feat: add Level 0 auto-delete for heartbeat log orphans

- 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-fleet.prose.md

@@ -9,9 +9,9 @@ description: >
   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).
-  UPDATED 2026-07-17: Strix Halo model swapped to Genesis Hermes V3 APEX (LuffyTheFox, 24GB, uncensored,
+  UPDATED 2026-07-17: Context reduced fleet-wide from 256K to 128K for stability.
+  Strix Halo model swapped to Genesis Hermes V3 APEX (LuffyTheFox, 24GB, uncensored,
   Hermes agent fine-tune, tensor repair, multimodal with mmproj).
-  RTX 5070 swapped to HauhauCS Gemma4-12B QAT Uncensored Balanced (Q4_K_M, 87 tok/s, 0/465 refusals).
   Instability observed near 100K at 256K. 128K is the stable ceiling.
   For larger context needs → fall back to external providers (deepseek).
   VRAM headroom improved: RTX 3090 ~70%, RTX 5070 ~65%.
@@ -87,32 +87,20 @@ 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) | qwen3.6-27B-code (ThinkingCap) | Whatever runs on RTX 3090 |
+| `gpu-dense` | RTX 3090 (.8) | qwen3.6-27B-code | Whatever runs on RTX 3090 |
 | `gpu-light` | RTX 5070 (.110) | gemma-4-12b | Whatever runs on RTX 5070 |
 
 **Backward compatibility**: Old model-specific names (qwen3.6-27B-code, gemma-4-12b, qwen3.6-35B-udq4) still work
 but are deprecated for agent configs. Only the stable aliases survive model swaps.
 
-## Current Model Assignments (2026-07-17)
+## Current Model Assignments (2026-07-15)
 
 | Model | GPU | Host | VRAM | Ctx | KV Cache | Parallel | Batch/Ubatch | Status |
 |-------|-----|------|------|-----|----------|----------|-------------|--------|
-| qwen3.6-27B-code (ThinkingCap) | RTX 3090 | .8 (llm-gpu) | ~20.9/24.6GB (85%) | **128K** | turbo4 | 1 | default | ✅ 68 tok/s |
+| qwen3.6-27B-code (MTP) | RTX 3090 | .8 (llm-gpu) | ~17/24.6GB (70%) | **128K** | turbo4 | 2 | default | ✅ 63 tok/s |
-| gemma-4-12b (HauhauCS QAT) | RTX 5070 | .110 (ocu-llm) | ~10.0/12.2GB (82%) | 128K | q4_0 | 1 | 2048/1024 | ✅ 87 tok/s |
+| gemma-4-12b | RTX 5070 | .110 (ocu-llm) | ~7.8/12.2GB (65%) | 128K | q4_0 | 2 | 2048/1024 | ✅ healthy |
 | Genesis Hermes V3 APEX | Strix Halo Vulkan | .15 (amdpve) | ~10GB/64GB | 128K | q4_0 | 1 | 4096/1024 | ✅ 65 tok/s |
 
-> **RTX 5070 model swap (2026-07-17)**: Switched from `gemma-4-12b-it-IQ4_NL` (Unsloth, 191 tok/s)
-> to `HauhauCS/Gemma4-12B-QAT-Uncensored-HauhauCS-Balanced` (Q4_K_M QAT, 87 tok/s).
-> Trade: 54% slower generation for QAT quality, 0/465 refusals, and agent-optimized tuning.
-> MTP draft also swapped: Q8_0 (444MB) → tuned draft (242MB), saving 200MB VRAM.
-> Role unchanged: gpu-light (vision, web extract, light auxiliary tasks).
-
-> **RTX 3090 model swap (2026-07-17)**: Switched from `Qwopus3.6-27B-v2-MTP-Q4_K_M` (63 tok/s)
-> to `bottlecapai/ThinkingCap-Qwen3.6-27B` (Q4_K_M QAT, 68 tok/s).
-> RL-finetuned: 50% fewer thinking tokens, MMLU-Pro 0.85 vs 0.83 base.
-> Self-spec MTP REQUIRED (crashes without it on turboquant build).
-> Added vision via mmproj (0.9GB). VRAM 85%.
-
 ## Routing Configuration (LiteLLM — July 2026)
 
 ### syslog-auto Weighted Pool (Direct GPU — bypasses router)
@@ -130,16 +118,16 @@ Note: All syslog-auto entries route directly to GPUs with `api_key: not-needed`.
 | Model | RPM Cap | Notes |
 |-------|---------|-------|
 | strix-moe (Hermes V3) | 40 | Tight cap — prevents Strix overload |
-| qwen3.6-27B-code | 500 | ThinkingCap Q4_K_M + MTP self-spec + vision, 68 tok/s |
+| qwen3.6-27B-code | 500 | High cap — primary workhorse |
-| gemma-4-12b | 500 | HauhauCS QAT Uncensored Balanced + MTP, 87 tok/s |
+| gemma-4-12b | 500 | High cap — IQ4_NL+MTP, 122 tok/s |
 
 ### Stable Aliases (for agent configs — never change)
 
 | Alias | RPM Cap | Routes To | Purpose |
 |-------|---------|-----------|---------|
 | `strix-moe` | 40 | Strix Halo | Compression tasks (MoE models) |
-| `gpu-dense` | 500 | RTX 3090 (ThinkingCap) | Heavy reasoning, code gen, delegation |
+| `gpu-dense` | 500 | RTX 3090 | Heavy reasoning |
-| `gpu-light` | 500 | RTX 5070 (HauhauCS QAT) | Vision, web extract, light tasks |
+| `gpu-light` | 500 | RTX 5070 | Vision, web extract, light tasks |
 
 ### Fallback Chains
 - gemma → qwen
@@ -261,8 +249,8 @@ If no SSH access, send Zulip DM via abiba-bot with vault update instructions.
 - **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 runs `--parallel 2`** with MTP draft (spec-type draft-mtp, spec-draft-n-max 2).
-- **RTX 3090 config**: `-c 131072 -ctk turbo4 -ctv turbo4 --parallel 1 --flash-attn on --cont-batching --spec-type draft-mtp --spec-draft-n-max 4`. ThinkingCap Qwen3.6-27B Q4_K_M (15.7GB) + mmproj (0.9GB) + MTP self-spec. VRAM: ~85%. Service: `/home/llmuser/llama-wrapper.sh`. ⚠️ MTP REQUIRED for stability on this turboquant build — model segfaults without `--spec-type draft-mtp`. Outputs reasoning_content (hidden from agent, improves answer quality).
+- **RTX 3090 config**: `-c 131072 -ctk turbo4 -ctv turbo4 --parallel 2 --flash-attn on --cont-batching --spec-type draft-mtp`. Context reduced to 128K (2026-07-17, was 256K). VRAM: ~70%. Service: `/home/llmuser/llama-wrapper.sh`.
-- **RTX 5070 config (2026-07-17)**: HauhauCS Gemma4-12B QAT Uncensored Balanced (Q4_K_M) + tuned MTP draft (242MB) at 128K context, single slot. Gen speed: 87 tok/s (vs 191 IQ4_NL). VRAM: ~10.0/12.2GB (~82%). Service: `/home/llmuser/llama-wrapper.sh`. Model: `Gemma4-12B-QAT-Uncensored-HauhauCS-Balanced-Q4_K_M.gguf`, MTP: `mtp-gemma-4-12B-it.gguf`, mmproj: `mmproj-Gemma4-12B-QAT-Uncensored-HauhauCS-Balanced-BF16.gguf`. Recommended sampling: temp 0.6, top_k 64, top_p 0.9, min_p 0.05, repeat_penalty 1.1.
+- **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 2026-07-16 against `/opt/inference-harness/litellm_config.yaml` on CT 116)**: gemma-4-12b 120s, qwen3.6-27B-code 300s, 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: `LuffyTheFox/Qwen3.6-35B-A3B-Uncensored-Genesis-Hermes-V3-GGUF` (APEX quant), alias `strix-moe`, 128K context, flash-attn + q4 KV, multimodal (mmproj loaded). Hermes agent fine-tune, tensor repair (SSM layers fixed via SVD), uncensored (0/465 refusals).
 - **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).
@@ -278,8 +266,8 @@ 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) | ThinkingCap-Qwen3.6-27B Q4_K_M | **68** | — | — | **128K** |
+| RTX 3090 (.8) | qwen3.6-27B-code (MTP) | **63** | — | — | **128K** |
-| RTX 5070 (.110) | HauhauCS QAT Uncensored Balanced | **87** | — | — | **128K** |
+| RTX 5070 (.110) | gemma-4-12b (IQ4_NL+MTP) | **191** | — | — | **128K** |
 | Strix Halo (.15) | Genesis Hermes V3 APEX | **65** | 140 | — | **128K** |
 
 Benchmarks from 2026-07-17. Strix Halo swapped to Genesis Hermes V3 APEX (LuffyTheFox). RTX 5070 MTP provides 2.7x speedup over pre-upgrade 70 tok/s.

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

@@ -5,14 +5,11 @@ description: >
   Standard Hermes configuration template for Syslog Solution LLC agents.
   Enforces shared infrastructure setup (Firecrawl, SearXNG, local models,
   RA-H OS MCP) while keeping agent-specific API keys and model choices.
-  UPDATED 2026-07-18: Compression model switched to `syslog-auto` (was `strix-moe`)
+  UPDATED 2026-07-16: Compression model is the stable alias `strix-moe` (NOT `ornith-1.0-35b`,
-  to relieve Strix Halo pressure. syslog-auto distributes compression across the
-  weighted pool (55% RTX 3090, 30% Strix Halo, 15% RTX 5070).
-  UPDATED 2026-07-16: Compression model was the stable alias `strix-moe` (NOT `ornith-1.0-35b`,
   which LiteLLM does not serve). All 3 GPUs verified at 128K (reduced from 256K 2026-07-17 for stability).
   Added Rule 12 (Context-Issue Diagnostic) + Rule 13 (.env fallback enforcement) from the
   2026-07-16 Mumuni root-cause investigation (WAL #1300).
-  UPDATED 2026-07-12: GPU workload redistributed. Compression → Strix Halo (later switched to syslog-auto 2026-07-18). RTX 3090 context verified at 128K. Infisical .env fallback required (Rule 3/13).
+  UPDATED 2026-07-12: GPU workload redistributed. Compression → Strix Halo. RTX 3090 context verified at 128K. Infisical .env fallback required (Rule 3/13).
 ---
 
 ## Maintains
@@ -133,9 +130,7 @@ mcp_servers:
 # ─── Compression ───
 compression:
   enabled: true
-  model: syslog-auto              # ⚠️ Switched from strix-moe 2026-07-18 to relieve Strix Halo.
+  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).
-                                  # syslog-auto distributes across weighted pool (55% RTX 3090,
-                                  # 30% Strix Halo, 15% RTX 5070). All GPUs at 128K.
   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
@@ -150,9 +145,8 @@ compression:
 #   model: gpu-light                 # stable alias (NOT raw "gemma-4-12b")
 #   base_url: http://192.168.68.116/v1
 #   api_key_env: LITELLM_API_KEY
-# Compression uses syslog-auto (switched from strix-moe 2026-07-18) to distribute
+# Do NOT use syslog-auto for auxiliary tasks — it routes to the primary GPU.
-# load across the weighted pool and relieve Strix Halo pressure.
+# gpu-light = RTX 5070 (12B), freeing the Strix Halo for agent reasoning.
-# Vision and web_extract use gpu-light = RTX 5070 (12B).
 # 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)
 # in agent configs — use the stable aliases so model swaps don't break agents.
@@ -172,7 +166,7 @@ auxiliary:
     timeout: 30
   compression:
     provider: harness
-    model: syslog-auto             # Switched from strix-moe 2026-07-18. Relieves Strix Halo pressure.
+    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)
@@ -253,30 +247,34 @@ The following MUST be identical across ALL profiles:
 - Apply to BOTH main config AND all sub-agent profiles
 - For agents needing longer outputs: raise to 8192, but never omit
 
-### Rule 7: Auxiliary Model Consistency (UPDATED 2026-07-18)
+### Rule 7: Auxiliary Model Consistency (UPDATED 2026-07-16)
-- Vision and web_extract use `gpu-light` (stable alias, RTX 5070 — 12GB, vision-optimized)
+- Vision and web_extract use `gemma-4-12b` (RTX 5070 — 12GB, vision-optimized)
-- Compression now uses `syslog-auto` (switched from `strix-moe` 2026-07-18) to distribute
+- Compression uses `strix-moe` (stable alias for Strix Halo — 64GB, 128K ctx, compression-optimized)
-  compression load across the weighted pool (55% RTX 3090, 30% Strix Halo, 15% RTX 5070).
+- **`strix-moe` is the only valid compression model name** — LiteLLM does NOT serve `ornith-1.0-35b`
-  This relieves Strix Halo pressure while keeping compression functional on all GPUs.
+  (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.
-- **`syslog-auto` is the valid compression model** — LiteLLM serves it as the weighted pool.
+- **OPERATIONAL DECISION (2026-07-23): Use `syslog-auto` for compression across all agents.**
-  Old configs with `strix-moe` for compression should be updated to `syslog-auto`.
+  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`
-- **Compression via syslog-auto**: Routes through the weighted pool. Strix Halo still handles
+- **Do NOT use `syslog-auto`** for auxiliary tasks — it routes unpredictably
-  ~30% of compression calls (at 60 RPM via pool vs 40 RPM direct), but the bulk (55%)
+- **Compression on Strix Halo**: The strix-moe alias routes to Strix Halo
-  goes to RTX 3090 which has ample spare capacity.
+  (64GB UMA, 128K context) — the designated compression GPU. This frees the
+  RTX 5070 for vision and web search, and the RTX 3090 for heavy reasoning.
 - The `compression:` block's `model` MUST match `auxiliary: compression: model`
 - The `compression: max_context_window: 131072` MUST match actual GPU capacity (128K)
 
-### Rule 8: GPU Workload Distribution (UPDATED 2026-07-18)
+### Rule 8: GPU Workload Distribution (UPDATED 2026-07-16)
-- **RTX 3090 (24GB, 128K ctx, qwen3.6-27B-code)**: Heavy reasoning, code gen, long conversations — also handles ~55% of compression via syslog-auto pool
+- **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 — handles ~15% of compression via syslog-auto pool
+- **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, Geneis Hermes V3 APEX)**: Agent reasoning, compression (~30% via syslog-auto pool), fallback for other GPUs
+- **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: gpu-light` (RTX 5070)
+  - `auxiliary.vision.model: gemma-4-12b` (RTX 5070)
-  - `auxiliary.web_extract.model: gpu-light` (RTX 5070)
+  - `auxiliary.web_extract.model: gemma-4-12b` (RTX 5070)
-  - `auxiliary.compression.model: syslog-auto` (distributed pool, switched from strix-moe 2026-07-18)
+  - `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
@@ -332,7 +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
 ```
@@ -358,15 +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)
 

memory-fixer.prose.md

@@ -0,0 +1,71 @@
+---
+kind: pattern
+name: memory-fixer
+description: >
+  Auto-fix low-hanging fruit in the graph. No judgment calls — only deterministic Level 1 operations. 
+  Escalate anything that needs Kwame's input.
+version: 1.1.0
+---
+
+# Memory Fixer
+
+## Purpose
+Auto-fix low-hanging fruit in the graph. No judgment calls — only deterministic Level 1 operations. Escalate anything that needs Kwame's input.
+
+## Level 0 Auto-Deletes (Allowed Without Approval)
+Ephemeral heartbeat and log nodes that violate "Logs NEVER go in the graph":
+
+- `[LITELLM-HEALTH]`, `[GPU-SELF-HEAL]`, `[PM2-SELF-HEAL]`
+- `[PROXMOX-MONITOR]`, `[GPU-MONITOR]`, `[INFRA-MONITOR]`, `[AGENT-HEALTH]`, `[DISK-GC]`
+- `[WAL]` entries older than 30 days
+
+**Condition:** node must be an orphan (no edges). Deleting a connected node risks breaking other nodes.
+
+**Method:** direct SQLite on `.65` (MCP has no delete tool):
+```bash
+ssh root@192.168.68.65 "sqlite3 /root/.local/share/RA-H/db/rah.sqlite \"
+DELETE FROM nodes WHERE id IN (
+  SELECT id FROM nodes WHERE id NOT IN (SELECT from_node_id FROM edges)
+  AND id NOT IN (SELECT to_node_id FROM edges)
+  AND title LIKE '[LITELLM-HEALTH]%' -- add more prefixes as needed
+);\""
+```
+
+## Level 1 Auto-Fixes (No Judgment Required)
+
+### 1. Missing `type` Field
+For nodes with content but no `metadata.type`:
+- Title contains "Proxmox" or "infrastructure" → `type: infrastructure`
+- Title contains "skill" or "how to" or "guide" → `type: skill`
+- Title contains "doc" or "template" or "brand" → `type: documentation`
+- Title starts with "WAL:" or "TASK:" → `type: note`
+- Title starts with "[LEARN]" → `type: documentation`
+- Otherwise → `type: note` (default)
+
+### 2. Missing `tenant` / `namespace`
+For any node with NULL tenant or namespace:
+```sql
+UPDATE nodes 
+SET metadata = json_set(
+  COALESCE(metadata, '{}'),
+  '$.tenant', 'syslogsolution',
+  '$.namespace', 'syslogsolution'
+)
+WHERE json_extract(metadata, '$.tenant') IS NULL 
+   OR json_extract(metadata, '$.namespace') IS NULL;
+```
+
+### 3. Staleness State Transitions
+Using the type-based windows from the memory-monitor contract:
+- Nodes stale > their window → transition to `state: review_pending`
+- Nodes in `review_pending` for >7 days → escalate to Kwame (Level 2)
+
+## Level 2 Escalations (Kwame Decision Required)
+1. **Nodes in `review_pending` >7 days** — Archive, refresh, or keep?
+2. **Orphan Nodes >90 days old** — Delete or Connect?
+3. **Potential Duplicate Nodes** — Same title or >70% overlap. Merge or Keep?
+4. **Conflicting Metadata** — Content suggests one tenant but metadata says another.
+
+## Logging
+Every Level 1 fix logged to `~/.hermes/logs/memory-fixer/YYYY-MM-DD.md`
+Every Level 2 escalation logged and delivered to Kwame.

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}")