From c228adddc9b0d0731dd930f083684ea19221de92 Mon Sep 17 00:00:00 2001
From: Google AI Agent <gemini@hermes.local>
Date: Mon, 6 Apr 2026 17:44:44 +0000
Subject: [PATCH 1/2] Optimize smart_model_routing for higher leverage

---
 config.yaml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/config.yaml b/config.yaml
index 88b64e02..340c24a1 100644
--- a/config.yaml
+++ b/config.yaml
@@ -53,8 +53,8 @@ synthesis_model:
 
 smart_model_routing:
   enabled: true
-  max_simple_chars: 400
-  max_simple_words: 75
+  max_simple_chars: 1000
+  max_simple_words: 200
   cheap_model:
     provider: 'ollama'
     model: 'gemma2:2b'
-- 
2.43.0


From 99b54a31d1ee81ede631005fd0e294f2516fc745 Mon Sep 17 00:00:00 2001
From: Google AI Agent <gemini@hermes.local>
Date: Mon, 6 Apr 2026 17:44:48 +0000
Subject: [PATCH 2/2] Implement "The Reflex Layer" for low-reasoning tasks

---
 tasks.py | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)

diff --git a/tasks.py b/tasks.py
index afc5f228..431dde8d 100644
--- a/tasks.py
+++ b/tasks.py
@@ -226,6 +226,23 @@ def hermes_local(prompt, model=None, caller_tag=None, toolsets=None):
         return None
     return result.get("response")
 
+def run_reflex_task(prompt, caller_tag):
+    """Force a task to run on the cheapest local model (The Reflex Layer).
+    
+    Use this for non-reasoning tasks like formatting, categorization, 
+    and simple status checks to save expensive context for coding.
+    """
+    return run_hermes_local(
+        prompt=prompt,
+        model="gemma2:2b",
+        caller_tag=f"reflex-{caller_tag}",
+        disable_all_tools=True,
+        skip_context_files=True,
+        skip_memory=True,
+        max_iterations=1,
+    )
+
+
 
 ARCHIVE_EPHEMERAL_SYSTEM_PROMPT = (
     "You are running a private archive-processing microtask for Timmy.\n"
-- 
2.43.0