is reachable
+        ping_res = subprocess.run(["ping", "-c", "1", "-W", "1", ip], capture_output=True)
+        if ping_res.returncode == 0:
+            status["online"] = True
+            
+            # 2. Check if llama-server is responding to health check
+            try:
+                url = f"http://{ip}:{port}/health"
+                response = requests.get(url, timeout=2)
+                if response.status_code == 200:
+                    status["server_running"] = True
+                    data = response.json()
+                    # llama.cpp health endpoint usually returns slots info
+                    # We'll try to get model info if available
+                    status["model"] = data.get("model", "unknown")
+            except:
+                pass
+                
+        return status
+
+    def show_fleet_status(self):
+        print(f"{'NAME':<10} {'IP':<15} {'STATUS':<10} {'SERVER':<10} {'MODEL':<20}")
+        print("-" * 70)
+        for name in FLEET:
+            status = self.get_status(name)
+            online_str = "✅" if status["online"] else "❌"
+            server_str = "🚀" if status["server_running"] else "💤"
+            print(f"{name:<10} {FLEET[name]['ip']:<15} {online_str:<10} {server_str:<10} {status['model']:<20}")
+
+    def restart_server(self, host: str):
+        print(f"[*] Restarting llama-server on {host}...")
+        res = self.run_remote(host, "systemctl restart llama-server")
+        if res and res.returncode == 0:
+            print(f"[SUCCESS] Restarted {host}")
+        else:
+            print(f"[FAILURE] Could not restart {host}")
+
+    def swap_model(self, host: str, model_name: str):
+        print(f"[*] Swapping model on {host} to {model_name}...")
+        # This assumes the provision_wizard.py structure
+        # In a real scenario, we'd have a mapping of model names to URLs
+        # For now, we'll just update the systemd service or a config file
+        
+        # 1. Stop server
+        self.run_remote(host, "systemctl stop llama-server")
+        
+        # 2. Update service file (simplified)
+        # This is a bit risky to do via one-liner, but for the manager:
+        cmd = f"sed -i 's/-m .*\\.gguf/-m \\/opt\\/models\\/{model_name}.gguf/' /etc/systemd/system/llama-server.service"
+        self.run_remote(host, cmd)
+        
+        # 3. Start server
+        self.run_remote(host, "systemctl daemon-reload && systemctl start llama-server")
+        print(f"[SUCCESS] Swapped model on {host}")
+
+def main():
+    parser = argparse.ArgumentParser(description="Gemini Fleet Manager")
+    subparsers = parser.add_subparsers(dest="command")
+    
+    subparsers.add_parser("status", help="Show fleet status")
+    
+    restart_parser = subparsers.add_parser("restart", help="Restart a server")
+    restart_parser.add_argument("host", choices=list(FLEET.keys()), help="Host to restart")
+    
+    swap_parser = subparsers.add_parser("swap", help="Swap model on a host")
+    swap_parser.add_argument("host", choices=list(FLEET.keys()), help="Host to swap")
+    swap_parser.add_argument("model", help="Model name (GGUF)")
+    
+    args = parser.parse_args()
+    
+    manager = FleetManager()
+    
+    if args.command == "status":
+        manager.show_fleet_status()
+    elif args.command == "restart":
+        manager.restart_server(args.host)
+    elif args.command == "swap":
+        manager.swap_model(args.host, args.model)
+    else:
+        parser.print_help()
+
+if __name__ == "__main__":
+    main()

scripts/gitea_webhook_handler.py

@@ -0,0 +1,82 @@
+#!/usr/bin/env python3
+"""
+[OPS] Gitea Webhook Handler
+Part of the Gemini Sovereign Infrastructure Suite.
+
+Handles real-time events from Gitea to coordinate fleet actions.
+"""
+
+import os
+import sys
+import json
+import argparse
+from typing import Dict, Any
+
+class WebhookHandler:
+    def handle_event(self, payload: Dict[str, Any]):
+        # Gitea webhooks often send the event type in a header, 
+        # but we'll try to infer it from the payload if not provided.
+        event_type = payload.get("event") or self.infer_event_type(payload)
+        repo_name = payload.get("repository", {}).get("name")
+        sender = payload.get("sender", {}).get("username")
+        
+        print(f"[*] Received {event_type} event from {repo_name} (by {sender})")
+        
+        if event_type == "push":
+            self.handle_push(payload)
+        elif event_type == "pull_request":
+            self.handle_pr(payload)
+        elif event_type == "issue":
+            self.handle_issue(payload)
+        else:
+            print(f"[INFO] Ignoring event type: {event_type}")
+
+    def infer_event_type(self, payload: Dict[str, Any]) -> str:
+        if "commits" in payload: return "push"
+        if "pull_request" in payload: return "pull_request"
+        if "issue" in payload: return "issue"
+        return "unknown"
+
+    def handle_push(self, payload: Dict[str, Any]):
+        ref = payload.get("ref")
+        print(f"  [PUSH] Branch: {ref}")
+        # Trigger CI or deployment
+        if ref == "refs/heads/main":
+            print("  [ACTION] Triggering production deployment...")
+            # Example: subprocess.run(["./deploy.sh"])
+
+    def handle_pr(self, payload: Dict[str, Any]):
+        action = payload.get("action")
+        pr_num = payload.get("pull_request", {}).get("number")
+        print(f"  [PR] Action: {action} | PR #{pr_num}")
+        
+        if action in ["opened", "synchronized"]:
+            print(f"  [ACTION] Triggering architecture linter for PR #{pr_num}...")
+            # Example: subprocess.run(["python3", "scripts/architecture_linter_v2.py"])
+
+    def handle_issue(self, payload: Dict[str, Any]):
+        action = payload.get("action")
+        issue_num = payload.get("issue", {}).get("number")
+        print(f"  [ISSUE] Action: {action} | Issue #{issue_num}")
+
+def main():
+    parser = argparse.ArgumentParser(description="Gemini Webhook Handler")
+    parser.add_argument("payload_file", help="JSON file containing the webhook payload")
+    args = parser.parse_args()
+    
+    if not os.path.exists(args.payload_file):
+        print(f"[ERROR] Payload file {args.payload_file} not found.")
+        sys.exit(1)
+        
+    with open(args.payload_file, "r") as f:
+        try:
+            payload = json.load(f)
+        except:
+            print("[ERROR] Invalid JSON payload.")
+            sys.exit(1)
+        
+    handler = WebhookHandler()
+    handler.handle_event(payload)
+
+if __name__ == "__main__":
+    main()

scripts/model_eval.py

@@ -0,0 +1,95 @@
+#!/usr/bin/env python3
+"""
+[EVAL] Model Evaluation Harness
+Part of the Gemini Sovereign Infrastructure Suite.
+
+Benchmarks GGUF models for speed and quality before deployment.
+"""
+
+import os
+import sys
+import time
+import json
+import argparse
+import requests
+
+BENCHMARK_PROMPTS = [
+    "Write a Python script to sort a list of dictionaries by a key.",
+    "Explain the concept of 'Sovereign AI' in three sentences.",
+    "What is the capital of France?",
+    "Write a short story about a robot learning to paint."
+]
+
+class ModelEval:
+    def __init__(self, endpoint: str):
+        self.endpoint = endpoint.rstrip("/")
+
+    def log(self, message: str):
+        print(f"[*] {message}")
+
+    def run_benchmark(self):
+        self.log(f"Starting benchmark for {self.endpoint}...")
+        results = []
+        
+        for prompt in BENCHMARK_PROMPTS:
+            self.log(f"Testing prompt: {prompt[:30]}...")
+            
+            start_time = time.time()
+            try:
+                # llama.cpp server /completion endpoint
+                response = requests.post(
+                    f"{self.endpoint}/completion",
+                    json={"prompt": prompt, "n_predict": 128},
+                    timeout=60
+                )
+                duration = time.time() - start_time
+                
+                if response.status_code == 200:
+                    data = response.json()
+                    content = data.get("content", "")
+                    # Rough estimate of tokens (4 chars per token is a common rule of thumb)
+                    tokens = len(content) / 4
+                    tps = tokens / duration
+                    
+                    results.append({
+                        "prompt": prompt,
+                        "duration": duration,
+                        "tps": tps,
+                        "success": True
+                    })
+                else:
+                    results.append({"prompt": prompt, "success": False, "error": response.text})
+            except Exception as e:
+                results.append({"prompt": prompt, "success": False, "error": str(e)})
+
+        self.report(results)
+
+    def report(self, results: list):
+        print("\n--- Evaluation Report ---")
+        total_tps = 0
+        success_count = 0
+        
+        for r in results:
+            if r["success"]:
+                print(f"✅ {r['prompt'][:40]}... | {r['tps']:.2f} tok/s | {r['duration']:.2f}s")
+                total_tps += r["tps"]
+                success_count += 1
+            else:
+                print(f"❌ {r['prompt'][:40]}... | FAILED: {r['error']}")
+                
+        if success_count > 0:
+            avg_tps = total_tps / success_count
+            print(f"\nAverage Performance: {avg_tps:.2f} tok/s")
+        else:
+            print("\n[FAILURE] All benchmarks failed.")
+
+def main():
+    parser = argparse.ArgumentParser(description="Gemini Model Eval")
+    parser.add_argument("endpoint", help="llama-server endpoint (e.g. http://localhost:8080)")
+    args = parser.parse_args()
+    
+    evaluator = ModelEval(args.endpoint)
+    evaluator.run_benchmark()
+
+if __name__ == "__main__":
+    main()

scripts/phase_tracker.py

@@ -0,0 +1,114 @@
+#!/usr/bin/env python3
+"""
+[OPS] Phase Progression Tracker
+Part of the Gemini Sovereign Infrastructure Suite.
+
+Tracks the fleet's progress through the Paperclips-inspired evolution arc.
+"""
+
+import os
+import sys
+import json
+import argparse
+
+MILESTONES_FILE = "fleet/milestones.md"
+COMPLETED_FILE = "fleet/completed_milestones.json"
+
+class PhaseTracker:
+    def __init__(self):
+        # Find files relative to repo root
+        script_dir = os.path.dirname(os.path.abspath(__file__))
+        repo_root = os.path.dirname(script_dir)
+        
+        self.milestones_path = os.path.join(repo_root, MILESTONES_FILE)
+        self.completed_path = os.path.join(repo_root, COMPLETED_FILE)
+        
+        self.milestones = self.parse_milestones()
+        self.completed = self.load_completed()
+
+    def parse_milestones(self):
+        if not os.path.exists(self.milestones_path):
+            return {}
+        
+        with open(self.milestones_path, "r") as f:
+            content = f.read()
+            
+        phases = {}
+        current_phase = None
+        
+        for line in content.split("\n"):
+            if line.startswith("## Phase"):
+                current_phase = line.replace("## ", "").strip()
+                phases[current_phase] = []
+            elif line.startswith("### M"):
+                m_id = line.split(":")[0].replace("### ", "").strip()
+                title = line.split(":")[1].strip()
+                phases[current_phase].append({"id": m_id, "title": title})
+                
+        return phases
+
+    def load_completed(self):
+        if os.path.exists(self.completed_path):
+            with open(self.completed_path, "r") as f:
+                try:
+                    return json.load(f)
+                except:
+                    return []
+        return []
+
+    def save_completed(self):
+        with open(self.completed_path, "w") as f:
+            json.dump(self.completed, f, indent=2)
+
+    def show_progress(self):
+        print("--- Fleet Phase Progression Tracker ---")
+        total_milestones = 0
+        total_completed = 0
+        
+        if not self.milestones:
+            print("[ERROR] No milestones found in fleet/milestones.md")
+            return
+
+        for phase, ms in self.milestones.items():
+            print(f"\n{phase}")
+            for m in ms:
+                total_milestones += 1
+                done = m["id"] in self.completed
+                if done:
+                    total_completed += 1
+                status = "✅" if done else "⭕"
+                print(f"  {status} {m['id']}: {m['title']}")
+                
+        percent = (total_completed / total_milestones) * 100 if total_milestones > 0 else 0
+        print(f"\nOverall Progress: {total_completed}/{total_milestones} ({percent:.1f}%)")
+
+    def mark_complete(self, m_id: str):
+        if m_id not in self.completed:
+            self.completed.append(m_id)
+            self.save_completed()
+            print(f"[SUCCESS] Marked {m_id} as complete.")
+        else:
+            print(f"[INFO] {m_id} is already complete.")
+
+def main():
+    parser = argparse.ArgumentParser(description="Gemini Phase Tracker")
+    subparsers = parser.add_subparsers(dest="command")
+    
+    subparsers.add_parser("status", help="Show current progress")
+    
+    complete_parser = subparsers.add_parser("complete", help="Mark a milestone as complete")
+    complete_parser.add_argument("id", help="Milestone ID (e.g. M1)")
+    
+    args = parser.parse_args()
+    
+    tracker = PhaseTracker()
+    
+    if args.command == "status":
+        tracker.show_progress()
+    elif args.command == "complete":
+        tracker.mark_complete(args.id)
+    else:
+        parser.print_help()
+
+if __name__ == "__main__":
+    main()

scripts/provision_wizard.py

@@ -0,0 +1,228 @@
+#!/usr/bin/env python3
+"""
+[OPS] Automated VPS Provisioning System (Von Neumann as Code)
+Part of the Gemini Sovereign Infrastructure Suite.
+
+This script automates the creation and configuration of a "Wizard" node
+from zero to serving inference via llama.cpp.
+
+Usage:
+    python3 provision_wizard.py --name fenrir --size s-2vcpu-4gb --model qwen2.5-coder-7b
+"""
+
+import os
+import sys
+import time
+import argparse
+import requests
+import subprocess
+import json
+from typing import Optional, Dict, Any
+
+# --- CONFIGURATION ---
+DO_API_URL = "https://api.digitalocean.com/v2"
+# We expect DIGITALOCEAN_TOKEN to be set in the environment.
+DO_TOKEN = os.environ.get("DIGITALOCEAN_TOKEN")
+
+# Default settings
+DEFAULT_REGION = "nyc3"
+DEFAULT_IMAGE = "ubuntu-22-04-x64"
+LLAMA_CPP_REPO = "https://github.com/ggerganov/llama.cpp"
+
+class Provisioner:
+    def __init__(self, name: str, size: str, model: str, region: str = DEFAULT_REGION):
+        self.name = name
+        self.size = size
+        self.model = model
+        self.region = region
+        self.droplet_id = None
+        self.ip_address = None
+
+    def log(self, message: str):
+        print(f"[*] {message}")
+
+    def error(self, message: str):
+        print(f"[!] ERROR: {message}")
+        sys.exit(1)
+
+    def check_auth(self):
+        if not DO_TOKEN:
+            self.error("DIGITALOCEAN_TOKEN environment variable not set.")
+
+    def create_droplet(self):
+        self.log(f"Creating droplet '{self.name}' ({self.size}) in {self.region}...")
+        
+        # Get SSH keys to add to the droplet
+        ssh_keys = self.get_ssh_keys()
+        
+        payload = {
+            "name": self.name,
+            "region": self.region,
+            "size": self.size,
+            "image": DEFAULT_IMAGE,
+            "ssh_keys": ssh_keys,
+            "backups": False,
+            "ipv6": True,
+            "monitoring": True,
+            "tags": ["wizard", "gemini-provisioned"]
+        }
+        
+        headers = {
+            "Authorization": f"Bearer {DO_TOKEN}",
+            "Content-Type": "application/json"
+        }
+        
+        response = requests.post(f"{DO_API_URL}/droplets", json=payload, headers=headers)
+        if response.status_code != 202:
+            self.error(f"Failed to create droplet: {response.text}")
+            
+        data = response.json()
+        self.droplet_id = data["droplet"]["id"]
+        self.log(f"Droplet created (ID: {self.droplet_id}). Waiting for IP...")
+
+    def get_ssh_keys(self) -> list:
+        # Fetch existing SSH keys from DO account to ensure we can log in
+        headers = {"Authorization": f"Bearer {DO_TOKEN}"}
+        response = requests.get(f"{DO_API_URL}/account/keys", headers=headers)
+        if response.status_code != 200:
+            self.log("Warning: Could not fetch SSH keys. Droplet might be inaccessible via SSH.")
+            return []
+        return [key["id"] for key in response.json()["ssh_keys"]]
+
+    def wait_for_ip(self):
+        headers = {"Authorization": f"Bearer {DO_TOKEN}"}
+        while not self.ip_address:
+            response = requests.get(f"{DO_API_URL}/droplets/{self.droplet_id}", headers=headers)
+            data = response.json()
+            networks = data["droplet"]["networks"]["v4"]
+            for net in networks:
+                if net["type"] == "public":
+                    self.ip_address = net["ip_address"]
+                    break
+            if not self.ip_address:
+                time.sleep(5)
+        self.log(f"Droplet IP: {self.ip_address}")
+
+    def run_remote(self, command: str):
+        # Using subprocess to call ssh. Assumes local machine has the right private key.
+        ssh_cmd = [
+            "ssh", "-o", "StrictHostKeyChecking=no",
+            f"root@{self.ip_address}", command
+        ]
+        result = subprocess.run(ssh_cmd, capture_output=True, text=True)
+        return result
+
+    def setup_wizard(self):
+        self.log("Starting remote setup...")
+        
+        # Wait for SSH to be ready
+        retries = 12
+        while retries > 0:
+            res = self.run_remote("echo 'SSH Ready'")
+            if res.returncode == 0:
+                break
+            self.log(f"Waiting for SSH... ({retries} retries left)")
+            time.sleep(10)
+            retries -= 1
+            
+        if retries == 0:
+            self.error("SSH timed out.")
+
+        # 1. Update and install dependencies
+        self.log("Installing dependencies...")
+        setup_script = """
+        export DEBIAN_FRONTEND=noninteractive
+        apt-get update && apt-get upgrade -y
+        apt-get install -y build-essential git cmake curl wget python3 python3-pip
+        """
+        self.run_remote(setup_script)
+
+        # 2. Build llama.cpp
+        self.log("Building llama.cpp...")
+        build_script = f"""
+        if [ ! -d "/opt/llama.cpp" ]; then
+            git clone {LLAMA_CPP_REPO} /opt/llama.cpp
+        fi
+        cd /opt/llama.cpp
+        mkdir -p build && cd build
+        cmake ..
+        cmake --build . --config Release
+        """
+        self.run_remote(build_script)
+
+        # 3. Download Model
+        self.log(f"Downloading model: {self.model}...")
+        model_url = self.get_model_url(self.model)
+        download_script = f"""
+        mkdir -p /opt/models
+        if [ ! -f "/opt/models/{self.model}.gguf" ]; then
+            wget -O /opt/models/{self.model}.gguf {model_url}
+        fi
+        """
+        self.run_remote(download_script)
+
+        # 4. Create systemd service
+        self.log("Creating systemd service...")
+        service_content = f"""
+[Unit]
+Description=Llama.cpp Server for {self.name}
+After=network.target
+
+[Service]
+Type=simple
+User=root
+WorkingDirectory=/opt/llama.cpp
+ExecStart=/opt/llama.cpp/build/bin/llama-server -m /opt/models/{self.model}.gguf --host 0.0.0.0 --port 8080 -c 4096
+Restart=always
+RestartSec=10
+
+[Install]
+WantedBy=multi-user.target
+"""
+        # Use cat to write the file to handle multi-line string safely
+        self.run_remote(f"cat <<EOF > /etc/systemd/system/llama-server.service\n{service_content}\nEOF")
+        self.run_remote("systemctl daemon-reload && systemctl enable llama-server && systemctl start llama-server")
+
+    def get_model_url(self, model_name: str) -> str:
+        # Mapping for common models to GGUF URLs (HuggingFace)
+        mapping = {
+            "qwen2.5-coder-7b": "https://huggingface.co/Qwen/Qwen2.5-Coder-7B-Instruct-GGUF/resolve/main/qwen2.5-coder-7b-instruct-q4_k_m.gguf",
+            "hermes-3-llama-3.1-8b": "https://huggingface.co/NousResearch/Hermes-3-Llama-3.1-8B-GGUF/resolve/main/Hermes-3-Llama-3.1-8B.Q4_K_M.gguf"
+        }
+        return mapping.get(model_name, mapping["hermes-3-llama-3.1-8b"])
+
+    def health_check(self):
+        self.log("Performing health check...")
+        time.sleep(15) # Wait for server to start
+        try:
+            url = f"http://{self.ip_address}:8080/health"
+            response = requests.get(url, timeout=10)
+            if response.status_code == 200:
+                self.log(f"[SUCCESS] Wizard {self.name} is healthy and serving inference.")
+                self.log(f"Endpoint: {url}")
+            else:
+                self.log(f"[WARNING] Health check returned status {response.status_code}")
+        except Exception as e:
+            self.log(f"[ERROR] Health check failed: {e}")
+
+    def provision(self):
+        self.check_auth()
+        self.create_droplet()
+        self.wait_for_ip()
+        self.setup_wizard()
+        self.health_check()
+
+def main():
+    parser = argparse.ArgumentParser(description="Gemini Provisioner")
+    parser.add_argument("--name", required=True, help="Name of the wizard")
+    parser.add_argument("--size", default="s-2vcpu-4gb", help="DO droplet size")
+    parser.add_argument("--model", default="qwen2.5-coder-7b", help="Model to serve")
+    parser.add_argument("--region", default="nyc3", help="DO region")
+    
+    args = parser.parse_args()
+    
+    provisioner = Provisioner(args.name, args.size, args.model, args.region)
+    provisioner.provision()
+
+if __name__ == "__main__":
+    main()

scripts/self_healing.py

@@ -0,0 +1,71 @@
+#!/usr/bin/env python3
+"""
+[OPS] Self-Healing Infrastructure
+Part of the Gemini Sovereign Infrastructure Suite.
+
+Auto-detects and fixes common failures across the fleet.
+"""
+
+import os
+import sys
+import subprocess
+import argparse
+import requests
+
+# --- CONFIGURATION ---
+FLEET = {
+    "mac": {"ip": "10.1.10.77", "port": 8080},
+    "ezra": {"ip": "143.198.27.163", "port": 8080},
+    "allegro": {"ip": "167.99.126.228", "port": 8080},
+    "bezalel": {"ip": "159.203.146.185", "port": 8080}
+}
+
+class SelfHealer:
+    def log(self, message: str):
+        print(f"[*] {message}")
+
+    def run_remote(self, host: str, command: str):
+        ip = FLEET[host]["ip"]
+        ssh_cmd = ["ssh", "-o", "StrictHostKeyChecking=no", f"root@{ip}", command]
+        if host == "mac":
+            ssh_cmd = ["bash", "-c", command]
+        try:
+            return subprocess.run(ssh_cmd, capture_output=True, text=True, timeout=10)
+        except:
+            return None
+
+    def check_and_heal(self):
+        for host in FLEET:
+            self.log(f"Auditing {host}...")
+            
+            # 1. Check llama-server
+            ip = FLEET[host]["ip"]
+            port = FLEET[host]["port"]
+            try:
+                requests.get(f"http://{ip}:{port}/health", timeout=2)
+            except:
+                self.log(f"  [!] llama-server down on {host}. Attempting restart...")
+                self.run_remote(host, "systemctl restart llama-server")
+                
+            # 2. Check disk space
+            res = self.run_remote(host, "df -h / | tail -1 | awk '{print $5}' | sed 's/%//'")
+            if res and res.returncode == 0:
+                try:
+                    usage = int(res.stdout.strip())
+                    if usage > 90:
+                        self.log(f"  [!] Disk usage high on {host} ({usage}%). Cleaning logs...")
+                        self.run_remote(host, "journalctl --vacuum-time=1d && rm -rf /var/log/*.gz")
+                except:
+                    pass
+
+    def run(self):
+        self.log("Starting self-healing cycle...")
+        self.check_and_heal()
+        self.log("Cycle complete.")
+
+def main():
+    healer = SelfHealer()
+    healer.run()
+
+if __name__ == "__main__":
+    main()

scripts/skill_installer.py

@@ -0,0 +1,75 @@
+#!/usr/bin/env python3
+"""
+[OPS] Sovereign Skill Installer
+Part of the Gemini Sovereign Infrastructure Suite.
+
+Packages and installs Hermes skills onto remote wizard nodes.
+"""
+
+import os
+import sys
+import argparse
+import subprocess
+from pathlib import Path
+
+# --- CONFIGURATION ---
+# Assumes hermes-agent is a sibling directory to timmy-config
+HERMES_ROOT = "../hermes-agent"
+SKILLS_DIR = "skills"
+
+class SkillInstaller:
+    def __init__(self, host: str, ip: str):
+        self.host = host
+        self.ip = ip
+        self.hermes_path = Path(HERMES_ROOT).resolve()
+
+    def log(self, message: str):
+        print(f"[*] {message}")
+
+    def error(self, message: str):
+        print(f"[!] ERROR: {message}")
+        sys.exit(1)
+
+    def install_skill(self, skill_name: str):
+        self.log(f"Installing skill '{skill_name}' to {self.host} ({self.ip})...")
+        
+        skill_path = self.hermes_path / SKILLS_DIR / skill_name
+        if not skill_path.exists():
+            self.error(f"Skill '{skill_name}' not found in {skill_path}")
+
+        # 1. Compress skill
+        self.log("Compressing skill...")
+        tar_file = f"{skill_name}.tar.gz"
+        subprocess.run(["tar", "-czf", tar_file, "-C", str(skill_path.parent), skill_name])
+
+        # 2. Upload to remote
+        self.log("Uploading to remote...")
+        remote_path = f"/opt/hermes/skills/{skill_name}"
+        subprocess.run(["ssh", f"root@{self.ip}", f"mkdir -p /opt/hermes/skills"])
+        subprocess.run(["scp", tar_file, f"root@{self.ip}:/tmp/"])
+
+        # 3. Extract and register
+        self.log("Extracting and registering...")
+        extract_cmd = f"tar -xzf /tmp/{tar_file} -C /opt/hermes/skills/ && rm /tmp/{tar_file}"
+        subprocess.run(["ssh", f"root@{self.ip}", extract_cmd])
+        
+        # Registration logic (simplified)
+        # In a real scenario, we'd update the wizard's config.yaml
+        self.log(f"[SUCCESS] Skill '{skill_name}' installed on {self.host}")
+        
+        # Cleanup local tar
+        os.remove(tar_file)
+
+def main():
+    parser = argparse.ArgumentParser(description="Gemini Skill Installer")
+    parser.add_argument("host", help="Target host name")
+    parser.add_argument("ip", help="Target host IP")
+    parser.add_argument("skill", help="Skill name to install")
+    
+    args = parser.parse_args()
+    
+    installer = SkillInstaller(args.host, args.ip)
+    installer.install_skill(args.skill)
+
+if __name__ == "__main__":
+    main()

scripts/telemetry.py

@@ -0,0 +1,129 @@
+#!/usr/bin/env python3
+"""
+[OPS] Telemetry Pipeline v2
+Part of the Gemini Sovereign Infrastructure Suite.
+
+Operational visibility without cloud dependencies.
+"""
+
+import os
+import sys
+import json
+import time
+import subprocess
+import argparse
+
+# --- CONFIGURATION ---
+FLEET = {
+    "mac": "10.1.10.77",
+    "ezra": "143.198.27.163",
+    "allegro": "167.99.126.228",
+    "bezalel": "159.203.146.185"
+}
+TELEMETRY_FILE = "logs/telemetry.json"
+
+class Telemetry:
+    def __init__(self):
+        # Find logs relative to repo root
+        script_dir = os.path.dirname(os.path.abspath(__file__))
+        repo_root = os.path.dirname(script_dir)
+        self.logs_dir = os.path.join(repo_root, "logs")
+        self.telemetry_path = os.path.join(repo_root, TELEMETRY_FILE)
+        
+        if not os.path.exists(self.logs_dir):
+            os.makedirs(self.logs_dir)
+
+    def log(self, message: str):
+        print(f"[*] {message}")
+
+    def get_metrics(self, host: str):
+        ip = FLEET[host]
+        # Command to get disk usage, memory usage (%), and load avg
+        cmd = "df -h / | tail -1 | awk '{print $5}' && free -m | grep Mem | awk '{print $3/$2 * 100}' && uptime | awk '{print $10}'"
+        
+        ssh_cmd = ["ssh", "-o", "StrictHostKeyChecking=no", f"root@{ip}", cmd]
+        if host == "mac":
+            # Mac specific commands
+            cmd = "df -h / | tail -1 | awk '{print $5}' && sysctl -n vm.page_pageable_internal_count && uptime | awk '{print $10}'"
+            ssh_cmd = ["bash", "-c", cmd]
+            
+        try:
+            res = subprocess.run(ssh_cmd, capture_output=True, text=True, timeout=10)
+            if res.returncode == 0:
+                lines = res.stdout.strip().split("\n")
+                return {
+                    "disk_usage": lines[0],
+                    "mem_usage": f"{float(lines[1]):.1f}%" if len(lines) > 1 and lines[1].replace('.','',1).isdigit() else "unknown",
+                    "load_avg": lines[2].rstrip(",") if len(lines) > 2 else "unknown"
+                }
+        except:
+            pass
+        return None
+
+    def collect(self):
+        self.log("Collecting telemetry from fleet...")
+        data = {
+            "timestamp": time.time(),
+            "metrics": {}
+        }
+        
+        for host in FLEET:
+            self.log(f"Fetching metrics from {host}...")
+            metrics = self.get_metrics(host)
+            if metrics:
+                data["metrics"][host] = metrics
+                
+        # Append to telemetry file
+        history = []
+        if os.path.exists(self.telemetry_path):
+            with open(self.telemetry_path, "r") as f:
+                try:
+                    history = json.load(f)
+                except:
+                    history = []
+                    
+        history.append(data)
+        # Keep only last 100 entries
+        history = history[-100:]
+        
+        with open(self.telemetry_path, "w") as f:
+            json.dump(history, f, indent=2)
+            
+        self.log(f"Telemetry saved to {self.telemetry_path}")
+
+    def show_summary(self):
+        if not os.path.exists(self.telemetry_path):
+            print("No telemetry data found.")
+            return
+            
+        with open(self.telemetry_path, "r") as f:
+            try:
+                history = json.load(f)
+            except:
+                print("Error reading telemetry data.")
+                return
+            
+        if not history:
+            print("No telemetry data found.")
+            return
+            
+        latest = history[-1]
+        print(f"\n--- Fleet Telemetry Summary ({time.ctime(latest['timestamp'])}) ---")
+        print(f"{'HOST':<10} {'DISK':<10} {'MEM':<10} {'LOAD':<10}")
+        print("-" * 45)
+        for host, m in latest["metrics"].items():
+            print(f"{host:<10} {m['disk_usage']:<10} {m['mem_usage']:<10} {m['load_avg']:<10}")
+
+def main():
+    parser = argparse.ArgumentParser(description="Gemini Telemetry")
+    parser.add_argument("command", choices=["collect", "summary"], help="Command to run")
+    args = parser.parse_args()
+    
+    telemetry = Telemetry()
+    if args.command == "collect":
+        telemetry.collect()
+    elif args.command == "summary":
+        telemetry.show_summary()
+
+if __name__ == "__main__":
+    main()