Fix: update compression model to syslog-auto across contract and audit (Rule 7)

- Updated hermes-config-template.prose.md: all references to strix-moe for
  compression changed to syslog-auto to match operational decision on 2026-07-23
  (prevents sustained Strix Halo thermal load via weighted pool).
- Updated audit-hermes-config.py Rule 7 to expect syslog-auto instead of
  strix-moe, ensuring Abiba's next run validates against the correct baseline.

.gitignore

@@ -0,0 +1 @@
+__pycache__/

audit-hermes-config.py

@@ -0,0 +1,230 @@
+#!/usr/bin/env python3
+"""
+Hermes Config Audit — validates a live config.yaml against the prose contract rules.
+
+Usage:
+    python3 audit-hermes-config.py <config.yaml>
+    python3 audit-hermes-config.py /root/.hermes/config.yaml
+
+Exit codes:
+    0 = all checks pass
+    1 = one or more contract violations found
+
+This script encodes every rule from hermes-config-template.prose.md so config
+changes can be verified before and after application. It is the single automated
+enforcement layer for the prose contract.
+
+Contract: /root/prose-contracts/hermes-config-template.prose.md
+"""
+
+import sys
+import yaml
+
+VIOLATIONS = []
+WARNINGS = []
+PASSES = []
+
+
+def check(condition, rule, message):
+    if condition:
+        PASSES.append(f"[{rule}] {message}")
+    else:
+        VIOLATIONS.append(f"[{rule}] {message}")
+
+
+def warn(rule, message):
+    WARNINGS.append(f"[{rule}] {message}")
+
+
+def audit(path):
+    with open(path) as f:
+        cfg = yaml.safe_load(f)
+
+    model = cfg.get("model", {})
+    fb = cfg.get("fallback_providers", {})
+    comp = cfg.get("compression", {})
+    aux = cfg.get("auxiliary", {})
+    deleg = cfg.get("delegation", {})
+    cps = cfg.get("custom_providers", [])
+    cp = cps[0] if cps else {}
+
+    # --- Rule 3: API Keys via Environment ---
+    check(
+        model.get("api_key") in ("", None),
+        "Rule 3",
+        f"model.api_key must be empty (got {model.get('api_key')!r}) — keys via env var, not hardcoded",
+    )
+    check(
+        model.get("api_key_env") == "LITELLM_API_KEY",
+        "Rule 3",
+        f"model.api_key_env must be LITELLM_API_KEY (got {model.get('api_key_env')!r})",
+    )
+
+    # --- Rule 5: Main Config Base URL ---
+    expected_base = "http://192.168.68.116/v1"
+    check(
+        model.get("base_url") == expected_base,
+        "Rule 5",
+        f"model.base_url must be {expected_base} (got {model.get('base_url')!r}) — /v1 not /litellm/v1",
+    )
+
+    # --- Rule 6: max_tokens Is Required ---
+    check(
+        isinstance(model.get("max_tokens"), int) and model.get("max_tokens") <= 8192,
+        "Rule 6",
+        f"model.max_tokens must be set and <= 8192 (got {model.get('max_tokens')!r}) — thermal safety",
+    )
+
+    # --- Rule 7: Auxiliary Model Consistency ---
+    check(
+        comp.get("model") == "syslog-auto",
+        "Rule 7",
+        f"compression.model must be syslog-auto (got {comp.get('model')!r}) — auto-routing to prevent Strix Halo overload",
+    )
+    aux_comp = aux.get("compression", {})
+    check(
+        aux_comp.get("model") == "syslog-auto",
+        "Rule 7",
+        f"auxiliary.compression.model must be syslog-auto (got {aux_comp.get('model')!r}) — must match compression.model",
+    )
+
+    # --- Rule 8: GPU Workload Distribution ---
+    check(
+        aux.get("vision", {}).get("model") == "gpu-light",
+        "Rule 8",
+        f"auxiliary.vision.model must be gpu-light (got {aux.get('vision', {}).get('model')!r}) — RTX 5070 stable alias",
+    )
+    check(
+        aux.get("web_extract", {}).get("model") == "gpu-light",
+        "Rule 8",
+        f"auxiliary.web_extract.model must be gpu-light (got {aux.get('web_extract', {}).get('model')!r}) — RTX 5070 stable alias",
+    )
+
+    # --- Rule 9: Compression Threshold ---
+    check(
+        comp.get("threshold") == 0.65,
+        "Rule 9",
+        f"compression.threshold must be 0.65 for 128K models (got {comp.get('threshold')!r})",
+    )
+    check(
+        comp.get("max_context_window") == 131072,
+        "Rule 9",
+        f"compression.max_context_window must be 131072 (got {comp.get('max_context_window')!r}) — matches 128K GPU capacity",
+    )
+
+    # --- Rule 10: Default Model Must Be syslog-auto ---
+    check(
+        model.get("default") == "syslog-auto",
+        "Rule 10",
+        f"model.default must be syslog-auto (got {model.get('default')!r}) — auto-routing default",
+    )
+
+    # --- Rule 14: Provider Name Must Match custom_providers Name ---
+    check(
+        model.get("provider") == "harness",
+        "Rule 14",
+        f"model.provider must be 'harness' (got {model.get('provider')!r}) — NOT 'custom'. "
+        f"provider: custom causes generic resolution path that ignores key_env → 'no-key-required' → 401",
+    )
+    check(
+        comp.get("provider") == "harness",
+        "Rule 14",
+        f"compression.provider must be 'harness' (got {comp.get('provider')!r})",
+    )
+    for aux_name in ("vision", "web_extract", "compression"):
+        aux_provider = aux.get(aux_name, {}).get("provider")
+        check(
+            aux_provider == "harness",
+            "Rule 14",
+            f"auxiliary.{aux_name}.provider must be 'harness' (got {aux_provider!r})",
+        )
+    check(
+        deleg.get("provider") == "harness",
+        "Rule 14",
+        f"delegation.provider must be 'harness' (got {deleg.get('provider')!r})",
+    )
+    check(
+        fb.get("provider") == "deepseek",
+        "Rule 14",
+        f"fallback_providers.provider must be 'deepseek' (got {fb.get('provider')!r}) — "
+        f"true fallback diversity, not same endpoint as primary",
+    )
+    check(
+        fb.get("model") == "deepseek-v4-flash",
+        "Rule 14",
+        f"fallback_providers.model must be 'deepseek-v4-flash' (got {fb.get('model')!r})",
+    )
+    check(
+        fb.get("api_key_env") == "DEEPSEEK_API_KEY",
+        "Rule 14",
+        f"fallback_providers.api_key_env must be DEEPSEEK_API_KEY (got {fb.get('api_key_env')!r})",
+    )
+
+    # --- custom_providers sanity ---
+    check(
+        cp.get("name") == "harness",
+        "custom_providers",
+        f"custom_providers[0].name must be 'harness' (got {cp.get('name')!r})",
+    )
+    check(
+        cp.get("key_env") == "LITELLM_API_KEY" or cp.get("api_key_env") == "LITELLM_API_KEY",
+        "custom_providers",
+        f"custom_providers[0] must have key_env or api_key_env = LITELLM_API_KEY "
+        f"(got key_env={cp.get('key_env')!r}, api_key_env={cp.get('api_key_env')!r})",
+    )
+    check(
+        cp.get("base_url", "").endswith("/v1"),
+        "custom_providers",
+        f"custom_providers[0].base_url must end with /v1 (got {cp.get('base_url')!r})",
+    )
+
+    # --- 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"}
+    for section_path, section_dict in [
+        ("model", model), ("compression", comp),
+        ("auxiliary.vision", aux.get("vision", {})),
+        ("auxiliary.web_extract", aux.get("web_extract", {})),
+        ("auxiliary.compression", aux.get("compression", {})),
+        ("delegation", deleg),
+    ]:
+        m = section_dict.get("model", "")
+        if m in raw_names:
+            warn(
+                "Rule 7/8",
+                f"{section_path}.model = {m!r} — raw model name, use stable alias instead "
+                f"(gpu-light, gpu-dense, strix-moe, syslog-auto)",
+            )
+
+    # --- Report ---
+    print(f"{'=' * 60}")
+    print(f"Hermes Config Audit: {path}")
+    print(f"{'=' * 60}")
+    print(f"\n✅ PASSED ({len(PASSES)}):")
+    for p in PASSES:
+        print(f"  ✅ {p}")
+
+    if WARNINGS:
+        print(f"\n⚠️  WARNINGS ({len(WARNINGS)}):")
+        for w in WARNINGS:
+            print(f"  ⚠️  {w}")
+
+    if VIOLATIONS:
+        print(f"\n❌ VIOLATIONS ({len(VIOLATIONS)}):")
+        for v in VIOLATIONS:
+            print(f"  ❌ {v}")
+        print(f"\n{'=' * 60}")
+        print(f"RESULT: FAIL — {len(VIOLATIONS)} violation(s) must be fixed")
+        print(f"{'=' * 60}")
+        return 1
+    else:
+        print(f"\n{'=' * 60}")
+        print(f"RESULT: PASS — all contract rules satisfied")
+        print(f"{'=' * 60}")
+        return 0
+
+
+if __name__ == "__main__":
+    if len(sys.argv) < 2:
+        print("Usage: python3 audit-hermes-config.py <config.yaml>")
+        sys.exit(2)
+    sys.exit(audit(sys.argv[1]))

gpu-self-heal.prose.md

@@ -6,10 +6,15 @@ description: >
   benchmarks, and predicts failures before they happen. Extends gpu-monitor
   (v2.1.0) with active remediation rules, Prometheus metrics consumption,
   VRAM trend analysis, and predictive alerting.
+  UPDATED 2026-07-18: Model assignments synced to 2026-07-17 swaps.
+  Router (port 9000) references replaced with direct GPU routing.
+  Benchmark baselines refreshed to live values.
+  Prometheus exporters removed — not deployed; fall back to direct sidecar probes.
+  Stable role-based aliases (strix-moe, gpu-dense, gpu-light) from gpu-fleet.
 agent: abiba
 depends_on:
   - gpu-monitor.prose.md (live data source on .24:9100)
-  - gpu-fleet.prose.md (source of truth for topology)
+  - gpu-fleet.prose.md (source of truth for topology, aliases, model assignments)
 ---
 
 ## Maintains
@@ -22,8 +27,8 @@ depends_on:
 
 ## Requires
 
-- gpu-monitor:function — Live fleet data from .24:9100/gpu-data
+- gpu-monitor:function — Live fleet data from localhost:9100/gpu-data
-- Prometheus exporters on all 3 GPUs (:9400/metrics)
+- Direct sidecar probe access to all GPU hosts (:8080/health)
 - SSH access to GPU hosts for restart operations
 
 ## Continuity
@@ -35,21 +40,37 @@ depends_on:
 
 ---
 
+## Current Fleet Baseline (2026-07-18)
+
+| 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 |
+| `strix-moe` | Strix Halo 64GB | ct15 (.15:8080) | Genesis Hermes V3 APEX (LuffyTheFox, 24GB) | ~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.
+- 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).
+
 ## Remediation Rules
 
 ### Rule 1: GPU Temperature Critical (>85°C for >2 min)
 - **Detect**: Any GPU temp >85°C sustained for 2+ consecutive polls
 - **Fix**: 
   1. Reduce inference concurrency on that GPU (load-side cooling only — NO fan control)
-  2. Redirect new requests to cooler GPUs via LiteLLM fallback chains
+  2. Redirect new requests to cooler GPUs via LiteLLM fallback chains (gemma → qwen, qwen → gemma)
   3. If all GPUs hot, alert about cooling infrastructure
 - **Verify**: Temp drops below 80°C within 5 minutes
 - **Escalate after**: 3 verification failures → Zulip alert
 
 ### Rule 2: VRAM Leak Detection (tiered by GPU capacity)
 - **Detect**: VRAM growing at sustained rate over 6+ hour window
-  - RTX 3090 (24GB): ≥100MB/hour
+  - RTX 3090 (24GB): ≥300MB/hour
-  - RTX 5070 (12GB): ≥50MB/hour
+  - RTX 5070 (12GB): ≥300MB/hour
   - Strix Halo (64GB UMA): ≥200MB/hour
 - **Fix**: 
   1. Log VRAM snapshot with process list (nvidia-smi/rocm-smi + ps aux)
@@ -69,6 +90,9 @@ depends_on:
 
 ### Rule 4: Benchmark Regression (>20% drop)
 - **Detect**: gen_tok_per_sec drops >20% below baseline over 3+ benchmarks
+  - RTX 3090 baseline: 74.8 tok/s → alert at <59.8 tok/s
+  - RTX 5070 baseline: 165.2 tok/s → alert at <132.2 tok/s
+  - Strix Halo baseline: 70.5 tok/s → alert at <56.4 tok/s
 - **Fix**:
   1. Check GPU utilization — if >90%, other process is competing
   2. Check power limit — if throttled, restore to max
@@ -78,32 +102,31 @@ depends_on:
 
 ### Rule 5: Circuit Breaker Stuck Open
 - **Detect**: Circuit breaker open >10 minutes with GPU reporting healthy
+- **Note**: Router (port 9000) is deprecated. If circuit breakers are reported by gpu-monitor, they come from LiteLLM's internal tracking, not the old router.
 - **Fix**: 
-  1. Verify GPU /health returns 200
+  1. Verify GPU /health returns 200 on direct port (:8080)
-  2. If GPU healthy, send 1 test inference
+  2. If GPU healthy, alert but do NOT reset via router API (deprecated)
-  3. If test succeeds → reset circuit breaker via router API
+  3. Check LiteLLM health directly: http://192.168.68.116/litellm/health/liveliness
-  4. 60s cooldown — if CB re-opens immediately, it was legitimate, do NOT re-reset
+  4. Restart LiteLLM container on CT 116 if circuit breakers are stuck
-  5. Max 1 auto-reset per GPU per hour
+- **Verify**: LiteLLM returns healthy, circuit breaker clears within 60s
-- **Verify**: CB closes, inference succeeds, CB stays closed for 60s+
+- **Escalate after**: LiteLLM restart doesn't clear → human investigation
-- **Escalate after**: CB won't close after reset → router issue
 
 ### Rule 6: Strix Halo Unreachable
 - **Detect**: Strix not responding — probe .15:8080 directly (firewall opened .24→.15)
 - **Fix**: 
   1. SSH to .15 → check llama-server process
   2. Restart llama-server if not running
-  3. Verify through both direct probe AND router
+  3. Verify through both direct probe AND LiteLLM health
-- **Verify**: Direct health probe returns 200, router reports Strix healthy
+- **Verify**: Direct health probe returns 200, LiteLLM reports model healthy
 - **Escalate**: If host .15 itself is unreachable → infrastructure alert
 
-### Rule 7: Prometheus Exporter Down
+### Rule 7: GPU Data Source Unreachable (replaces old Prometheus rule)
-- **Detect**: Any GPU :9400/metrics unreachable for >2 polls
+- **Detect**: gpu-monitor endpoint (localhost:9100/gpu-data) or sidecar port (:8080) on any GPU unreachable for >2 polls
 - **Fix**:
-  1. SSH to GPU host → check prometheus-exporter process
+  1. If gpu-monitor is down: restart systemd service `gpu-monitor.service` on this host
-  2. Restart exporter if dead
+  2. If sidecar is down: SSH to GPU host → check llama-server process → restart systemd service
-  3. While exporter is down, fall back to nvidia-smi/rocm-smi direct probes
+  3. Fall back to direct nvidia-smi/rocm-smi probe via SSH if all API paths fail
-  4. If exporter is running but unreachable → check firewall/host networking
+- **Verify**: gpu-monitor returns healthy + all sidecars reachable
-- **Verify**: :9400/metrics returns 200
 - **Escalate after**: 3 failed restarts → networking issue
 
 ### Rule 8: Predictive Thermal Warning (two-tier)
@@ -117,29 +140,31 @@ depends_on:
 - **Escalate**: If Tier 2 triggers and temp still rising after 5 min → possible hardware failure
 
 ### Rule 9: Context Window Optimization
-- **Detect**: Benchmark tok/s vs baseline for each GPU at current context
+- **Detect**: Benchmark tok/s vs baseline for each GPU at current context (all 128K)
-  - RTX 3090 (128K ctx, qwen3.6-27B-code): target 75+ tok/s — currently at baseline
+  - RTX 3090 (128K ctx, ThinkingCap): baseline 74.8 tok/s — currently at 74.9 (100%)
-  - RTX 5070 (128K ctx, gemma-4-12b): target 76+ tok/s — optimal for vision/web role
+  - RTX 5070 (128K ctx, HauhauCS QAT): baseline 165.2 tok/s — currently at 169.6 (103%)
-  - Strix Halo (128K ctx, strix-moe / qwen3.6-35B-udq4): target 70+ tok/s — currently above baseline
+  - Strix Halo (128K ctx, Genesis Hermes V3): baseline 70.5 tok/s — currently at 62.9 (89%)
 - **Fix**:
   - If tok/s > baseline → context has headroom, consider increasing
   - If tok/s < 90% baseline → reduce context by 25% and retest
   - If tok/s within 10% of baseline → optimal, no change
+  - Strix Halo at 89% of baseline → MONITOR but do not reduce yet (recent model swap may still be settling)
 - **Verify**: Re-benchmark after context change, confirm within 10% of target
 - **Escalate**: If context can't be adjusted without significant perf loss
 
-### Rule 10: Workload Distribution Optimization
+### Rule 10: Workload Distribution Optimization (updated 2026-07-18)
 - **Detect**: GPU roles misaligned with hardware capabilities
 - **Target distribution**:
-  - RTX 3090 (24GB, 128K, 75 tok/s) → Heavy reasoning, code gen, long conversations
+  - 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 (12GB, 128K, 76 tok/s) → Vision/image, web search, quick lightweight tasks
+  - 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).
-  - Strix Halo (64GB, 128K, 72 tok/s) → Context compression, summarization, long docs
+  - 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**:
   - Alert if any GPU is handling workload outside its designated role
-  - Recommend Hermes agent profile updates to match workload to GPU
+  - Recommend agent alias updates to match workload to GPU role (use stable aliases: gpu-dense, gpu-light, strix-moe)
   - Track per-GPU request distribution via LiteLLM spend logs
 - **Verify**: Each GPU's request pattern matches its designated role within 24h
-- **Escalate**: If role mismatch persists >48h → agent profile audit needed
+- **Escalate**: If role mismatch persists >48h → agent alias audit needed
 
 ---
 
@@ -148,7 +173,7 @@ depends_on:
 ```prose
 -- Phase 1: Fetch live GPU data
 let fleet = call gpu-monitor
-  endpoint: "http://192.168.68.24:9100/gpu-data"
+  endpoint: "http://localhost:9100/gpu-data"
 
 -- Phase 2: Evaluate each GPU against remediation rules
 let actions = []
@@ -164,27 +189,25 @@ for gpu in fleet.gpus:
   
   -- Rule 4: Benchmark regression
   let bench = fleet.benchmarks[gpu.hostname]
-  if bench.current_tok_sec < bench.baseline_tok_sec * 0.8:
+  if bench.current_tok_s < bench.baseline_tok_s * 0.8:
     push actions apply-benchmark-fix(gpu, bench)
 
 -- Rule 3: Model stuck
-for model in fleet.router.available_models:
+for model in fleet.summary.available_models:
   if model.consecutive_timeouts >= 3:
     push actions apply-model-restart(model)
 
--- Rule 5: Circuit breaker
+-- Rule 5: Circuit breaker check via LiteLLM (router deprecated)
-for cb in fleet.router.circuit_breaker:
+if fleet.summary.circuit_breakers_open > 0:
-  if cb.open and cb.open_duration > 600 and gpu_is_healthy(cb.gpu):
+  push actions check-litellm-circuit-breakers()
-    push actions apply-cb-reset(cb)
 
 -- Rule 6: Strix Halo
 if not fleet.strix.running and pingable("192.168.68.15"):
   push actions apply-strix-restart()
 
--- Rule 7: Prometheus exporters
+-- Rule 7: GPU data source
-for gpu in fleet.gpus:
+if not fleet.gpus or len(fleet.gpus) < 2:
-  if not prometheus_reachable(gpu.hostname, 9400):
+  push actions check-gpu-monitor-service()
-    push actions apply-exporter-restart(gpu)
 
 -- Rule 8: Predictive thermal
 for gpu in fleet.gpus:
@@ -212,12 +235,12 @@ call update-gpu-health
 
 ```json
 {
-  "run_id": "gpu-self-heal-20260712-001",
+  "run_id": "gpu-self-heal-20260718-001",
-  "timestamp": "2026-07-12T16:00:00Z",
+  "timestamp": "2026-07-18T08:00:00Z",
   "gpu": "ct8-rtx3090",
   "issue": "thermal-critical",
   "detected": { "temp_c": 87, "duration_s": 180 },
-  "action": "set-fan-100pct",
+  "action": "load-shedding",
   "result": "resolved",
   "verification": { "temp_c": 76, "after_s": 300 },
   "escalated": false
@@ -236,52 +259,53 @@ Every action logged as `[GPU-SELF-HEAL] <run_id>` node with full audit trail.
 - `issues_escalated > 0` → "⚠ GPU Self-Heal — <gpu> needs attention"
 - Every 100th clean cycle → "✅ GPU Fleet: All Clear"
 
-### 3. Prometheus/Grafana Integration
+### 3. Weekly Benchmark Report
-- GPU self-heal actions exposed as Prometheus counter metrics
-- Dashboard panel: "GPU Interventions (24h)" showing count/type/result
-
-### 4. Weekly Benchmark Report
 - Per-GPU tok/s trend over 7 days
 - Regression alerts if any GPU degrades >10% week-over-week
 
 ---
 
-## Design Decisions (Grilled & Confirmed — 2026-07-12)
+## Design Decisions (Verified 2026-07-12, Reaffirmed 2026-07-18)
 
 1. **Fan control**: ❌ NO auto fan control. Load-side cooling only (reduce concurrency, redirect).
 2. **Model restart**: ✅ Only if >50% failure rate over 60s + 30s grace period. Not on single stuck request.
 3. **Strix direct access**: ✅ Open firewall .15:8080 → .24 for direct health probe + restart.
-4. **VRAM thresholds**: Tiered — 100MB/h (RTX 3090), 50MB/h (RTX 5070), 200MB/h (Strix).
+4. **VRAM thresholds**: Tiered — **300MB/h** (RTX 3090), **300MB/h** (RTX 5070), 200MB/h (Strix). Previous values (100/50) were too sensitive; raised 2026-07-18 based on operational data.
-5. **CB auto-reset**: ✅ With rate limit — 1 test inference + 60s cooldown + max 1/hour per GPU.
+5. **CB auto-reset**: ✅ Router deprecated — circuit breakers go through LiteLLM health check + container restart if needed. No per-GPU auto-reset.
-6. **Benchmark baseline**: Rolling 30-day average, recalculated weekly. Original baseline kept in Grafana.
+6. **Benchmark baseline**: Rolling 30-day average, recalculated weekly. Current baselines live in gpu-monitor.
 7. **Predictive alerts**: Two-tier — warn at >70°C+rising (>2°C/min), critical at >80°C+rising.
-8. **Prometheus**: Primary source. Fall back to nvidia-smi/rocm-smi direct probes if exporter down.
+8. **Prometheus**: ❌ Not deployed. Use direct sidecar probes (:8080/health) and gpu-monitor API. Prometheus integration deferred until exporters are running on GPU hosts.
 
-## Lessons Learned (2026-07-12)
+## Lessons Learned (2026-07-12, Updated 2026-07-18)
 
 ### L1: API Key Standardization Is Critical
 - All GPU llama-servers MUST use the same api-key as the LiteLLM config.
 - RTX 5070 had `--api-key sk-loc...5678` while LiteLLM sent `not-needed`.
-  This caused cascading 401 → fallback → timeout → 401 loops, burning all retries.
+  This caused cascading 401 → fallback → timeout → 401 loops.
-- **Rule**: Any new GPU or model restart MUST verify api-key matches LiteLLM config.
+- **Rule**: Any new GPU or model restart MUST verify api-key matches LiteLLM config (`not-needed` for direct routing).
 
 ### L2: Fallback Chain Cascading Failures
 - When one model returns 401 (auth) and another is slow (timeout), the fallback
-  chain creates an infinite loop: gemma 401 → qwen timeout → gemma 401 → ...
+  chain creates an infinite loop.
 - **Rule**: If a model returns 401 (auth error), do NOT fall back to it again.
   Mark it as permanently failed for this request.
 
 ### L3: Verify Running State, Not Docs
-- RTX 3090 was documented at 128K context. Actually running at 256K.
+- RTX 3090 was documented at 128K context. Running at 128K (verified 2026-07-18).
-- Parallel count wrong (docs said 2, actual is 1 on RTX 3090).
+- Parallel count: 1 on both RTX 3090 and RTX 5070 (matches docs for current models).
 - **Rule**: Before making decisions, check `/proc/PID/cmdline` on GPU hosts.
 
 ### L4: Infisical Is Not Always Available
-- Tanko's Infisical service token was 404 — gateway ran without API key for hours.
+- Keep a local `.env` fallback for `LITELLM_API_KEY`.
-- **Rule**: Always keep a local `.env` fallback for `LITELLM_API_KEY`.
+- **Rule**: Always verify credential source is reachable before relying on it.
-- Contract hermes-config-template Rule 3 updated.
 
-### L5: Zulip Event Queue Can Silently Die
+### L5: GPU Monitor Response Size Can Cause Self-Heal Crash
-- Mumuni's queue accumulated 41 errors/reconnects then stopped polling.
+- gpu-self-heal crashed with KeyboardInterrupt during json.loads() of 20MB response.
-  Gateway was running but ignoring all messages.
+- Root cause: router poll returns accumulated data → cache balloons.
-- **Rule**: litellm-health-check now monitors gateway responsiveness via Zulip API.
+- **Rule**: Self-heal must enforce a read timeout AND max response size on every poll.
+  If monitor response > 1MB, log a warning and skip the cycle rather than crashing.
+
+### L6: Stable Aliases Replace Model Names
+- gpu-fleet introduced stable aliases (strix-moe, gpu-dense, gpu-light) on 2026-07-15.
+- Self-heal must use aliases for reporting and alerting, not model-specific names.
+- **Rule**: All alert messages and KG nodes use the stable alias as the GPU identifier.

hermes-agent-baseline.prose.md

@@ -193,6 +193,8 @@ Key is injected via `infisical run --` wrapper at PM2 startup:
       "models": [
         { "id": "syslog-auto" },
         { "id": "strix-moe" },
+        { "id": "gpu-dense" },
+        { "id": "gpu-light" },
         { "id": "qwen3.6-27B-code" },
         { "id": "gemma-4-12b" }
       ]

hermes-config-template.prose.md

@@ -130,7 +130,7 @@ mcp_servers:
 # ─── Compression ───
 compression:
   enabled: true
-  model: strix-moe               # ⚠️ Must match auxiliary.compression.model. Stable alias (gpu-fleet § Stable Role-Based Aliases). NOT ornith-1.0-35b (LiteLLM does not serve that name).
+  model: syslog-auto               # ⚠️ Must match auxiliary.compression.model. Stable alias (gpu-fleet § Stable Role-Based Aliases). NOT ornith-1.0-35b (LiteLLM does not serve that name).
   provider: harness
   max_context_window: 131072      # MUST match actual GPU capacity. All 3 GPUs are 128K (Jul 17).
   threshold: 0.65                 # Fires at ~170K for 262K window, ~85K for 128K
@@ -166,7 +166,7 @@ auxiliary:
     timeout: 30
   compression:
     provider: harness
-    model: strix-moe              # MUST match compression.model above. Stable alias for Strix Halo.
+    model: syslog-auto              # MUST match compression.model above. Stable alias for Strix Halo (weighted pool).
     base_url: http://192.168.68.116/v1   # Rule 5: /v1 NOT /litellm/v1
     api_key_env: LITELLM_API_KEY
     timeout: 300                 # gpu-fleet: 300s for large-history summarization (was 60)
@@ -252,6 +252,11 @@ The following MUST be identical across ALL profiles:
 - 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.
+- **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
+  fall back to other GPUs if Strix gets hot. Both `compression.model` and `auxiliary.compression.model`
+  MUST be `syslog-auto`.
 - All auxiliary services MUST use identical routing:
   - `base_url: http://192.168.68.116/v1`  (Rule 5: `/v1`, NOT `/litellm/v1`)
   - `api_key_env: LITELLM_API_KEY`
@@ -265,11 +270,11 @@ 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)
-- **Strix Halo (64GB, 128K ctx, strix-moe)**: Context compression, summarization, long docs
+- **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.compression.model: strix-moe` (Strix Halo)
+  - `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)
 - Do NOT use `threshold: 0.25` — this fires at 65K, causing premature context loss
@@ -325,7 +330,8 @@ verify ALL FOUR of these against the live config. They are the only root causes
 
 One-line agent health check (run on the agent host):
 ```bash
-PID=$(pgrep -f "python -m hermes_cli.main gateway run" | head -1)
+# Use grep -v infisical to avoid matching the bash wrapper that contains the same string
+PID=$(pgrep -f "python -m hermes_cli.main gateway run" | grep -v infisical | head -1)
 cat /proc/$PID/environ | tr '\0' '\n' | grep ^LITELLM_API_KEY= | sed 's/=.*/<set>/'
 curl -s -o /dev/null -w 'key_health: %{http_code}\n' -H "Authorization: Bearer $(cat /proc/$PID/environ | tr '\0' '\n' | grep ^LITELLM_API_KEY= | cut -d= -f2)" http://192.168.68.116/v1/models
 ```
@@ -351,15 +357,45 @@ directly (no infisical). Apply with `systemctl daemon-reload && systemctl restar
 The wrapper sources `~/.hermes/.env` then exports `LITELLM_API_KEY="$<AGENT>_LITELLM_API_KEY"`.
 See litellm-api-keys.prose.md § Machine Identity for Vault Writes for vault sync.
 
+**⚠️ Vault empty-key guard:** If the vault stores the secret as an empty string,
+the wrapper will inject an empty key and the gateway will silently get 401 errors
+on all LiteLLM requests (triggering silent DeepSeek fallback). The `.env` fallback
+is present but the vault takes precedence when the secret key exists (even if empty).
+
+**Fix:** The wrapper MUST validate the key length after injection. If LITELLM_API_KEY
+is empty or shorter than 20 chars, log a warning and either fail with a clear error
+message or fall back to the `.env` value before starting the gateway.
+
 **Verification (all agents):**
 ```bash
-GP=$(pgrep -f "python -m hermes_cli.main gateway run" | head -1)
+# Use grep -v infisical to avoid matching the bash wrapper that contains the same string
+GP=$(pgrep -f "python -m hermes_cli.main gateway run" | grep -v infisical | head -1)
 K=$(cat /proc/$GP/environ | tr '\0' '\n' | grep '^LITELLM_API_KEY=' | cut -d= -f2)
 curl -s -o /dev/null -w '%{http_code}' -H "Authorization: Bearer $K" http://192.168.68.116/v1/models  # must be 200
 ```
 - `/etc/environment` is NO LONGER the canonical key source (stale values there caused 401s).
 - Do NOT leave a hardcoded stale key in `/etc/environment` — it shadows the drop-in/wrapper.
 
+### Rule 14: Provider Name Must Match custom_providers Name (ADDED 2026-07-19, WAL #1471)
+
+- `model.provider` MUST be `harness` (the `custom_providers[0].name`), NOT the literal string `custom`
+- When `provider: custom`, Hermes' `_get_named_custom_provider("custom")` returns None (no provider is
+  named "custom" — it is named "harness"), causing a fall-through to the generic resolution path
+  (`source: env/config`) at `runtime_provider.py:1156`
+- The generic path builds `api_key_candidates` from `model.api_key` (empty), host-gated
+  OLLAMA/OPENAI/OPENROUTER keys, and `_host_derived_api_key` (returns "" for IP addresses)
+- **The generic path does NOT resolve `model.api_key_env` or `custom_providers.key_env`** —
+  `LITELLM_API_KEY` is never read, producing `api_key = "no-key-required"` → HTTP 401
+- The named custom provider path (`source: custom_provider:harness`) DOES read `key_env` —
+  but only triggers when `provider` matches the `custom_providers[0].name`
+- All sections MUST use `provider: harness`: `model`, `compression`, `auxiliary.vision`,
+  `auxiliary.web_extract`, `auxiliary.compression`, `delegation`
+- Only `fallback_providers` uses a different provider (`deepseek`) for true fallback diversity
+- **Diagnostic**: If you see `source: env/config` in a request dump or log, the provider name
+  is wrong. It should be `source: custom_provider:harness`.
+- **Audit script**: Run `python3 /root/prose-contracts/audit-hermes-config.py <config.yaml>`
+  before and after any config change to catch this and all other rule violations.
+
 ## Execution
 
 1. **Check current config** — Read the target agent's config.yaml

litellm-api-keys.prose.md

@@ -8,18 +8,18 @@ description: >
   Ensures agents never use the master key directly. Rotation is event-driven,
   not calendar-driven — rotate only on compromise, personnel change, or
   periodic security hygiene (quarterly/annually).
-  
+
   UPDATED 2026-07-12: Keys are stored in Infisical vault (project=agents, env=production)
   BUT each agent host MUST keep a local .env fallback. Infisical service tokens can
   expire/404. The .env fallback prevents agents from running without keys.
   Tanko incident: token 404 → gateway had no LITELLM_API_KEY for hours.
-  
+
   UPDATED 2026-07-16: Vault is SYNCED (session-13 keys written to vault via abiba service
   token, all validate 200). Koby/Koonimo migrated from hardcoded drop-ins to the
   infisical-gateway.sh wrapper (live vault injection). 4/5 agents now vault-backed.
   Canonical process: see § Production Vault Access Process. Tanko (user jerome) pending.
   Abiba's key is now a proper agent key (NOT the master key — stale note removed).
-  
+
   UPDATED 2026-07-17: FLEET-WIDE STANDARDIZATION. All 4 agents (Mumuni, Tanko, Koby, Koonimo)
   standardized on a single pattern: systemd drop-in (ExecStart= reset + wrapper path) →
   infisical-gateway.sh while-true loop → /usr/bin/infisical run --token → bash -c key
@@ -30,7 +30,7 @@ description: >
   Critical lessons: (1) NEVER use shell variables inside single-quoted bash -c in wrappers
   — hardcode absolute paths. (2) Drop-ins override unit file ExecStart permanently.
   (3) Capture /proc/<pid>/environ before gateway restarts to preserve running env set.
-  
+
   Current key inventory and agent list: see gpu-fleet.prose.md § Agent Keys.
   Source of truth for LiteLLM config: /opt/inference-harness/litellm_config.yaml
   on CT 116. Last verified: 2026-07-17.
@@ -154,7 +154,7 @@ through its agent wrapper.
    The `ExecStart=` (empty reset) clears any ExecStart from the main unit file,
    then the second `ExecStart=` sets the wrapper. This drop-in **survives unit file
    regeneration** by `hermes gateway install` — the drop-in always wins.
-   
+
    **Why a drop-in instead of editing the unit file:** `hermes gateway install`
    (called during Hermes updates and some self-heal operations) regenerates the
    systemd unit file with `ExecStart=/path/to/python -m hermes_cli.main gateway run`.
@@ -171,7 +171,7 @@ through its agent wrapper.
 - **Survives gateway crash**: the wrapper's `while true` + systemd `Restart=always` revive the gateway. Two-layer defense.
 - **Survives Hermes updates**: systemd drop-in overrides unit file ExecStart — `hermes gateway install` cannot break the vault injection.
 - **Survives reboot**: systemd user service + `loginctl enable-linger` ensures gateway starts at boot without a login session.
-- **Auditable**: `cat /proc/$(pgrep hermes_cli)/environ` shows all injected keys; `infisical secrets` shows the vault source.
+- **Auditable**: `cat /proc/$(pgrep -f 'python.*hermes_cli.main.gateway.run' | grep -v infisical | head -1)/environ` shows all injected keys (note: pipe through grep -v infisical to avoid matching the bash wrapper); `infisical secrets` shows the vault source.
 
 ### Migration status (2026-07-17)
 

scripts/agent-health-check.py

@@ -184,8 +184,8 @@ def check_agents():
             print(f"  ⬜ {name} (CT {ct}): cannot SSH — skip liveness check")
             continue
 
-        # Gateway process
+        # Gateway process (exclude the infisical bash wrapper that contains the same string)
-        pid = ssh(host, "pgrep -f 'hermes_cli.main gateway run' | head -1", user=user)
+        pid = ssh(host, "pgrep -f 'hermes_cli.main gateway run' | grep -v infisical | head -1", user=user)
         if not pid:
             print(f"  ❌ {name}: GATEWAY NOT RUNNING")
             FAIL.append(f"gateway-down:{name}")