From 27e8a770448033a5e03227eee68206911b17eed3 Mon Sep 17 00:00:00 2001
From: Brad Stein <Brad.Stein@gmail.com>
Date: Tue, 27 Jan 2026 21:16:47 -0300
Subject: [PATCH] atlasbot: add model fallback and rollout

---
 services/comms/atlasbot-deployment.yaml |  4 +++-
 services/comms/scripts/atlasbot/bot.py  | 24 +++++++++++++++++-------
 2 files changed, 20 insertions(+), 8 deletions(-)

diff --git a/services/comms/atlasbot-deployment.yaml b/services/comms/atlasbot-deployment.yaml
index 97567eb..7414f1e 100644
--- a/services/comms/atlasbot-deployment.yaml
+++ b/services/comms/atlasbot-deployment.yaml
@@ -16,7 +16,7 @@ spec:
       labels:
         app: atlasbot
       annotations:
-        checksum/atlasbot-configmap: manual-atlasbot-71
+        checksum/atlasbot-configmap: manual-atlasbot-72
         vault.hashicorp.com/agent-inject: "true"
         vault.hashicorp.com/role: "comms"
         vault.hashicorp.com/agent-inject-secret-turn-secret: "kv/data/atlas/comms/turn-shared-secret"
@@ -83,6 +83,8 @@ spec:
               value: http://ollama.ai.svc.cluster.local:11434
             - name: OLLAMA_MODEL
               value: qwen2.5:14b-instruct
+            - name: OLLAMA_FALLBACK_MODEL
+              value: qwen2.5:14b-instruct-q4_0
             - name: OLLAMA_TIMEOUT_SEC
               value: "600"
             - name: ATLASBOT_THINKING_INTERVAL_SEC
diff --git a/services/comms/scripts/atlasbot/bot.py b/services/comms/scripts/atlasbot/bot.py
index 47458ea..2c93b75 100644
--- a/services/comms/scripts/atlasbot/bot.py
+++ b/services/comms/scripts/atlasbot/bot.py
@@ -17,6 +17,7 @@ ROOM_ALIAS = "#othrys:live.bstein.dev"
 
 OLLAMA_URL = os.environ.get("OLLAMA_URL", "https://chat.ai.bstein.dev/")
 MODEL = os.environ.get("OLLAMA_MODEL", "qwen2.5-coder:7b-instruct-q4_0")
+FALLBACK_MODEL = os.environ.get("OLLAMA_FALLBACK_MODEL", "")
 API_KEY = os.environ.get("CHAT_API_KEY", "")
 OLLAMA_TIMEOUT_SEC = float(os.environ.get("OLLAMA_TIMEOUT_SEC", "480"))
 ATLASBOT_HTTP_PORT = int(os.environ.get("ATLASBOT_HTTP_PORT", "8090"))
@@ -3133,14 +3134,23 @@ def _ollama_call(
     if lock:
         lock.acquire()
     try:
-        with request.urlopen(r, timeout=OLLAMA_TIMEOUT_SEC) as resp:
-            data = json.loads(resp.read().decode())
-            msg = data.get("message") if isinstance(data, dict) else None
-            if isinstance(msg, dict):
-                raw_reply = msg.get("content")
+        try:
+            with request.urlopen(r, timeout=OLLAMA_TIMEOUT_SEC) as resp:
+                data = json.loads(resp.read().decode())
+        except error.HTTPError as exc:
+            if exc.code == 404 and FALLBACK_MODEL and FALLBACK_MODEL != payload["model"]:
+                payload["model"] = FALLBACK_MODEL
+                r = request.Request(endpoint, data=json.dumps(payload).encode(), headers=headers)
+                with request.urlopen(r, timeout=OLLAMA_TIMEOUT_SEC) as resp:
+                    data = json.loads(resp.read().decode())
             else:
-                raw_reply = data.get("response") or data.get("reply") or data
-            reply = _normalize_reply(raw_reply) or "I'm here to help."
+                raise
+        msg = data.get("message") if isinstance(data, dict) else None
+        if isinstance(msg, dict):
+            raw_reply = msg.get("content")
+        else:
+            raw_reply = data.get("response") or data.get("reply") or data
+        reply = _normalize_reply(raw_reply) or "I'm here to help."
         if use_history:
             history[hist_key].append(f"Atlas: {reply}")
         return reply