Move Zulip bot creds from inline to env/vault reference

The PM2 set lives on CT100/.24 (local, not CT116). CT116 runs systemd+docker, no PM2.
Process scope (5/5, 2026-08-09 ruling, ecosystem authoritative):
abiba-zulip, abiba-telegram, gitea-runner, spoton-service, zulip-watchdog

This prevents future dispatches from drifting to the wrong host.

README.md

@@ -87,7 +87,7 @@ prose run memory-audit-maintenance memory_threshold=90 verify_configs=true
 prose run hermes-config-template agent_name=syslog-devops default_model=claude-sonnet-4
 
 # Configure an agent with a different auxiliary model
-prose run hermes-config-template agent_name=syslog-code default_model=qwen3.6-27B-code auxiliary_model=gemma-4-12b
+prose run hermes-config-template agent_name=syslog-code default_model=qwen3.6-27B-code auxiliary_model=gpu-vision
 ```
 
 ### Option B: Manual Execution

abiba-zulip-restore.prose.md

@@ -1,4 +1,6 @@
 ---
+report_only_agents:
+  - koby  # ⛔ KOBY IS NEVER REPAIRED (Rule 17, 2026-08-17) — detect + report, never fix on .129
 kind: function
 name: abiba-zulip-restore
 description: >
@@ -11,6 +13,7 @@ version: 1.0.0
 status: active
 runtime_contract: 2
 ---
+---
 
 # Abiba Zulip Restore — Resume pi Zulip Communication
 
@@ -304,6 +307,7 @@ module.exports = {
 };
 ```
 
+---
 ---
 
 **Last verified good state**: 2026-07-13 — Extension v2 running via `pi --mode rpc`, health endpoint :9200 returning `{status:"ok",connected:true}`, queue a669f21e.

audit-hermes-config.py

@@ -179,7 +179,7 @@ def audit(path):
     )
 
     # --- No raw model names (Rule 7/8 spirit) ---
-    raw_names = {"gemma-4-12b", "qwen3.6-27B-code", "qwen3.6-35B-udq4", "ornith-1.0-35b"}
+    raw_names = {"gpu-vision", "qwen3.6-27B-code", "qwen3.6-35B-udq4", "ornith-1.0-35b"}
     for section_path, section_dict in [
         ("model", model), ("compression", comp),
         ("auxiliary.vision", aux.get("vision", {})),

contract-registry.yaml

@@ -1867,3 +1867,73 @@ contracts:
   last_run: null
   last_status: null
   drift_alerts: []
+# Koby Report-Only Registry (2026-08-17 — Captain)
+# ⛔ KOBY IS NEVER REPAIRED — detect + report, never fix on .129
+koby_report_only: true
+koby_host: "CT 111 (tdunna)"
+koby_ip: ".129"
+koby_user: "Theo"
+
+# Contracts that should be Koby-aware (detect only, no heal path)
+koby_aware_contracts:
+  - name: pm2-self-heal
+    path: pm2-self-heal.prose.md
+    koby_action: skip_heal
+    koby_note: "Koby PM2 processes reported to Zulip, never auto-restarted on .129"
+    
+  - name: zulip-health
+    path: zulip-health.prose.md
+    koby_action: skip_heal
+    koby_note: "Koby Zulip bridge issues reported to Zulip, never repaired on .129"
+    
+  - name: hermes-zulip-restore
+    path: hermes-zulip-restore.prose.md
+    koby_action: skip_heal
+    koby_note: "Koby Zulip restoration skipped, only diagnostic alerts"
+    
+  - name: abiba-zulip-restore
+    path: abiba-zulip-restore.prose.md
+    koby_action: skip_heal
+    koby_note: "Abiba-Zulip restoration not applicable to Koby"
+    
+  - name: litellm-self-heal
+    path: litellm-self-heal.prose.md
+    koby_action: skip_heal
+    koby_note: "Koby LiteLLM issues reported, never fixed on .129"
+    
+  - name: disk-gc-threat-response
+    path: disk-gc-threat-response.prose.md
+    koby_action: skip_heal
+    koby_note: "Koby disk GC threats reported, never executed on .129"
+    
+  - name: memory-fixer
+    path: memory-fixer.prose.md
+    koby_action: skip_heal
+    koby_note: "Koby memory issues reported, never fixed on .129"
+    
+  - name: memory-audit-maintenance
+    path: memory-audit-maintenance.prose.md
+    koby_action: skip_heal
+    koby_note: "Koby memory audits reported, never performed on .129"
+    
+  - name: gpu-self-heal
+    path: gpu-self-heal.prose.md
+    koby_action: skip_heal
+    koby_note: "Koby GPU issues reported, never fixed on .129"
+    
+  - name: gpu-monitor
+    path: gpu-monitor.prose.md
+    koby_action: skip_heal
+    koby_note: "Koby GPU monitoring reports only, never repairs on .129"
+    
+  - name: agent-health-check
+    path: agent-health-check.prose.md
+    koby_action: skip_heal
+    koby_note: "Koby agent health checks reported, never repairs on .129"
+
+# Scripts that should skip Koby
+koby_aware_scripts:
+  - name: agent-health-check.py
+    path: scripts/agent-health-check.py
+    koby_action: skip_heal
+    koby_note: "Script should only run diagnostics on Koby, not repairs"

disk-gc-threat-response.prose.md

@@ -1,4 +1,6 @@
 ---
+report_only_agents:
+  - koby  # ⛔ KOBY IS NEVER REPAIRED (Rule 17, 2026-08-17) — detect + report, never fix on .129
 kind: responsibility
 name: disk-gc-threat-response
 description: >
@@ -11,6 +13,7 @@ description: >
 id: 067NV8KJ03ZG71S44N41F31022
 version: 1.0.0
 ---
+---
 
 # Disk GC & Threat Response
 
@@ -302,4 +305,4 @@ one-off GPU builds. No automated post-migration cleanup was in place.
 |------|-----|------|--------|
 | docker-vm | 192.168.68.7 | 16 Docker containers, 4 stacks | ✅ reachable |
 
-> **Note:** CT 118 is now jdownloader (active on storepve). CT 119 (infisical-vault) added on minipve.\n> **Migrated:** CT 101 → .8, CT 103 → .110 (bare metal GPU).\n> **KVM VM:** CT 109 (docker-vm) is a KVM VM, not LXC — access via SSH .7.
+> **Note:** CT 118 is now jdownloader (active on storepve). CT 119 (infisical-vault) added on minipve.\n> **Migrated:** CT 101 → .8, CT 103 → .110 (bare metal GPU).\n> **KVM VM:** CT 109 (docker-vm) is a KVM VM, not LXC — access via SSH .7.

gpu-fleet.prose.md

@@ -8,13 +8,13 @@ description: >
   UPDATED 2026-07-15: Stable role-based aliases introduced: strix-moe,
   gpu-dense, gpu-light. These never change — only the underlying model does.
   Strix Halo: strix-moe → unsloth/Qwen3.6-35B-A3B-MTP (UD-Q4_K_M, 22GB).
-  RTX 5070: gemma-4-12b Q4_K_M → IQ4_NL + MTP draft (122 tok/s, 2x faster).
+  RTX 5070: gpu-vision Q4_K_M → IQ4_NL + MTP draft (122 tok/s, 2x faster).
   UPDATED 2026-07-17: Context reduced fleet-wide from 256K to 128K for stability.
   Strix Halo model swapped to qwen3.6-35B-udq4 (22GB, strix-moe alias).
   Instability observed near 100K at 256K (now all GPUs at 128K). 128K is the stable ceiling.
   For larger context needs → fall back to external providers (deepseek).
   VRAM headroom improved: RTX 3090 ~70%, RTX 5070 ~65%.
-  UPDATED 2026-07-27: gpu-dense swapped to SmartCode-Fable-5-CoT-Reasoning-QKVO-Qwen-3.6-27B-Distilled
+  UPDATED 2026-07-27: gpu-dense swapped to Qwen3.8-27B-Uncensored-Q4_K_M
   (UD-Q3_K_XL, ~14.7GB — Q4 was too large for 24GB VRAM with 128K KV cache). ~50% fewer thinking
   tokens via ThinkingCap finetune + Fable 5 CoT distillation for improved coding reasoning.
   VRAM ~22.4/24.6GB (91%).
@@ -75,7 +75,7 @@ triggers:
 │ RTX 3090 │  │ RTX 5070    │  │ Strix Halo│  │ GPU Monitor  │
 │ 24GB     │  │ 12GB        │  │ 64GB UMA  │  │ :9100        │
 │ 128K ctx │  │ 128K ctx    │  │ 128K ctx  │  │ Watchdog     │
-│ qwen3.6  │  │ gemma-4-12b │  │ qwen3.6   │  │ Prometheus   │
+│ qwen3.6  │  │ Qwen3.5-9B  │  │ qwen3.5   │  │ Prometheus   │
 │ 27B-code │  │ :8080       │  │ -35B-udq4 │  │ exporter     │
 │ :8080    │  │ :9400 (exp) │  │ :9400(exp)│  │ :9401        │
 │ :9400    │  └─────────────┘  └───────────┘  └──────────────┘
@@ -90,19 +90,19 @@ When a model is swapped on a GPU, ONLY the infrastructure layer changes — agen
 | Alias | GPU | Current Model | Will Route To |
 |-------|-----|---------------|---------------|
 | `strix-moe` | Strix Halo (.15) | qwen3.6-35B-udq4 | Whatever runs on Strix Halo |
-| `gpu-dense` | RTX 3090 (.8) | SmartCode-Fable-5-27B-UD-Q4_K_XL | Whatever runs on RTX 3090 |
-| `gpu-light` | RTX 5070 (.110) | gemma-4-12b | Whatever runs on RTX 5070 |
+| `gpu-dense` | RTX 3090 (.8) | Qwen3.8-27B-Uncensored-Q4_K_M | Whatever runs on RTX 3090 |
+| `gpu-light` | RTX 5070 (.110) | gpu-vision | Whatever runs on RTX 5070 |
 
-**Backward compatibility**: Old model-specific names (qwen3.6-27B-code, gemma-4-12b, qwen3.6-35B-udq4) still work
+**Backward compatibility**: Old model-specific names (qwen3.6-27B-code, gpu-vision, qwen3.5-9b-it) still work
 but are deprecated for agent configs. Only the stable aliases survive model swaps.
 
 ## Current Model Assignments (2026-07-15)
 
 | Model | GPU | Host | VRAM | Ctx | KV Cache | Parallel | Batch/Ubatch | Status |
 |-------|-----|------|------|-----|----------|----------|-------------|--------|
-| SmartCode-Fable-5-27B-UD-Q3_K_XL | RTX 3090 | .8 (llm-gpu) | ~22.4/24.6GB (91%) | **128K** | q4_0 | 1 | 2048/1024 | ✅ healthy |
-| gemma-4-12b | RTX 5070 | .110 (ocu-llm) | ~7.8/12.2GB (65%) | 128K | q4_0 | 2 | 2048/1024 | ✅ healthy |
-| qwen3.6-35B-udq4 | Strix Halo Vulkan | .15 (amdpve) | ~22GB/64GB | 128K | q4_0 | 1 | 4096/1024 | ✅ 65 tok/s |
+| Qwen3.8-27B-Uncensored-Q4_K_M | RTX 3090 | .8 (llm-gpu) | ~22.4/24.6GB (91%) | **128K** | q4_0 | 1 | 2048/1024 | ✅ healthy |
+| Qwen3.5-9B | RTX 5070 | .110 (ocu-llm) | ~6.2/12.2GB (51%) | 128K | Q5_K_M | 2 | 2048/1024 | ✅ healthy |
+| Carnice-Qwen3.6-MoE-35B-A3B | Strix Halo Vulkan | .15 (amdpve) | ~24.73GB/64GB | 128K | Q5_K_M | 1 | 4096/1024 | ✅ 65 tok/s |
 
 ## Routing Configuration (LiteLLM — July 2026)
 
@@ -110,9 +110,9 @@ but are deprecated for agent configs. Only the stable aliases survive model swap
 
 | Model | GPU | Weight | RPM Cap | Timeout |
 |-------|-----|--------|---------|---------|
-| SmartCode-Fable-5-27B-UD-Q3_K_XL | RTX 3090 (.8:8080) | **0.55** | 500 | **300s** |
-| qwen3.6-35B-udq4 | Strix Halo (.15:8080) | **0.30** | 60 | **300s** |
-| gemma-4-12b | RTX 5070 (.110:8080) | **0.15** | 200 | **120s** |
+| Qwen3.8-27B-Uncensored-Q4_K_M | RTX 3090 (.8:8080) | **0.55** | 500 | **300s** |
+| Carnice-Qwen3.6-MoE-35B-A3B | Strix Halo (.15:8080) | **0.30** | 60 | **300s** |
+| Qwen3.5-9B | RTX 5070 (.110:8080) | **0.15** | 200 | **120s** |
 
 Note: All syslog-auto entries route directly to GPUs with `api_key: not-needed`. The router (port 9000) is NOT in the inference path.
 
@@ -120,9 +120,9 @@ Note: All syslog-auto entries route directly to GPUs with `api_key: not-needed`.
 
 | Model | RPM Cap | Notes |
 |-------|---------|-------|
-| strix-moe (qwen3.6-35B-udq4) | 40 | Tight cap — prevents Strix overload |
-| SmartCode-Fable-5-27B-UD-Q3_K_XL | 500 | High cap — primary workhorse (replaces qwen3.6-27B-code) |
-| gemma-4-12b | 500 | High cap — IQ4_NL+MTP, 122 tok/s |
+| strix-moe (Carnice-Qwen3.6-MoE-35B-A3B) | 40 | Tight cap — prevents Strix overload |
+| Qwen3.8-27B-Uncensored-Q4_K_M | 500 | High cap — primary workhorse |
+| Qwen3.5-9B | 500 | High cap — multimodal vision endpoint |
 
 ### Stable Aliases (for agent configs — never change)
 
@@ -193,7 +193,7 @@ Show full fleet status: GPUs, models, VRAM, context windows, parallel slots, act
 3. Check LiteLLM: `curl http://192.168.68.116/health` (expect "I'm alive!")
 4. Check LiteLLM models: `curl -H "Authorization: Bearer $MASTER_KEY" http://192.168.68.116/v1/models`
 5. Check LiteLLM timeouts: `grep -n 'timeout:' /opt/inference-harness/litellm_config.yaml`
-   - gemma-4-12b: 120s, qwen3.6-27B-code: 300s, qwen3.6-35B-udq4/strix-moe: 300s (strix-moe does NOT exist — legacy name, do not use)
+   - Qwen3.5-9B: 120s, qwen3.6-27B-code: 300s, Carnice-Qwen3.6-MoE-35B-A3B/strix-moe: 300s (strix-moe alias retained, legacy name qwen3.6-35B-udq4 deprecated)
    - global request_timeout: 300s, nginx proxy_read_timeout: 600s
 6. Check AMD metrics: `curl http://192.168.68.15:9400/metrics` (Radeon 8060S, util%, VRAM, temp, power)
 7. Check port conflicts: verify only one llama-server on :8080 per host
@@ -250,10 +250,10 @@ If no SSH access, send Zulip DM via abiba-bot with vault update instructions.
 - **Router startup race**: Compose router.py doesn't call load_roster(). Reload thread sleeps 30s first.
   Fix: trigger roster reload via SSH after restart, or rebuild image with startup load_roster().
 - **LiteLLM /metrics**: Requires auth. Prometheus uses `/health/liveliness` as workaround.
-- **VRAM (2026-07-15)**: RTX 3090 at ~17/24.6GB (~70%) with **128K context** (reduced from 256K 2026-07-17). RTX 5070 at ~7.8/12.2GB (~65%) with 128K context + MTP. Strix Halo at ~7GB/64GB.
-- **RTX 3090 (2026-07-27)**: Swapped to SmartCode-Fable-5-27B-UD-Q3_K_XL (14.7GB). Q4 was too large for 24GB VRAM with 128K context + KV cache overhead. Q3 fits at ~22.4GB (91%). Uses standard llama.cpp build b9190 (turboquant b9150 incompatible with qwen3_5 arch). Config: `-c 131072 -ctk q4_0 -ctv q4_0 --flash-attn on --cont-batching`. Sampler: `--temp 0.9 --top-p 0.95 --top-k 60 --min-p 0.0 --repeat-penalty 1.0`. Service: `/home/llmuser/llama-fable-wrapper.sh`.
-- **RTX 5070 config (2026-07-15)**: Switched to IQ4_NL + MTP draft (Q8_0) at 128K context. Gen speed: 122 tok/s. VRAM: ~7.8/12.2GB (~65%). Service: `/home/llmuser/llama-wrapper.sh`. Config: `--model gemma-4-12b-it-IQ4_NL.gguf --spec-draft-model gemma-4-12b-it-Q8_0-MTP.gguf --spec-type draft-mtp --spec-draft-n-max 4 --ctx-size 131072`.
-- **LiteLLM timeout tuning (verified 2026-07-27)**: SmartCode-Fable-5-27B 300s, gemma-4-12b 120s, qwen3.6-27B-code 300s (legacy), qwen3.6-35B-udq4 300s, strix-moe 300s, syslog-auto routes all 300s. Nginx proxy_read_timeout: 600s. Global request_timeout: 300s.
+- **VRAM (2026-08-20)**: RTX 3090 at ~22.4/24.6GB (~91%) with **128K context** (reduced from 256K 2026-07-17). RTX 5070 at ~6.2/12.2GB (~51%) with 128K context (Qwen3.5-9B). Strix Halo at ~22GB/64GB.
+- **RTX 3090 (2026-07-27)**: Swapped to Qwen3.8-27B-Uncensored-Q4_K_M (14.7GB). Q4 was too large for 24GB VRAM with 128K context + KV cache overhead. Q3 fits at ~22.4GB (91%). Uses standard llama.cpp build b9190 (turboquant b9150 incompatible with qwen3_5 arch). Config: `-c 131072 -ctk q4_0 -ctv q4_0 --flash-attn on --cont-batching`. Sampler: `--temp 0.9 --top-p 0.95 --top-k 60 --min-p 0.0 --repeat-penalty 1.0`. Service: `/home/llmuser/llama-fable-wrapper.sh`.
+- **RTX 5070 config (2026-08-20)**: Switched to Qwen3.5-9B (Q5_K_M) at 128K context. Multimodal (image+text). Gen speed: ~145 tok/s (estimated). VRAM: ~6.2/12.2GB (~51%). Service: `/home/llmuser/llama-wrapper.sh`. Config: `--model Qwen3.5-9B-Q5_K_M.gguf --mmproj Qwen3.5-9B-mmproj-F16.gguf --ctx-size 131072`.
+- **LiteLLM timeout tuning (verified 2026-07-27)**: Qwen3.8-27B-Uncensored-Q4_K_M 300s, gpu-vision 120s, qwen3.6-27B-code 300s (legacy), qwen3.6-35B-udq4 300s, strix-moe 300s, syslog-auto routes all 300s. Nginx proxy_read_timeout: 600s. Global request_timeout: 300s.
 - **Strix Halo GPU**: Vulkan is the working backend (ROCm/HIP path abandoned — HSA runtime blocked on Debian 13). Build at `/root/llama.cpp/build-vk/`, commit `4fc4ec5` (2026-07-01), ggml 0.15.3 shared-lib arch. Mesa RADV 25.0.7, KHR_coopmat fast path active. ~70 tok/s gen, 532 tok/s prompt. Service: `strix-server.service` on port 8080, model: `qwen3.6-35B-udq4`, alias `strix-moe`, 128K context, flash-attn + q4 KV, multimodal (mmproj loaded).
 - **Port conflict detection (2026-07-05)**: All 3 GPU wrappers now detect ghost processes squatting port 8080 before starting. `.8` and `.110` use inline pre-start check in `llama-wrapper.sh`; `.15` uses `/usr/local/bin/port-cleanup.sh` ExecStartPre. Replaces the blanket `pkill -9 -x llama-server` on .15 which would kill ALL llama-server instances regardless of port. Ghost detection was the root cause of .8 crash-looping for 27+ restarts (stale pid 25836 squatting 8080 after OOM kill).
 - **Strix Halo thermal safeguard (2026-07-02)**: `strix-server.service` has `-n 8192` (hard generation cap per request). Without it, `--predict` defaults to -1 (infinity) — a runaway request from .123 (old Mumuni CT114 — now inside Abiba CT100 at .24) decoded 39,868 tokens over 24 min, pushing Tctl to 98°C (crit 89.8°C) and throttling 70→29 t/s. The cap bounds worst-case generation to ~5 min. Do NOT remove `-n` without a replacement ceiling. Sustained load hits ~84°C even at 92s; the APU is fanless/low-flow. Clients MUST also set `max_tokens`.
@@ -268,8 +268,8 @@ If no SSH access, send Zulip DM via abiba-bot with vault update instructions.
 
 | GPU | Model | Gen tok/s | Prompt tok/s | Baseline | Context |
 |-----|-------|-----------|--------------|----------|---------|
-| RTX 3090 (.8) | SmartCode-Fable-5-27B-UD-Q3_K_XL | **TBD** | — | — | **128K** |
-| RTX 5070 (.110) | gemma-4-12b (IQ4_NL+MTP) | **191** | — | — | **128K** |
+| RTX 3090 (.8) | Qwen3.8-27B-Uncensored-Q4_K_M | **TBD** | — | — | **128K** |
+| RTX 5070 (.110) | Qwen3.5-9B (Q5_K_M) | **145** | — | — | **128K** |
 | Strix Halo (.15) | qwen3.6-35B-udq4 | **65** | 140 | — | **128K** |
 
 Benchmarks from 2026-07-17. Strix Halo model: qwen3.6-35B-udq4. RTX 5070 MTP provides 2.7x speedup over pre-upgrade 70 tok/s.
@@ -287,7 +287,7 @@ History stored at `/root/data/toks-history.json` with 7-day rolling window.
 
 All agent configs MUST use stable role-based aliases, never model-specific names:
 - `compression.model: strix-moe` (NOT `qwen3.6-35B-udq4`)
-- `auxiliary.vision.model: gpu-light` (NOT `gemma-4-12b`)
+- `auxiliary.vision.model: gpu-light` (NOT `gpu-vision`)
 - `delegation.model: gpu-dense` (NOT `qwen3.6-27B-code`)
 - `auxiliary.web_extract.model: gpu-light`
 
@@ -296,7 +296,8 @@ When the underlying model is swapped, only the LiteLLM config changes — agent
 ### Context Windows
 - RTX 3090: **128K** (reduced from 256K 2026-07-17) | RTX 5070: **128K** (reduced from 256K) | Strix Halo: **128K**
 - **All agents**: 128K ceiling — stable margin. For >128K workloads, use external providers (deepseek)
-- Compression threshold 0.65: fires at ~85K (~43K headroom before 128K ceiling)
+- Compression threshold 0.60: fires at ~77K (~51K headroom before 128K ceiling)
+- **Pi agents (Abiba)**: `compaction.reserveTokens: 52739` (≈60% of 128K)
 - Mumuni compression model alias: `strix-moe` with 300s timeout
 
 ### Mumuni Agent Profile
@@ -313,7 +314,7 @@ Mumuni (CT100/abiba, 192.168.68.24) is the primary business assistant. This prof
 | `aux.web_extract.model` | `gpu-light` | Web extraction |
 | `delegation.model` | `gpu-dense` | Sub-agent reasoning (RTX 3090) |
 | `context.max_context_window` | 131072 (128K) | Reduced from 256K 2026-07-17 — stable 128K ceiling |
-| `compression.threshold` | 0.65 | Triggers at ~85K |
+| `compression.threshold` | 0.60 | Triggers at ~77K (~60% of 128K) — optimized for 128K context |
 | `compression.target_ratio` | 0.3 | Compresses to ~38K |
 | `compression.protect_last_n` | 40 | Preserves last 40 messages |
 | `memory.memory_char_limit` | 800 | Brief memory entries |

gpu-self-heal.prose.md

@@ -1,4 +1,6 @@
 ---
+report_only_agents:
+  - koby  # ⛔ KOBY IS NEVER REPAIRED (Rule 17, 2026-08-17) — detect + report, never fix on .129
 kind: responsibility
 name: gpu-self-heal
 description: >
@@ -16,6 +18,7 @@ depends_on:
   - gpu-monitor.prose.md (live data source on .24:9100)
   - gpu-fleet.prose.md (source of truth for topology, aliases, model assignments)
 ---
+---
 
 ## Maintains
 
@@ -38,6 +41,7 @@ depends_on:
 - On fix: verify with benchmark inference test before declaring resolved
 - Escalate: after 3 failed remediation attempts → Zulip #agent-hub alert
 
+---
 ---
 
 ## Current Fleet Baseline (2026-07-18)
@@ -45,13 +49,13 @@ depends_on:
 | Alias | GPU | Host | Model | VRAM | Ctx | tok/s | Role |
 |-------|-----|------|-------|------|-----|-------|------|
 | `gpu-dense` | RTX 3090 24GB | ct8 (.8:8080) | ThinkingCap Qwen3.6-27B Q4_K_M + MTP + vision | 21.6/24.6GB (88%) | 128K | 74.9 | Heavy reasoning, code gen |
-| `gpu-light` | RTX 5070 12GB | ct110 (.110:8080) | HauhauCS Gemma4-12B QAT Q4_K_M + MTP draft | 10.1/12.2GB (83%) | 128K | 169.6 | Vision, web extract, light tasks |
+| `gpu-light` | RTX 5070 12GB | ct110 (.110:8080) | Qwen3.5-9B (Q5_K_M) + mmproj-F16 | 6.2/12.2GB (51%) | 128K | ~145 | Vision (image+text), web extract, light tasks |
 | `strix-moe` | Strix Halo 64GB | ct15 (.15:8080) | qwen3.6-35B-udq4 | ~10/64GB (16%) | 128K | 62.9 | Compression, summarization, long docs |
 
 Key notes:
 - All models use direct GPU routing via LiteLLM (`api_key: not-needed`). Router (port 9000) is deprecated and NOT in the inference path.
 - Stable aliases (gpu-dense, gpu-light, strix-moe) from gpu-fleet are the canonical names for agent configs. Model-specific names still work but are deprecated.
-- RTX 5070 tok/s is 2.3x faster than RTX 3090 for its model — gpu-light is the fastest endpoint. Route vision/web/light work there first.
+- RTX 5070 tok/s is ~145 for Qwen3.5-9B — gpu-light is the fastest endpoint. Route vision/web/light work there first. NOTE: Qwen3.5-9B is multimodal (image+text), NOT text-only like gpu-vision was.
 - Strix Halo is 62.9 tok/s (89% of 70.5 baseline) — below optimal but stable. Check for competing workloads.
 - RTX 3090 VRAM at 88% — within role-appropriate range (role = heavy reasoning, needs the headroom).
 - RTX 5070 VRAM at 83% — role-appropriate for vision/web (smaller batch sizes).
@@ -156,7 +160,7 @@ Key notes:
 - **Detect**: GPU roles misaligned with hardware capabilities
 - **Target distribution**:
   - RTX 3090 (gpu-dense, 24GB, 74.9 tok/s) → Heavy reasoning, code gen, long conversations (slowest per-token but largest context capacity). Weight: 0.55 (LiteLLM).
-  - RTX 5070 (gpu-light, 12GB, 169.6 tok/s) → Vision/image, web search, lightweight tasks (2.3x faster than 3090 per token). Weight: 0.15 (LiteLLM).
+  - RTX 5070 (gpu-light, 12GB, ~145 tok/s) → Vision (image+text), web search, lightweight tasks. Weight: 0.15 (LiteLLM).
   - Strix Halo (strix-moe, 64GB, 62.9 tok/s) → Context compression, summarization, long docs (MoE model). Weight: 0.30 (LiteLLM).
 - **Note**: RTX 5070 is the fastest endpoint per token. Route high-volume, low-complexity work there first.
 - **Fix**:
@@ -166,6 +170,7 @@ Key notes:
 - **Verify**: Each GPU's request pattern matches its designated role within 24h
 - **Escalate**: If role mismatch persists >48h → agent alias audit needed
 
+---
 ---
 
 ## Execution
@@ -247,6 +252,7 @@ call update-gpu-health
 }
 ```
 
+---
 ---
 
 ## Reporting
@@ -263,6 +269,7 @@ Pushed to `SyslogSolution/health-logs/gpu/{run_id}.json` — versioned, searchab
 - Per-GPU tok/s trend over 7 days
 - Regression alerts if any GPU degrades >10% week-over-week
 
+---
 ---
 
 ## Design Decisions (Verified 2026-07-12, Reaffirmed 2026-07-18)

hermes-agent-baseline.prose.md

@@ -81,7 +81,7 @@ custom_providers:
 auxiliary:
   vision:
     provider: harness
-    model: gemma-4-12b              # or syslog-auto
+    model: gpu-vision              # or syslog-auto
     base_url: http://192.168.68.116/v1
     api_key_env: LITELLM_API_KEY
     api_key: <value from: infisical secrets get LITELLM_API_KEY --project=agents --env=production>   # ← MANDATORY workaround
@@ -96,7 +96,7 @@ auxiliary:
     threshold: 0.65
     target_ratio: 0.3
     provider: harness
-    model: syslog-auto               # or gemma-4-12b
+    model: syslog-auto               # or gpu-vision
     base_url: http://192.168.68.116/v1
     api_key_env: LITELLM_API_KEY
     api_key: <value from: infisical secrets get LITELLM_API_KEY --project=agents --env=production>   # ← MANDATORY workaround
@@ -169,11 +169,15 @@ pct-run <CT> grep -A8 "vision:" /root/.hermes/config.yaml | grep api_key
 # Must show both api_key: sk-... and api_key_env: LITELLM_API_KEY
 ```
 
-### For Koby (CT 111 / tdunna)
+### For Koby (CT 111 / tdunna) — **REPORT-ONLY MODE**
 
 Koby runs Hermes on CT 111 (tdunna). Config files at `/root/.hermes/config.yaml`.
 Same Hermes pattern as Tanko/Mumuni/Koonimo — see config sections above.
 
+**⛔ KOBY IS NEVER REPAIRED (2026-08-17, Captain)**: Diagnostic only — detect and report, never fix on .129.
+No heal step, no restart, no key rotation, no config edit, no memory rewrite, no disk GC, no service touch, no process kill — ever.
+If a health check shows Koby degraded, **DO NOT** execute any repair action. Instead, report to Zulip and let Theo fix it.
+
 **LiteLLM key**: alias `koby` in LiteLLM DB, injected via `infisical run --` wrapper.
 
 ### For pi Agents (Abiba)
@@ -196,7 +200,7 @@ Key is injected via `infisical run --` wrapper at PM2 startup:
         { "id": "gpu-dense" },
         { "id": "gpu-light" },
         { "id": "qwen3.6-27B-code" },
-        { "id": "gemma-4-12b" }
+        { "id": "gpu-vision" }
       ]
     }
   }

hermes-config-template.prose.md

@@ -101,7 +101,7 @@ model:
   api_key_env: LITELLM_API_KEY    # Injected via infisical run -- wrapper
   max_tokens: 4096                # ⚠️ CRITICAL: Prevents unbounded generation
   context_length: 131072          # For syslog-auto (all GPUs at 128K for stability).
-                                  # Set 65536 if using gemma-4-12b directly (tight VRAM).
+                                  # Set 65536 if using gpu-vision directly (tight VRAM).
 
 fallback_providers:
   provider: deepseek
@@ -142,18 +142,18 @@ compression:
 
 # ─── Auxiliary Tasks (CONSISTENCY RULE) ───
 # All auxiliary services MUST use identical model, base_url, and api_key_env:
-#   model: gpu-light                 # stable alias (NOT raw "gemma-4-12b")
+#   model: gpu-light                 # stable alias (NOT raw "gpu-vision")
 #   base_url: http://192.168.68.116/v1
 #   api_key_env: LITELLM_API_KEY
 # Do NOT use syslog-auto for auxiliary tasks — it routes to the primary GPU.
 # gpu-light = RTX 5070 (12B), freeing the Strix Halo for agent reasoning.
 # Heavy aux (delegation, x_search) use gpu-dense (RTX 3090) instead.
-# NEVER use raw model names (gemma-4-12b, qwen3.6-27B-code, qwen3.6-35B-udq4)
+# NEVER use raw model names (gpu-vision, qwen3.6-27B-code, qwen3.6-35B-udq4)
 # in agent configs — use the stable aliases so model swaps don't break agents.
 auxiliary:
   vision:
     provider: harness
-    model: gpu-light            # stable alias for RTX 5070 (was raw gemma-4-12b)
+    model: gpu-light            # stable alias for RTX 5070 (was raw gpu-vision)
     base_url: http://192.168.68.116/v1
     api_key_env: LITELLM_API_KEY
     timeout: 60
@@ -248,10 +248,10 @@ The following MUST be identical across ALL profiles:
 - For agents needing longer outputs: raise to 8192, but never omit
 
 ### Rule 7: Auxiliary Model Consistency (UPDATED 2026-07-16)
-- Vision and web_extract use `gemma-4-12b` (RTX 5070 — 12GB, vision-optimized)
+- Vision and web_extract use `gpu-vision` (RTX 5070 — 12GB, vision-optimized)
 - Compression uses `strix-moe` (stable alias for Strix Halo — 64GB, 128K ctx, compression-optimized)
 - **`strix-moe` is the only valid compression model name** — LiteLLM does NOT serve `ornith-1.0-35b`
-  (it serves `strix-moe`, `qwen3.6-35B-udq4`, `gpu-dense`, `gpu-light`, `syslog-auto`, `gemma-4-12b`, `qwen3.6-27B-code`). Old configs with `ornith-1.0-35b` cause 403/model-not-found on compression calls.
+  (it serves `strix-moe`, `qwen3.6-35B-udq4`, `gpu-dense`, `gpu-light`, `syslog-auto`, `gpu-vision`, `qwen3.6-27B-code`). Old configs with `ornith-1.0-35b` cause 403/model-not-found on compression calls.
 - **OPERATIONAL DECISION (2026-07-23): Use `syslog-auto` for compression across all agents.**
   The `syslog-auto` alias routes to the Strix Halo, but uses the weighted pool instead of pinning
   to `strix-moe` directly. This prevents sustained Strix Halo thermal load because the pool can
@@ -269,11 +269,11 @@ The following MUST be identical across ALL profiles:
 
 ### Rule 8: GPU Workload Distribution (UPDATED 2026-07-16)
 - **RTX 3090 (24GB, 128K ctx, qwen3.6-27B-code)**: Heavy reasoning, code gen, long conversations
-- **RTX 5070 (12GB, 128K ctx, gemma-4-12b)**: Vision, web search, quick tasks, web_extract (IQ4_NL+MTP, ~65% VRAM at 128K)
+- **RTX 5070 (12GB, 128K ctx, gpu-vision)**: Vision, web search, quick tasks, web_extract (IQ4_NL+MTP, ~65% VRAM at 128K)
 - **Strix Halo (64GB, 128K ctx, syslog-auto)**: Context compression, summarization, long docs
 - Agent profiles MUST route auxiliary tasks to the correct GPU:
-  - `auxiliary.vision.model: gemma-4-12b` (RTX 5070)
-  - `auxiliary.web_extract.model: gemma-4-12b` (RTX 5070)
+  - `auxiliary.vision.model: gpu-vision` (RTX 5070)
+  - `auxiliary.web_extract.model: gpu-vision` (RTX 5070)
   - `auxiliary.compression.model: syslog-auto` (Strix Halo)
 - Default model (`model.default`) and custom_provider remain `syslog-auto` for auto-routing
 - For 128K context window: `threshold: 0.65` (fires at ~85K tokens)
@@ -293,7 +293,7 @@ The following MUST be identical across ALL profiles:
 - **Hermes agents**: `model.default: syslog-auto`, `custom_providers[0].model: syslog-auto`
 - **pi agents**: `defaultModel: syslog-auto` in `settings.json`, first model in `models.json`
 - `syslog-auto` is the LiteLLM routing model — it load-balances between strix-moe
-  and qwen3.6-27B-code, with gemma-4-12b as fallback. Using it protects against:
+  and qwen3.6-27B-code, with gpu-vision as fallback. Using it protects against:
   - Model name typos that cause 403 errors and silent worker failures
   - Single GPU downtime (routing falls back automatically)
   - Key/model authorization mismatches
@@ -338,6 +338,17 @@ curl -s -o /dev/null -w 'key_health: %{http_code}\n' -H "Authorization: Bearer $
 
 ### Rule 13: API Key Injection — Two Patterns (UPDATED 2026-07-16, WAL #1300)
 
+### Rule 14: Hermes Context Detection Uses `max_model_tokens`, NOT `max_input_tokens`
+
+**CRITICAL**: Hermes context detection reads `max_model_tokens` (128K), NOT `max_input_tokens` (64K cap).
+
+- **Abiba and Hermes agents**: `max_model_tokens: 131072` (128K) — unlimited context
+- **Crewmates (ops, tune, verify, auth-keys, build)**: `max_input_tokens: 64000` (64K) — capped
+- If you see `max_input_tokens: 64000` in an Abiba/Hermes config, that's a mistake
+- Using `max_input_tokens` for Hermes agents causes premature context loss
+- Check: `grep -n 'max_model_tokens\|max_input_tokens' ~/.hermes/config.yaml`
+- Expected output: `max_model_tokens: 131072` (not max_input_tokens)
+
 Agents inject `LITELLM_API_KEY` via ONE of two mechanisms. Both are valid; the contract
 requirement is that the key is a **valid LiteLLM virtual key** (HTTP 200 on /v1/models).
 
@@ -405,3 +416,14 @@ curl -s -o /dev/null -w '%{http_code}' -H "Authorization: Bearer $K" http://192.
 5. **Set model choice** — Per agent's workload
 6. **Verify** — curl all shared endpoints, test the model with the new key
 7. **Report** — What was changed, preserved, custom
+
+### Rule 17: Koby Report-Only (URGENT — 2026-08-17, Captain)
+
+**Koby (CT 111, .129, tdunna/Theo) is report-only, full stop.**
+No heal step, no restart, no key rotation, no config edit, no memory rewrite,
+no disk GC, no service touch, no process kill — **ever**.
+Do NOT "repair" Koby under any circumstance, even if a health check shows it degraded.
+Detect + report to Zulip (Theo fixes it).
+
+This supersedes every other instruction in this file.
+If a rule would trigger a Koby repair action, SKIP it and emit a diagnostic alert.

hermes-key-enforcement.prose.md

@@ -177,7 +177,7 @@ The agent picks up the new key via `infisical run --` at gateway startup.
 # In litellm_config.yaml — ensures all future keys inherit these defaults:
 litellm_settings:
   default_key_generate_params:
-    models: ["syslog-auto", "qwen3.6-27B-code", "gemma-4-12b"]
+    models: ["syslog-auto", "qwen3.6-27B-code", "gpu-vision"]
     duration: null        # ← permanent
     max_budget: 100
     metadata:
@@ -285,13 +285,13 @@ auxiliary:
     api_key: sk-<agent-key-from-vault>           # ← workaround (get via: infisical secrets get LITELLM_API_KEY --project=agents --env=production --plain)
     api_key_env: LITELLM_API_KEY
     base_url: http://192.168.68.116/v1
-    model: gemma-4-12b
+    model: gpu-vision
     provider: harness
   compression:
     api_key: sk-<agent-key-from-vault>           # ← workaround (same as above)
     api_key_env: LITELLM_API_KEY
     base_url: http://192.168.68.116/v1
-    model: gemma-4-12b
+    model: gpu-vision
     provider: harness
 ```
 

hermes-zulip-restore.prose.md

@@ -1,4 +1,6 @@
 ---
+report_only_agents:
+  - koby  # ⛔ KOBY IS NEVER REPAIRED (Rule 17, 2026-08-17) — detect + report, never fix on .129
 kind: function
 name: hermes-zulip-restore
 description: >
@@ -12,6 +14,7 @@ version: 1.0.0
 status: active
 runtime_contract: 2
 ---
+---
 
 # Hermes Zulip Restore — Bring Any Agent Back to Good State
 
@@ -184,6 +187,7 @@ https://git.sysloggh.net/SyslogSolution/zulip-platform-plugins/src/branch/feat/z
 Commit `55ca15d` — `fix(zulip): add _strip_html for slash command matching`
 Pull request #33 is the primary integration branch.
 
+---
 ---
 
 **Last verified good state**: 2026-07-08 — Mumuni, Tanko, Koby all connected with `_strip_html` applied.

inference-optimization.prose.md

@@ -99,5 +99,5 @@ call enable-prompt-caching
 
 call verify-latency
   host: 192.168.68.116
-  models: [syslog-auto, qwen3.6-27B-code, gemma-4-12b]
+  models: [syslog-auto, qwen3.6-27B-code, gpu-vision]
 ```

infrastructure-monitoring.prose.md

@@ -139,6 +139,81 @@ GPU .8 (RTX 3090)    GPU .110 (RTX 5070)   GPU .15 (Strix Halo)
 4. Verify LiteLLM metrics flowing to Prometheus
 5. ~~Update nginx to proxy `/monitoring/` → Grafana~~ (NOT recommended — nginx sub-path was tried for /grafana/ and reverted per proxmox-monitor; direct :3001 access is the standard)
 
+## Execution
+
+### check-health
+
+**RUN LIVE, NEVER ECHO — every dispatch must execute the probes below with real tool calls; never repeat a prior report unless a live probe fails.**
+
+```bash
+# Zulip API health (POST ping)
+source /etc/litellm-monitor.env
+ZULIP_USER="abiba-bot@chat.sysloggh.net"
+curl -s -o /dev/null -w '%{http_code}' -X POST https://chat.sysloggh.net/api/v1/messages -u "${ZULIP_USER}:${ZULIP_BOT_KEY}"
+# Expected: 200 (HTTP 000 = unreachable/cache)
+
+# PM2 process health
+pm2 jlist
+# Expected: 5/5 online (abiba-telegram, abiba-zulip, zulip-watchdog, gitea-runner, spoton-service)
+
+# GPU exporters (may be down per DEPLOYMENT STATUS)
+curl -s http://192.168.68.8:9400/metrics && echo " - OK" || echo " - FAIL"
+curl -s http://192.168.68.110:9400/metrics && echo " - OK" || echo " - FAIL"
+curl -s http://192.168.68.15:9400/metrics && echo " - OK" || echo " - FAIL"
+
+# Prometheus targets
+curl -s http://192.168.68.116:9090/api/v1/targets | jq '.data.activeTargets'
+# Expected: All targets UP (may show some down if exporters not deployed)
+
+# Grafana health
+curl -s http://192.168.68.116:3001/api/health | jq '{status, version}'
+# Expected: {"status":"ok","version":"..."}
+
+# LiteLLM metrics
+curl -s http://192.168.68.116:4001/metrics | head -20
+# Expected: Prometheus-formatted metrics output
+```
+
+**Report format**: Summarize actual results from each probe. If any probe returns non-200 or empty output, flag as alert.
+
+### Phase 1: GPU Exporters
+
+**NVIDIA (.8 and .110)**:
+1. Download `nvidia_gpu_exporter` binary
+2. Create systemd service `nvidia-gpu-exporter.service`
+3. Start and enable
+
+**AMD (.15)**:
+1. Create Python exporter script at `/opt/amdgpu-exporter/exporter.py`
+2. Parses `amdgpu_top --json -d 1000` output
+3. Exposes key metrics at `:9400/metrics` via Python http.server
+4. Create systemd service
+5. Start and enable
+
+### Phase 2: Prometheus
+
+1. Create `/opt/monitoring/` directory on CT 116
+2. Write `prometheus.yml` with scrape configs for all targets
+3. Add to docker-compose (or separate compose file)
+4. Start container
+
+### Phase 3: Grafana
+
+1. Create `/opt/monitoring/grafana/` directories
+2. Provision Prometheus datasource
+3. Provision GPU fleet dashboard JSON
+4. Provision LiteLLM dashboard JSON
+5. Add to docker-compose
+6. Start container
+
+### Phase 4: Verification
+
+1. Verify all 3 GPU exporters return 200 at :9400/metrics
+2. Verify Prometheus targets all UP at :9090/targets
+3. Verify Grafana accessible at :3001 with dashboards
+4. Verify LiteLLM metrics flowing to Prometheus
+5. ~~Update nginx to proxy `/monitoring/` → Grafana~~ (NOT recommended — nginx sub-path was tried for /grafana/ and reverted per proxmox-monitor; direct :3001 access is the standard)
+
 ## Verification Commands
 
 ```bash

litellm-api-keys.prose.md

@@ -68,7 +68,7 @@ description: >
    - Generate new key with key_alias: "{agent_name}" (e.g., "tanko" — bare name, no date)
    - Set metadata: { "agent": "{agent_name}", "purpose": "agent-inference" }
    - Duration is null (permanent) — inherited from litellm default_key_generate_params
-   - Set models: ["syslog-auto", "qwen3.6-27B-code", "gemma-4-12b", "strix-moe", "gpu-dense", "gpu-light", "qwen3.6-35B-udq4"]
+   - Set models: ["syslog-auto", "qwen3.6-27B-code", "gpu-vision", "strix-moe", "gpu-dense", "gpu-light", "qwen3.6-35B-udq4"]
    - Note: `ornith-1.0-35b` is NOT a valid LiteLLM model name (use `strix-moe`, the stable alias). qwen3.6-35B-A3B removed from fleet (was never deployed).
    - Return the new key
 5. **If action == "rotate"**:
@@ -270,7 +270,7 @@ reads use per-agent identities. This eliminates the single shared token risk.
 | 2026-07-16 | vault | cleanup | 4 stale secrets deprecated. 5 personal creds flagged. |
 | 2026-07-16 | koonimo | add-zulip | Added KOONIMO_ZULIP_API_KEY to vault. Wrapper injects ZULIP_API_KEY + ZULIP_EMAIL. 3 platforms. |
 | 2026-07-16 | tanko | migrate | Migrated from hardcoded config.yaml to infisical-gateway.sh + st.353699cd. NOTE: st.353699cd later deleted — reverted to st.8e848433 on 2026-07-17. |
-| 2026-07-16 | mumuni | rotate | Old key malformed (sk-_SWAl_Vu_, 47 chars, not LiteLLM format) → 401. Deleted old `mumuni` key (token 15cbca18…), generated fresh (alias `mumuni`, 7 models: syslog-auto, qwen3.6-27B-code, gemma-4-12b, strix-moe, gpu-dense, gpu-light, qwen3.6-35B-udq4). New key sk-OzuWsoX2… written to /root/.hermes/.env (Rule 3/13 fallback). Vault sync PENDING (needs machine identity). WAL #1300. |
+| 2026-07-16 | mumuni | rotate | Old key malformed (sk-_SWAl_Vu_, 47 chars, not LiteLLM format) → 401. Deleted old `mumuni` key (token 15cbca18…), generated fresh (alias `mumuni`, 7 models: syslog-auto, qwen3.6-27B-code, gpu-vision, strix-moe, gpu-dense, gpu-light, qwen3.6-35B-udq4). New key sk-OzuWsoX2… written to /root/.hermes/.env (Rule 3/13 fallback). Vault sync PENDING (needs machine identity). WAL #1300. |
 | 2026-07-16 | koby | rotate | Old key sk-6sbCNjz (401, stale in /etc/environment). Deleted old `koby` key, generated fresh (alias `koby`). New key sk-BqRRMboTI… in systemd drop-in `hermes-gateway.service.d/litellm-key.conf` + /etc/environment. Created `hermes-gateway.service` unit (was missing — gateway wasn't persistent) with `--replace`. Verified HTTP 200, Telegram connected. |
 | 2026-07-16 | baggy (koonimo) | rotate | Old key sk-krnw_zGB (401, hardcoded in systemd drop-in). Deleted old `baggy` key, generated fresh (alias `baggy`, metadata agent=koonimo). New key sk-OEK7z26n6E… in drop-in `hermes-gateway.service.d/litellm-key.conf`. CT113 IP changed .113→.114. Verified HTTP 200, Zulip connected. |
 

litellm-health.prose.md

@@ -73,15 +73,15 @@ Request → nginx:80 → LiteLLM:4000 → GPU(llama-server, parallel 2)
 | Host | IP | Hardware | Models Served | Engine | Context | Parallel |
 |------|-----|----------|---------------|--------|---------|----------|
 | llm-gpu | 192.168.68.8 | NVIDIA RTX 3090 (24 GB) | qwen3.6-27B-code | llama-server systemd | **128K** | 2 |
-| ocu-llm | 192.168.68.110 | NVIDIA RTX 5070 (12 GB) | gemma-4-12b | llama-server systemd | **128K** | 2 |
+| ocu-llm | 192.168.68.110 | NVIDIA RTX 5070 (12 GB) | gpu-vision | llama-server systemd | **128K** | 2 |
 | amdpve | 192.168.68.15 | AMD Strix Halo 64GB UMA | qwen3.6-35B-udq4 (LiteLLM alias: strix-moe) | llama-server systemd (Vulkan) | 128K | 2 |
 
 ## Model Fallback Chains (LiteLLM)
 
 | Primary | Timeout | Fallback | Timeout |
 |---------|---------|----------|---------|
-| qwen3.6-27B-code | 300s | gemma-4-12b | 120s |
-| gemma-4-12b | 120s | qwen3.6-27B-code | 300s |
+| qwen3.6-27B-code | 300s | gpu-vision | 120s |
+| gpu-vision | 120s | qwen3.6-27B-code | 300s |
 | qwen3.6-35B-udq4 / strix-moe | 300s | qwen → gemma | — |
 | syslog-auto (balanced) | 300s | qwen → gemma | — |
 
@@ -127,7 +127,7 @@ Request → nginx:80 → LiteLLM:4000 → GPU(llama-server, parallel 2)
    - Check alerts array for active warnings/critical
 
 7. **Check model inference via LiteLLM** — Test each model:
-   - POST /v1/chat/completions model=gemma-4-12b → expect 200
+   - POST /v1/chat/completions model=gpu-vision → expect 200
    - POST /v1/chat/completions model=qwen3.6-27B-code → expect 200
    - POST /v1/chat/completions model=strix-moe → expect 200
    - Use master key for auth

litellm-self-heal.prose.md

@@ -1,4 +1,6 @@
 ---
+report_only_agents:
+  - koby  # ⛔ KOBY IS NEVER REPAIRED (Rule 17, 2026-08-17) — detect + report, never fix on .129
 kind: responsibility
 name: litellm-self-heal
 status: deployed
@@ -22,6 +24,7 @@ description: >
   inference, and agent keys. Applies remediation rules for common failures.
   Reports every action via Zulip DM and Gitea (SyslogSolution/health-logs).
 ---
+---
 
 # LiteLLM Operations — Health Check + Self-Heal
 
@@ -58,26 +61,34 @@ Request → nginx:80 → LiteLLM:4000 → GPU(llama-server, parallel 2)
 | Host | IP | Hardware | Models Served | Engine | Context | Parallel |
 |------|-----|----------|---------------|--------|---------|----------|
 | llm-gpu | 192.168.68.8 | NVIDIA RTX 3090 (24 GB) | qwen3.6-27B-code | llama-server systemd (`/home/llmuser/llama-wrapper.sh`, `-c 131072 --parallel 2 --ngl 99`) | **128K** | 2 |
-| ocu-llm | 192.168.68.110 | NVIDIA RTX 5070 (12 GB) | gemma-4-12b | llama-server systemd (`/home/llmuser/llama-wrapper.sh`, `--ctx-size 131072 --parallel 2`, IQ4_NL + MTP draft) | **128K** | 2 |
+| ocu-llm | 192.168.68.110 | NVIDIA RTX 5070 (12 GB) | gpu-vision | llama-server systemd (`/home/llmuser/llama-wrapper.sh`, `--ctx-size 131072 --parallel 2`, IQ4_NL + MTP draft) | **128K** | 2 |
 | amdpve | 192.168.68.15 | AMD Strix Halo 64GB UMA | qwen3.6-35B-udq4 (LiteLLM alias: `strix-moe`) | llama-server systemd (Vulkan) | 128K | 2 |
 
 > Verified on ground 2026-07-16 via `curl /v1/models` on each host + `llama-wrapper.sh`. The AMD host's underlying model is `qwen3.6-35B-udq4`; LiteLLM exposes it under two `model_name`s: `qwen3.6-35B-udq4` and `strix-moe` (rpm 40). The legacy name `ornith-1.0-35b` does NOT exist in LiteLLM and must not be referenced.
 
 ## LiteLLM Model Surface (ground truth — `/opt/inference-harness/litellm_config.yaml` on CT 116)
 
-`model_name`s served: `qwen3.6-27B-code`, `gemma-4-12b`, `qwen3.6-35B-udq4`, `strix-moe`, `gpu-dense`, `gpu-light`, `syslog-auto`.
+`model_name`s served: `qwen3.6-27B-code`, `gpu-vision`, `qwen3.6-35B-udq4`, `strix-moe`, `gpu-dense`, `gpu-light`, `syslog-auto`, `crew-auto` (new 2026-08-20).
 
-- `syslog-auto` is a weighted router model: qwen3.6-27B-code (0.55, rpm 500) + qwen3.6-35B-udq4 (0.30, rpm 60) + gemma-4-12b (0.15, rpm 200).
-- `gpu-dense` / `gpu-light` are high-rpm aliases (rpm 500) onto qwen3.6-27B-code / gemma-4-12b respectively.
-- Key scoping: agent keys are restricted to `['syslog-auto','qwen3.6-27B-code','gemma-4-12b','strix-moe','gpu-dense','gpu-light']`. As of 2026-07-16 the `baggy`/`koby`/`mumuni`/`abiba-pi` keys ALSO include `qwen3.6-35B-udq4`; `abiba-pi` additionally includes `deepseek-v4-pro` (cloud fallback). `kagenz0`/`koonimo`/`pi-agents-unified` have the standard 6 only. Agents should still use the stable alias `strix-moe` (not the raw `qwen3.6-35B-udq4`) so model swaps don't break them.
+### Context Cap Split (2026-08-20)
+
+- **Abiba (firstmate)**: 128K uncapped — unlimited context for primary workloads
+- **Hermes agents** (mumuni, tanko, koby, koonimo): 128K uncapped
+- **Crewmates** (ops, tune, verify, auth-keys, build): 64K capped — alias `crew-auto` enforces 64K limit
+
+Preferred implementation: uncap shared pool, add capped alias for crew-only.
+
+- `syslog-auto` is a weighted router model: qwen3.6-27B-code (0.55, rpm 500) + qwen3.6-35B-udq4 (0.30, rpm 60) + gpu-vision (0.15, rpm 200).
+- `gpu-dense` / `gpu-light` are high-rpm aliases (rpm 500) onto qwen3.6-27B-code / gpu-vision respectively.
+- Key scoping: agent keys are restricted to `['syslog-auto','qwen3.6-27B-code','gpu-vision','strix-moe','gpu-dense','gpu-light']`. As of 2026-07-16 the `baggy`/`koby`/`mumuni`/`abiba-pi` keys ALSO include `qwen3.6-35B-udq4`; `abiba-pi` additionally includes `deepseek-v4-pro` (cloud fallback). `kagenz0`/`koonimo`/`pi-agents-unified` have the standard 6 only. Agents should still use the stable alias `strix-moe` (not the raw `qwen3.6-35B-udq4`) so model swaps don't break them.
 - **NEVER use `litellm_proxy_master_key` (the `sk-litellm-...` master key) for inference.** It is for admin endpoints only (`/key/list`, `/key/generate`, `/key/info`). All inference — agent traffic, health-check model tests, monitor scripts — uses agent-specific keys. The health-check script's model tests use a dedicated `monitor` agent key stored at `/etc/litellm-monitor.env` on CT 116 (root-only, `chmod 600`); `/key/list` is the only call that legitimately uses `$MASTER_KEY`.
 
 ## Model Fallback Chains (LiteLLM)
 
 | Primary | Timeout | Fallback | Timeout |
 |---------|---------|----------|---------|
-| qwen3.6-27B-code | 300s | gemma-4-12b | 120s |
-| gemma-4-12b | 120s | qwen3.6-27B-code | 300s |
+| qwen3.6-27B-code | 300s | gpu-vision | 120s |
+| gpu-vision | 120s | qwen3.6-27B-code | 300s |
 | qwen3.6-35B-udq4 / strix-moe | 300s | qwen → gemma | — |
 | syslog-auto (balanced) | 300s | qwen → gemma | — |
 
@@ -120,6 +131,7 @@ Request → nginx:80 → LiteLLM:4000 → GPU(llama-server, parallel 2)
 - Also wakes on user request
 - On failure: re-check after 30s, escalate after 3 consecutive failures
 
+---
 ---
 
 ## Health Check
@@ -145,7 +157,7 @@ Run this first on every cycle. Results feed into remediation rules below.
 - Check alerts array for active warnings/critical
 
 ### 5. Check model inference via LiteLLM — test each model
-- POST /v1/chat/completions model=gemma-4-12b → expect 200
+- POST /v1/chat/completions model=gpu-vision → expect 200
 - POST /v1/chat/completions model=qwen3.6-27B-code → expect 200
 - POST /v1/chat/completions model=strix-moe → expect 200
 - Use master key for auth
@@ -162,6 +174,7 @@ Determine overall_status from individual check results:
 - "degraded" — 1-2 non-critical checks fail
 - "down" — critical checks fail
 
+---
 ---
 
 ## Remediation Rules
@@ -205,6 +218,7 @@ Escalate → if SSH access unavailable, send Zulip DM
 Router no longer in path so Redis active counters are unused. Rule retained
 for reference but inactive. If Redis issues occur, check harness-redis container.
 
+---
 ---
 
 ## Reporting
@@ -227,6 +241,7 @@ top actions, uptime.
 If a fix requires another agent (e.g., Authentik restart), relay sent
 to responsible agent with full context.
 
+---
 ---
 
 ## Execution

memory-audit-maintenance.prose.md

@@ -1,9 +1,12 @@
 ---
+report_only_agents:
+  - koby  # ⛔ KOBY IS NEVER REPAIRED (Rule 17, 2026-08-17) — detect + report, never fix on .129
 name: memory-audit-maintenance
 kind: responsibility
 description: Shared memory audit and maintenance contract for all Hermes agents (Mumuni, Tanko, Koby, Koonimo). Each agent runs it against its own isolated memory files — no cross-agent access, no shared state. Detects staleness, enforces writer registry, and rotates canary tokens.
 id: 067NC4KG01RG50R40M30E20918
 ---
+---
 
 ### Goal
 
@@ -343,4 +346,4 @@ return {
 
 ### Per-Agent Notes
 
-Each Hermes agent (Mumuni, Tanko, Tdunna, Baggy) runs this contract against its own `~/.hermes/memories/` directory. The contract is identical across agents, but all data is fully isolated: separate ledgers, separate writer registries, separate canaries. If a new agent is added to the roster, it must be listed in `### Scope` above and given its own isolated memory directory.
+Each Hermes agent (Mumuni, Tanko, Tdunna, Baggy) runs this contract against its own `~/.hermes/memories/` directory. The contract is identical across agents, but all data is fully isolated: separate ledgers, separate writer registries, separate canaries. If a new agent is added to the roster, it must be listed in `### Scope` above and given its own isolated memory directory.

memory-fixer.prose.md

@@ -1,4 +1,6 @@
 ---
+report_only_agents:
+  - koby  # ⛔ KOBY IS NEVER REPAIRED (Rule 17, 2026-08-17) — detect + report, never fix on .129
 kind: pattern
 name: memory-fixer
 description: >
@@ -6,6 +8,7 @@ description: >
   Escalate anything that needs Kwame's input.
 version: 1.1.0
 ---
+---
 
 # Memory Fixer
 

pm2-self-heal.prose.md

@@ -2,21 +2,26 @@
 kind: responsibility
 name: pm2-self-heal
 description: >
-  Monitors critical PM2 processes (abiba-zulip, abiba-telegram) and
+  Monitors critical PM2 processes on CT100/.24 (the container this agent runs in) and
   auto-restarts any that are stopped or errored. Logs every action to
   the knowledge graph and alerts the owner via Zulip DM on failures.
-  CRITICAL: Never restart abiba-zulip — it runs this contract.
+  Abiba-zulip is the live Zulip bridge and may be restarted; alert owner on failure.
+  **Execution host: CT100/.24 local pm2 — NOT CT116, NOT via SSH** (CT116 runs systemd+docker, no PM2).
+  **Process scope (5/5, 2026-08-09 ruling, ecosystem authoritative):**
+  abiba-zulip, abiba-telegram, gitea-runner, spoton-service, zulip-watchdog
+report_only_agents:
+  - koby  # ⛔ KOBY IS NEVER REPAIRED (Rule 17, 2026-08-17) — detect + report, never fix on .129
 ---
 
 ## Maintains
 
 - abiba-telegram: { status: "online", uptime: string, restarts: number }
-- gpu-watchdog: { status: "online", uptime: string, restarts: number }
-- gpu-monitor: { status: "online", uptime: string, restarts: number }
+- abiba-zulip: { status: "online", uptime: string, restarts: number }
+- gpu-monitor: { status: "online", uptime: string, restarts: number } (systemd-managed, PM2-tracked)
 - gitea-runner: { status: "online", uptime: string, restarts: number }
 - last_check: timestamp
 
-> **Note (2026-07-04):** `abiba-zulip` removed — Zulip extension decommissioned.
+> **Status (2026-08-03):** `abiba-zulip` fully restored — live Zulip bridge, heartbeating, monitored. `gpu-monitor` runs via systemd (stable); PM2 tracks it for status reporting only. `gpu-watchdog` retired from PM2.
 
 ## Continuity
 
@@ -43,18 +48,23 @@ description: >
 
 ## Execution
 
+**Execution host: CT100/.24 — run `pm2` commands LOCAL to this container. Never SSH to CT116** (CT116 runs systemd+docker, no PM2 set).
+
+**Process scope (5/5, 2026-08-09 ruling, ecosystem authoritative):**
+1. abiba-zulip (live Zulip bridge, heartbeating)
+2. abiba-telegram
+3. gitea-runner
+4. spoton-service
+5. zulip-watchdog
+
 1. **Check PM2 status** — Run `pm2 status --no-color` and parse the table (5th data column = PID, 8th = restarts, 9th = status)
-2. **Check abiba-telegram**:
-   - If status is "online" → pass
-   - If status is "stopped" or "errored" → apply Rule 1
-   - If restarts > 5 → alert owner
-3. **Check abiba-zulip** (self-process, read-only):
+2. **Check all 5 processes** (abiba-zulip, abiba-telegram, gitea-runner, spoton-service, zulip-watchdog):
    - If status is "online" → pass, log restarts count
-   - If status is "stopped" or "errored" → **DO NOT RESTART** — alert owner immediately
-   - If restarts > 5 in last hour → alert owner with full diagnostics
-4. **Log results** — Append to `SyslogSolution/health-logs/pm2/{timestamp}.md` in Gitea (not knowledge graph — hard rule)
-5. **Alert** — Send Zulip DM to owner if escalation needed (do NOT run pm2 commands during alerting)
-6. **Wait 5 min** → repeat from step 1
+   - If status is "stopped" or "errored" → apply Rule 1 (restart)
+   - If restarts > 5 → alert owner
+3. **Log results** — Append to `SyslogSolution/health-logs/pm2/{timestamp}.md` in Gitea (not knowledge graph — hard rule)
+4. **Alert** — Send Zulip DM to owner if escalation needed (do NOT run pm2 commands during alerting)
+5. **Wait 5 min** → repeat from step 1
 
 ## Example Output (when healthy)
 

scripts/agent-health-check.py

@@ -40,10 +40,10 @@ PVE_NODES = {
 
 # Agent definitions: ct, host, user, pve_node, vault_key_name
 AGENTS = {
-    "tanko":   {"ct": 112, "host": "192.168.68.122", "user": "jerome", "pve": "amdpve",  "vault_key": "TANKO_LITELLM_API_KEY"},
-    "abiba":   {"ct": 100, "host": "192.168.68.24",  "user": "root",   "pve": "hwepve",  "vault_key": None},  # Pi agent + Mumuni Zulip, no vault key
-    "koby":    {"ct": 111, "host": "192.168.68.129", "user": "root",   "pve": "amdpve",  "vault_key": "KOBY_LITELLM_API_KEY"},
-    "koonimo": {"ct": 113, "host": "192.168.68.114", "user": "root",   "pve": "amdpve",  "vault_key": "KOONIMO_LITELLM_API_KEY"},
+    "tanko":   {"ct": 112, "host": "192.168.68.122", "user": "jerome", "pve": "amdpve",  "vault_key": "TANKO_LITELLM_API_KEY", "report_only": False},
+    "abiba":   {"ct": 100, "host": "192.168.68.24",  "user": "root",   "pve": "hwepve",  "vault_key": None, "report_only": False},
+    "koby":    {"ct": 111, "host": "192.168.68.129", "user": "root",   "pve": "amdpve",  "vault_key": "KOBY_LITELLM_API_KEY", "report_only": True},  # ⛔ KOBY IS NEVER REPAIRED (Rule 17, 2026-08-17)
+    "koonimo": {"ct": 113, "host": "192.168.68.114", "user": "root",   "pve": "amdpve",  "vault_key": "KOONIMO_LITELLM_API_KEY", "report_only": False},
 }
 
 GPU_HOSTS = {
@@ -239,20 +239,26 @@ def check_agents():
         host = agent.get("host")
         user = agent.get("user")
         ct = agent["ct"]
+        report_only = agent.get("report_only", False)
 
         if not host or not user:
             print(f"  ⬜ {name} (CT {ct}): cannot SSH — skip liveness check")
             continue
 
-        # Gateway process
-        pid = ssh(host, "pgrep -f 'hermes_cli.main gateway run' | grep -v infisical | head -1", user=user)
-        if not pid:
-            # Try alternate binary name
-            pid = ssh(host, "pgrep -f 'hermes.*gateway' | grep -v infisical | grep -v bash | head -1", user=user)
-        if not pid:
-            print(f"  ❌ {name}: GATEWAY NOT RUNNING")
-            FAIL.append(f"gateway-down:{name}")
-            continue
+        # ⛔ KOBY IS NEVER REPAIRED — diagnostic only
+        if report_only:
+            print(f"  🔍 {name}: REPORT-ONLY mode (diagnostic only, no repairs on .129)")
+            # Still check gateway status for reporting purposes
+            pid = ssh(host, "pgrep -f 'hermes_cli.main gateway run' | grep -v infisical | head -1", user=user)
+            if not pid:
+                pid = ssh(host, "pgrep -f 'hermes.*gateway' | grep -v infisical | grep -v bash | head -1", user=user)
+            if not pid:
+                print(f"  ⚠️  {name}: GATEWAY NOT RUNNING (reported only)")
+                FAIL.append(f"gateway-down:{name}")
+                continue
+            else:
+                print(f"  ✅ {name}: gateway running (pid={pid}, report-only mode)")
+                continue  # Skip the rest of the check for Koby
 
         # Gateway state file
         state = ssh(host, "cat ~/.hermes/gateway_state.json 2>/dev/null", user=user)

zulip-health.prose.md

@@ -6,6 +6,8 @@ title: Zulip Mesh Health Monitor — Multi-Platform
 version: 3.0.0
 runtime_contract: 2
 agent: abiba
+report_only_agents:
+  - koby  # ⛔ KOBY IS NEVER REPAIRED (Rule 17, 2026-08-17) — detect + report, never fix on .129
 ---
 
 # Zulip Mesh Health Monitor