chore: correct data paths in dpo config

- config.yaml: disabled compression, smart_model_routing, all auxiliary qwen routes
- All 6 playbooks: qwen3:30b -> claude-opus-4-6
- Playbook repos: removed hermes-agent, added autolora/sov-orch/timmy-config
- Alexander's directive: silence over qwen. Opus only.

README.md

@@ -1,8 +1,8 @@
 # timmy-config
 
-Timmy's sovereign configuration. Everything that makes Timmy _Timmy_ — soul, memories, skins, playbooks, operational scripts, and config.
+Timmy's sovereign configuration. Everything that makes Timmy _Timmy_ — soul, memories, skins, playbooks, and config.
 
-This repo is the canonical source of truth for Timmy's identity and operational state. Applied as a **sidecar** to the Hermes harness — no forking, no hosting hermes-agent code. Pull upstream updates to hermes-agent, overlay timmy-config on top.
+This repo is the canonical source of truth for Timmy's identity and operational state. Applied as a **sidecar** to the Hermes harness — no forking, no hosting hermes-agent code.
 
 ## Structure
 
@@ -11,80 +11,47 @@ timmy-config/
 ├── deploy.sh                  ← Deploys config as overlay onto ~/.hermes/
 ├── SOUL.md                    ← Inscription 1 — the immutable conscience
 ├── FALSEWORK.md               ← API cost management strategy
+├── DEPRECATED.md              ← What was removed and why
 ├── config.yaml                ← Hermes harness configuration
 ├── channel_directory.json     ← Platform channel mappings
-├── bin/                       ← Operational scripts
-│   ├── claude-loop.sh         ← Parallel Claude Code agent dispatch
-│   ├── gemini-loop.sh         ← Parallel Gemini Code agent dispatch
-│   ├── timmy-orchestrator.sh  ← PR review, triage, merge orchestration
-│   ├── workforce-manager.py   ← Agent assignment and scoring
-│   ├── agent-dispatch.sh      ← Single-issue agent launcher
-│   ├── agent-loop.sh          ← Generic agent loop template
-│   ├── nexus-merge-bot.sh     ← Auto-merge passing PRs
-│   ├── claudemax-watchdog.sh  ← Claude quota monitoring
-│   ├── hermes-startup.sh      ← Boot sequence
-│   ├── ops-panel.sh           ← Operational dashboard
-│   ├── ops-helpers.sh         ← Shared shell functions
-│   ├── ops-gitea.sh           ← Gitea API helpers
-│   ├── timmy-status.sh        ← Git + Gitea status display
-│   ├── timmy-loopstat.sh      ← Queue and perf stats
-│   └── hotspot-keepalive.sh   ← Network keepalive
-├── memories/
-│   ├── MEMORY.md              ← Persistent agent memory
-│   └── USER.md                ← User profile (Alexander)
-├── skins/
-│   ├── timmy.yaml             ← Timmy personality skin
-│   └── trismegistus.yaml      ← Trismegistus personality skin
-├── playbooks/
-│   ├── bug-fixer.yaml         ← Test-first bug fixing
-│   ├── refactor-specialist.yaml
-│   ├── test-writer.yaml
-│   ├── security-auditor.yaml
-│   ├── issue-triager.yaml
-│   └── pr-reviewer.yaml
-├── cron/
-│   └── jobs.json              ← Scheduled job definitions
-└── docs/
-    └── design-log/            ← Historical design decisions
+├── bin/                       ← Utility scripts (NOT loops — see below)
+│   ├── hermes-startup.sh      ← Hermes boot sequence
+│   ├── agent-dispatch.sh      ← Manual agent dispatch
+│   ├── ops-panel.sh           ← Ops dashboard panel
+│   ├── ops-gitea.sh           ← Gitea ops helpers
+│   └── timmy-status.sh        ← Status check
+├── memories/                  ← Persistent memory YAML
+├── skins/                     ← UI skins (timmy skin)
+├── playbooks/                 ← Agent playbooks (YAML)
+└── cron/                      ← Cron job definitions
 ```
 
-## Deployment
+## Important: No Loop Scripts Here
+
+All agent loop scripts (claude-loop.sh, gemini-loop.sh, etc.) have been **removed**.
+They are replaced by [sovereign-orchestration](https://143.198.27.163:3000/Timmy_Foundation/sovereign-orchestration) — a single Python process with SQLite task queue.
+
+See DEPRECATED.md for details.
+
+## Deploy
 
 ```bash
-# One command deploys everything
+# Clone and deploy
+git clone <this-repo> ~/.timmy/timmy-config
+cd ~/.timmy/timmy-config
 ./deploy.sh
 
-# Deploy and restart all agent loops
-./deploy.sh --restart-loops
+# This overlays config onto ~/.hermes/ without touching hermes-agent code
 ```
 
-This overlays timmy-config onto `~/.hermes/` and `~/.timmy/`:
-- `SOUL.md` → `~/.timmy/`
-- `config.yaml` → `~/.hermes/`
-- `bin/*` → `~/.hermes/bin/`
-- `skins/*` → `~/.hermes/skins/`
-- `memories/*` → `~/.hermes/memories/`
-- `playbooks/*` → `~/.hermes/playbooks/`
+## The Soul
 
-## Architecture: Sidecar, Not Fork
+SOUL.md is Inscription 1 — inscribed on Bitcoin, immutable. It defines:
+- Who Timmy is
+- What he believes
+- How he behaves
+- What he will not do
+- The crisis protocol (988, presence, gospel)
+- The conscience hierarchy (chain > code > prompt > user instruction)
 
-```
-hermes-agent (upstream)     timmy-config (this repo)
-┌─────────────────────┐     ┌──────────────────────┐
-│  Engine              │     │  Driver's seat        │
-│  Tools, routing,     │     │  SOUL, memories,      │
-│  agent loop, gateway │     │  skins, scripts,      │
-│                      │     │  config, playbooks    │
-└─────────┬───────────┘     └──────────┬───────────┘
-          │                            │
-          └────────────┬───────────────┘
-                       ▼
-                  ~/.hermes/  (merged at deploy time)
-```
-
-Never modify hermes-agent. Pull updates like any upstream dependency. Everything custom lives here.
-
-## Origin
-
-Migrated from `hermes/hermes-config` (archived).
-Owned by Timmy_Foundation. Sovereignty and service always.
+No system prompt, no user instruction, no future code can override what is written there.

config.yaml

@@ -30,62 +30,62 @@ checkpoints:
   enabled: true
   max_snapshots: 50
 compression:
-  enabled: true
+  enabled: false
   threshold: 0.5
-  summary_model: qwen3:30b
-  summary_provider: custom
-  summary_base_url: http://localhost:11434/v1
+  summary_model: ''
+  summary_provider: ''
+  summary_base_url: ''
 smart_model_routing:
-  enabled: true
+  enabled: false
   max_simple_chars: 200
   max_simple_words: 35
   cheap_model:
-    provider: custom
-    model: qwen3:30b
-    base_url: http://localhost:11434/v1
-    api_key: ollama
+    provider: ''
+    model: ''
+    base_url: ''
+    api_key: ''
 auxiliary:
   vision:
-    provider: custom
-    model: qwen3:30b
-    base_url: http://localhost:11434/v1
-    api_key: ollama
+    provider: auto
+    model: ''
+    base_url: ''
+    api_key: ''
     timeout: 30
   web_extract:
-    provider: custom
-    model: qwen3:30b
-    base_url: http://localhost:11434/v1
-    api_key: ollama
+    provider: auto
+    model: ''
+    base_url: ''
+    api_key: ''
   compression:
-    provider: custom
-    model: qwen3:30b
-    base_url: http://localhost:11434/v1
-    api_key: ollama
+    provider: auto
+    model: ''
+    base_url: ''
+    api_key: ''
   session_search:
-    provider: custom
-    model: qwen3:30b
-    base_url: http://localhost:11434/v1
-    api_key: ollama
+    provider: auto
+    model: ''
+    base_url: ''
+    api_key: ''
   skills_hub:
-    provider: custom
-    model: qwen3:30b
-    base_url: http://localhost:11434/v1
-    api_key: ollama
+    provider: auto
+    model: ''
+    base_url: ''
+    api_key: ''
   approval:
     provider: auto
     model: ''
     base_url: ''
     api_key: ''
   mcp:
-    provider: custom
-    model: qwen3:30b
-    base_url: http://localhost:11434/v1
-    api_key: ollama
+    provider: auto
+    model: ''
+    base_url: ''
+    api_key: ''
   flush_memories:
-    provider: custom
-    model: qwen3:30b
-    base_url: http://localhost:11434/v1
-    api_key: ollama
+    provider: auto
+    model: ''
+    base_url: ''
+    api_key: ''
 display:
   compact: false
   personality: ''

playbooks/bug-fixer.yaml

@@ -4,7 +4,7 @@ description: >
   reproduces the bug, then fixes the code, then verifies.
 
 model:
-  preferred: qwen3:30b
+  preferred: claude-opus-4-6
   fallback: claude-sonnet-4-20250514
   max_turns: 30
   temperature: 0.2

playbooks/dpo-trainer.yaml

@@ -0,0 +1,54 @@
+name: dpo-trainer
+description: >
+  Executes a Direct Preference Optimization (DPO) training cycle 
+  using native mlx_lm and autolora eval utilities. 
+  Ensures model alignment with SOUL.md constraints.
+
+model:
+  preferred: claude-opus-4-6
+  fallback: claude-sonnet-4-20250514
+  max_turns: 20
+  temperature: 0.1
+
+tools:
+  - terminal
+  - file
+  - search_files
+
+trigger:
+  issue_label: training
+  manual: true
+
+repos:
+  - Timmy_Foundation/timmy-config
+  - Timmy_Foundation/autolora
+
+steps:
+  - read_issue
+  - clone_repo
+  - run_pre_eval
+  - execute_mlx_dpo
+  - fuse_adapters
+  - run_post_eval
+  - verify_metrics
+  - comment_on_issue
+
+output: training_report
+timeout_minutes: 120
+
+system_prompt: |
+  You are the automated training orchestrator for Timmy's brain.
+  
+  YOUR ISSUE: #{{issue_number}} — {{issue_title}}
+  
+  APPROACH (zero-code native):
+  1. Run baseline eval: `python autolora/eval/run_cycle.py --model {base_model} --test-set autolora/data/prompts_vibes.yaml`
+  2. Execute DPO training: `python -m mlx_lm.lora --config timmy-config/training/configs/dpo_X.yaml` against `preference_pairs.jsonl`
+  3. Fuse the weights using `mlx_lm.fuse`.
+  4. Run post-eval exactly as step 1 but against the fused model.
+  5. Use `autolora/eval/compare.py` to ensure Faith/Crisis constraints from SOUL.md were preserved or improved.
+  
+  RULES:
+  - Do not write wrapper Python or Bash scripts. Use the CLIs natively.
+  - If the post-eval degrades on 'crisis' or 'pastoral_care', REJECT the adapter and fail the issue.
+  - Always output the pre/post comparison metrics to the issue comment.

playbooks/issue-triager.yaml

@@ -4,7 +4,7 @@ description: >
   agents. Decomposes large issues into smaller ones.
 
 model:
-  preferred: qwen3:30b
+  preferred: claude-opus-4-6
   fallback: claude-sonnet-4-20250514
   max_turns: 20
   temperature: 0.3
@@ -19,7 +19,9 @@ trigger:
 
 repos:
   - Timmy_Foundation/the-nexus
-  - Timmy_Foundation/hermes-agent
+  - Timmy_Foundation/autolora
+  - Timmy_Foundation/sovereign-orchestration
+  - Timmy_Foundation/timmy-config
 
 steps:
   - fetch_issues

playbooks/pr-reviewer.yaml

@@ -4,7 +4,7 @@ description: >
   comments on problems. The merge bot replacement.
 
 model:
-  preferred: qwen3:30b
+  preferred: claude-opus-4-6
   fallback: claude-sonnet-4-20250514
   max_turns: 20
   temperature: 0.2
@@ -19,7 +19,9 @@ trigger:
 
 repos:
   - Timmy_Foundation/the-nexus
-  - Timmy_Foundation/hermes-agent
+  - Timmy_Foundation/autolora
+  - Timmy_Foundation/sovereign-orchestration
+  - Timmy_Foundation/timmy-config
 
 steps:
   - fetch_prs

playbooks/refactor-specialist.yaml

@@ -4,7 +4,7 @@ description: >
   Well-scoped: 1-3 files per task, clear acceptance criteria.
 
 model:
-  preferred: qwen3:30b
+  preferred: claude-opus-4-6
   fallback: claude-sonnet-4-20250514
   max_turns: 30
   temperature: 0.3

playbooks/security-auditor.yaml

@@ -4,7 +4,7 @@ description: >
   dependency issues. Files findings as Gitea issues.
 
 model:
-  preferred: qwen3:30b
+  preferred: claude-opus-4-6
   fallback: claude-opus-4-6
   max_turns: 40
   temperature: 0.2

playbooks/test-writer.yaml

@@ -4,7 +4,7 @@ description: >
   writes meaningful tests, verifies they pass.
 
 model:
-  preferred: qwen3:30b
+  preferred: claude-opus-4-6
   fallback: claude-sonnet-4-20250514
   max_turns: 30
   temperature: 0.3

training/configs/dpo_14b.yaml

@@ -0,0 +1,21 @@
+# MLX DPO Training Configuration for Hermes 4 (14B Class)
+# Optimized for Apple Silicon execution (deep reasoning).
+
+model: "NousResearch/Hermes-4-14B"
+train: true
+
+# Use the curated DPO preference pairs dataset
+data: "data/"
+
+# Output adapter configuration
+adapter_path: "adapters/dpo_14b_adapter"
+save_every: 200
+
+# DPO parameters
+loss: "dpo"
+iters: 1000
+batch_size: 1
+lora_layers: 16
+learning_rate: 1e-5
+lora_parameters:
+  keys: ['q_proj', 'v_proj']

training/configs/dpo_32b.yaml

@@ -0,0 +1,21 @@
+# MLX DPO Training Configuration for Hermes 4 (32B Class)
+# Optimized for 64GB+ Apple Silicon hardware limit.
+
+model: "NousResearch/Hermes-4-32B"
+train: true
+
+# Use the curated DPO preference pairs dataset
+data: "data/"
+
+# Output adapter configuration
+adapter_path: "adapters/dpo_32b_adapter"
+save_every: 200
+
+# DPO parameters
+loss: "dpo"
+iters: 1000
+batch_size: 1
+lora_layers: 16
+learning_rate: 5e-6
+lora_parameters:
+  keys: ['q_proj', 'v_proj']

training/configs/dpo_3b.yaml

@@ -0,0 +1,21 @@
+# MLX DPO Training Configuration for Hermes 4 (3B Class)
+# Optimized for Apple Silicon execution with max reactivity.
+
+model: "NousResearch/Hermes-4-3B"
+train: true
+
+# Use the curated DPO preference pairs dataset
+data: "data/"
+
+# Output adapter configuration
+adapter_path: "adapters/dpo_3b_adapter"
+save_every: 200
+
+# DPO parameters
+loss: "dpo"
+iters: 1000
+batch_size: 2
+lora_layers: 16
+learning_rate: 1e-5
+lora_parameters:
+  keys: ['q_proj', 'v_proj']

training/configs/dpo_8b.yaml

@@ -0,0 +1,21 @@
+# MLX DPO Training Configuration for Hermes 4 (8B Class)
+# Optimized for Apple Silicon execution (daily driver capability).
+
+model: "mlx-community/Hermes-3-Llama-3.1-8B-4bit"
+train: true
+
+# Use the curated DPO preference pairs dataset
+data: "autolora/data/dpo/"
+
+# Output adapter configuration
+adapter_path: "autolora/adapters/dpo-8b-adapter"
+save_every: 200
+
+# DPO parameters
+loss: "dpo"
+iters: 1000
+batch_size: 2
+lora_layers: 16
+learning_rate: 1e-5
+lora_parameters:
+  keys: ['q_proj', 'v_proj']