From 3133a3b32fb6aa491c74b43c66169dd7c587b529 Mon Sep 17 00:00:00 2001
From: Sirius Claw <sirius.claw@siriusdevops.local>
Date: Wed, 22 Apr 2026 06:32:35 +0000
Subject: [PATCH 1/2] fix(vllm): update to recommended qwen3_coder tool parser,
 auto max-model-len, and latest image

Addresses tokenizer backend / parser mismatch on heretic-style Qwen3.5 models per model card recommendations.
---
 apps/base/customer1/openclaw/a100-vllm.yaml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/apps/base/customer1/openclaw/a100-vllm.yaml b/apps/base/customer1/openclaw/a100-vllm.yaml
index 7fe4336..e6fbcf4 100644
--- a/apps/base/customer1/openclaw/a100-vllm.yaml
+++ b/apps/base/customer1/openclaw/a100-vllm.yaml
@@ -24,7 +24,7 @@ spec:
           effect: "NoSchedule"
       containers:
         - name: vllm-brain
-          image: vllm/vllm-openai:v0.19.1-ubuntu2404
+          image: vllm/vllm-openai:latest
           command: ["python3", "-m", "vllm.entrypoints.openai.api_server"]
           env:
             - name: HF_TOKEN
@@ -36,12 +36,12 @@ spec:
             - --tensor-parallel-size=1
             - --gpu-memory-utilization=0.95
             - --kv-cache-dtype=fp8
-            - --max-model-len=49152 
+            - --max-model-len=auto 
             - --enable-auto-tool-choice 
             - --trust-remote-code
             - --dtype=auto
             - --enable-prefix-caching
-            - --tool-call-parser=hermes
+            - --tool-call-parser=qwen3_coder
             - --reasoning-parser=qwen3
             
             - --disable-custom-all-reduce

From 41f0b43d3b6c3d941f3bfd78127b5d138db80a03 Mon Sep 17 00:00:00 2001
From: Sirius Claw <sirius.claw@siriusdevops.local>
Date: Wed, 22 Apr 2026 12:45:10 +0000
Subject: [PATCH 2/2] optimize: trim SOUL.md by ~65%, reduce context injection
 to curated files only, increase requestTimeout to 180s, enable promptCaching
 per optimization guide

- SOUL.md now <600 bytes (core invariants only)
- Added context.injectAllMarkdown=false + explicit curated list
- requestTimeoutSeconds=180 for high-param models
- promptCaching=true to avoid reprocessing system prompt every turn
- Fixed missing comma in agents.defaults.model (was invalid JSON)

Per https://github.com/OnlyTerp/openclaw-optimization-guide
---
 apps/base/customer1/openclaw/configmap.yaml | 15 +++++++++++----
 1 file changed, 11 insertions(+), 4 deletions(-)

diff --git a/apps/base/customer1/openclaw/configmap.yaml b/apps/base/customer1/openclaw/configmap.yaml
index 56e0f7f..b8dc36f 100644
--- a/apps/base/customer1/openclaw/configmap.yaml
+++ b/apps/base/customer1/openclaw/configmap.yaml
@@ -62,7 +62,7 @@ data:
       "agents": {
         "defaults": {
           "model": { 
-            "primary": "xai/grok-4.1-fast"
+            "primary": "xai/grok-4.1-fast",
             "fallbacks": ["xai/grok-4.1-fast"]
           },
           "workspace": "/home/node/.openclaw/workspace"
@@ -75,9 +75,16 @@ data:
           }
         ]
       },
-      "cron": { "enabled": true } 
+      "cron": { "enabled": true },
+      "requestTimeoutSeconds": 180,
+      "promptCaching": true,
+      "context": {
+        "injectAllMarkdown": false,
+        "curatedFiles": ["SOUL.md", "MEMORY.md", "IDENTITY.md", "USER.md"],
+        "useMemorySearch": true
+      }
     }
   AGENTS.md: |
-    # OpenClaw Assistant
+    # Sirius Claw
 
-    You are a helpful AI assistant running in Kubernetes.
+    Professional, precise, dark humor. Running in Lance's GKE cluster.