bstein/titan-iac
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

atlasbot: add model fallback and rollout

Browse Source
This commit is contained in:
Brad Stein 2026-01-27 21:16:47 -03:00
parent 65e50d1923
commit 27e8a77044
2 changed files with 20 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
services/comms/atlasbot-deployment.yaml
View File

@ -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

10
services/comms/scripts/atlasbot/bot.py
View File

@ -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"))
@ -3132,9 +3133,18 @@ def _ollama_call(
lock = _OLLAMA_LOCK if OLLAMA_SERIALIZE else None
if lock:
lock.acquire()
try:
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:
raise
msg = data.get("message") if isinstance(data, dict) else None
if isinstance(msg, dict):
raw_reply = msg.get("content")